research article investigation of antileishmanial activities of … · 2019. 7. 30. · research...
TRANSCRIPT
Research ArticleInvestigation of Antileishmanial Activities ofAcridines Derivatives against Promastigotes and AmastigotesForm of Parasites Using Quantitative Structure ActivityRelationship Analysis
Samir Chtita1 Mounir Ghamali1 Rachid Hmamouchi1 Bouhya Elidrissi1
Mohamed Bourass2 Majdouline Larif3 Mohammed Bouachrine4 and Tahar Lakhlifi1
1MCNSL Faculty of Science Moulay Ismail University Meknes Morocco2Faculty of Sciences Sidi Mohamed Ben Abdellah University Fez Morocco3Separation Process Laboratory Faculty of Science Ibn Tofail University Kenitra Morocco4High School of Technology Moulay Ismail University Meknes Morocco
Correspondence should be addressed to Samir Chtita samirchtitagmailcom
Received 13 August 2016 Revised 22 September 2016 Accepted 27 September 2016
Academic Editor Michael D Sevilla
Copyright copy 2016 Samir Chtita et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
In a search of newer and potent antileishmanial (against promastigotes and amastigotes form of parasites) drug a series of 60variously substituted acridines derivatives were subjected to a quantitative structure activity relationship (QSAR) analysis forstudying interpreting and predicting activities and designing new compounds by using multiple linear regression and artificialneural network (ANN) methods The used descriptors were computed with Gaussian 03 ACDChemSketch Marvin Sketch andChemOffice programs The QSAR models developed were validated according to the principles set up by the Organisation forEconomic Co-operation and Development (OECD) The principal component analysis (PCA) has been used to select descriptorsthat show a high correlation with activities The univariate partitioning (UP) method was used to divide the dataset into trainingand test sets The multiple linear regression (MLR) method showed a correlation coefficient of 0850 and 0814 for antileishmanialactivities against promastigotes and amastigotes forms of parasites respectively Internal and external validations were used todetermine the statistical quality of QSAR of the two MLR models The artificial neural network (ANN) method considering therelevant descriptors obtained from the MLR showed a correlation coefficient of 0933 and 0918 with 7-3-1 and 6-3-1 ANN modelsarchitecture for antileishmanial activities against promastigotes and amastigotes forms of parasites respectively The applicabilitydomain of MLR models was investigated using simple and leverage approaches to detect outliers and outsides compounds Theeffects of different descriptors in the activities were described and used to study and design new compounds with higher activitiescompared to the existing ones
1 Introduction
For hundreds of years Leishmaniasis a disease caused bya number of species of protozoan parasites belonging tothe genus Leishmania is recognized as an important publichealth problem throughout the world [1ndash3] Currently theleishmaniases are considered to be endemic in 88 countriesand according to World Health Organization (WHO) [4]twelve million people are infected with about two to three
million new cases each year and 350million people are underrisk of infection it is a major public health problem partic-ularly in Latin America Africa and Asia [5ndash9] To date novaccine against any clinical form of Leishmaniasis has beencommercialized and treatment relies only on chemotherapywhich has been based on the use of pentavalent antimo-nial drugs Other medications such as pentamidine andamphotericin B have been used as alternative drugs forresistant parasites With the emergence of some resistant
Hindawi Publishing CorporationAdvances in Physical ChemistryVolume 2016 Article ID 5137289 16 pageshttpdxdoiorg10115520165137289
2 Advances in Physical Chemistry
strains the toxicity of current drugs severe side effects andhigh cost andor restricted therapeutic spectrum a need fordevelopment of new and safer drugs is warranted [2 3 10]A great number of natural and synthetic compounds havebeen tested in the past years in antileishmanial assays Theirstructures are diverse and often contain nitrogen heterocyclessuch as quinolines pyrimidines acridines phenothiazinesand indoles [11ndash13]
Many experiments have been performed with the com-pounds bearing the heterocyclic ring structures to exploretheir effectiveness against Leishmania These studies sug-gested their similar pharmacophoric feature of the hetero-cyclic scaffold as a potential target for drug discovery ofantileishmanial drugs [14]
Leishmania parasites exist in two forms one is promastig-otes and the other is amastigotes The promastigotes areflagellated and found in sand fly while the amastigotes areovoid and nonflagellated form of Leishmania [15] Antileish-manial activity is performed against promastigotes and thenamastigotes forms of parasites Heterocyclic system may alsobe formed by fusion with other rings either carbocyclic orheterocyclic
Since their discovery in the 1880s acridines family havedemonstrated a broad spectrum of pharmacological prop-erties [16] First employed as antibacterial agents duringthe beginning of the twentieth century [17] They havebeen rapidly revealing interesting antiproliferative activitiesagainst both protozoa and tumor cells [18 19] Consequentlythey have been extensively used in antiparasitic chemother-apy and a wide range of new acridines derivatives have beensynthesized and successfully assessed for their antileishma-nial activities [20 21]
In order to open a new way in antileishmanial drugresearch a series of sixty acridines derivatives were synthe-sized [22ndash24] and studied for their antileishmanial (againstpromastigotes and amastigotes form of parasites) activitiesThe aim of this study was to develop a QSAR model able tocorrelate the structural features of the acridines derivativeswith their biological activities
In general the QSAR methods are based on the assump-tion that the activity of a certain chemical compound relatedto its structure through a certain mathematical algorithmThis relationship can be used in the prediction interpre-tation and assessment of new compounds with desiredactivities reducing and rationalizing time efforts and cost ofsynthesis and new product development
The basic assumption to drive a QSARmodel is presenteddue to a mathematical function of the chemical propertieswhich is related to the effect (activity) Therefore the effectis like the function ldquo119891rdquo of the chemical properties ldquo119909rdquo119910 = 119891(119909) To find this algorithm we use a number ofchemical compounds with known values of the studied effect(119910) For each chemical compound we calculate a series ofparameters (called chemical descriptors) Then we find analgorithm that provides a quite accurate value similar tothe real experimental value The final step is to check if theobtained algorithm is able to predict the activity values forother chemicals not used to build up the model (externalvalidation)
Indeed it is very important to generate a model whichworked not only for the chemical substances used withinthe training set but also for other similar chemicals Con-sequently the challenge is to define the correct statisticalproperties of the model
2 Materials and Methods
The current QSAR study investigates prediction and inter-pretation of the studied compounds and was also used fordesigning new proposed compounds by using linear andnonlinear methods It consists of four stages selection ofdataset and generation of molecular descriptors descriptiveanalysis statistical analysis (prediction and evaluation) andsuggestion of novel compounds
A flow chart for the development of the QSAR modelalong with the various validation methods used in this workis demonstrated in Figure 1
21 Selection of Dataset and Generation ofMolecular Descriptors
211 Selection of Dataset In this stage the datasets ofthe antileishmanial activities (against promastigotes andamastigotes forms of parasites) of various acridine derivatives(4-monosubstituted acridines 36-disubstituted acridines45-disubstituted acridines and 7-monosubstituted 9-chloroand 9-amino-2-methoxy acridines) were collected from pre-vious works [22ndash24] The molecular structures of the studiedmolecules with their antileishmanial activities are presentedin Table 1 All experimental IC50 antileishmanial activityvalues (120583M) were converted to the negative logarithm ofIC50 (pIC50 = minuslog
10(IC50))
212 Molecular Descriptors Generation A wide variety ofmolecular descriptors was calculated using Gaussian 03ACDChemSketch Marvin Sketch and ChemOffice pro-grams [25ndash28] to predict the correlation between thesedescriptors for the studied molecules with their antileishma-nial activities and to develop linear (multiple linear regression(MLR)) and nonlinear (artificial neural network (ANN))models Tables 3 and 4 show the selected descriptors (usingthe PCA method see more in descriptive analysis results) tobe used in this study
22 Descriptive Analysis In this stage the principal compo-nent analysis (PCA) was used to determine the nonlinearityand nonmulticollinearity among variables (descriptors) andto select descriptors that correlate with the activity Afterthat the univariate partitioning (UP) method was used toform dissimilar clusters of compounds to which the querycompounds would be compared for determining the degreeof similarity and dividing the dataset into training and testsets
23 Statistical Analysis (Models Development and Evaluation)In this stage linear and nonlinear QSAR models weredeveloped and evaluated to predict the activities of the testcompoundsThe study we conducted consists of the multiple
Advances in Physical Chemistry 3
N
1
Training set (promastigotes) N NHA NHD MTI
1 14330 4752 5141 3 2 82052 13080 4353 5023 3 2 108354 10070 4124 4469 3 2 255717 8800 0805 4167 5 2 355298 7820 0141 3989 7 2 45907
62 8310 3311 4511 5 2 39911
Obs25332204063406811207
1797Molecular descriptors
minus17minus12minus07minus02
0308131823
minus2 minus1 0 1 2
TrainingValidationTest
23
4
Activity = f (descriptors)
Preparation of the dataset
Calculation of molecular descriptors
Activity
Statistical analysis (QSAR models)
Internal and external validation of QSAR
models
Novel compounds with suggested
activities
N∘ 120583 120596 middot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middot
middot middot middot
pIC50
R5
R6
R7
R4
R1
R2
R3
Pred (pIC50 AMA)pIC50 AMA
Pred (pIC50 AMA)
pIC
50
AM
A
Figure 1 Flow chart of the methodology used in this work
linear regression (MLR) available in the XLSTAT softwareand the artificial neural network (ANN) available in theMatlab software
In order to propose models and to evaluate quanti-tatively the physicochemical effects of the substituents onthe activities of molecules we submitted the data matrixconstituted obviously from the used variables (descriptors)corresponding to the dataset molecules to a descendant MLRanalysis and to an ANN We use the coefficients 119877 1198772 1198772adjMSE and 119875 value to select the best regression performance[29] where119877 is the correlation coefficient1198772 is the coefficientof determination 1198772adj is the coefficient adjusted for degreesof freedom Mean squared error (MSE) is the standard errorof the coefficient of each descriptor and of the global modelwhich gives an indication of the valid inclusion of a descriptorin a QSAR model 119875 value is the probability (119875) of Fisherstatistics (119865) which gives an indication of the probability thata QSAR is a chance correlation
In order to assess the significance of the models andaccurate prediction ability for new compounds
(i) we use an internal validation procedure (leave-one-out cross validation) whereby one compound isremoved and the rebuilt model with the remainingmolecules is used to predict the response of theeliminated compound This one is then returned anda second is removed and the cycle is repeated and
so on until all compounds have been removed oneby one and an overall correlation coefficient 119877cv iscomputed
(ii) after the model is built an external prediction isnecessary In this one the obtained model was usedto predict the activities of a test set comprisingcompounds that are similar to those not used in thetraining set This is usually performed by splitting adataset into a training set and a test set typically ina 15 ratio Further before performing the externalvalidation of a model it is very important to check forthe presence of systematic error that violates the basicassumptions of the least squares regression model Ifhigh systematic error (bias) is present in the modelthen suchmodel should be discarded and performingany external validation test is of no use on such biasedmodel Xternal Validation Plus is a tool that checksthe presence of systematic errors in the model andfurther computes all the required external validationparameters while judging the performance of actualprediction quality of a QSARmodel based on recentlyproposed MAE-based criteria [30]
(iii) a model is valid only within its training domainand new molecules must be considered as belongingto the domain before the model is applied (OECDPrinciple 3 [31]) Without applicability domain (AD)
4 Advances in Physical Chemistry
Table1Ch
emicalstructurea
ndantileishmanialactivities
ofstu
died
compo
unds
(see
Figure
8)
Num
ber
R 1R 2
R 3R 4
R 5R 6
R 7pIC 50
b
PRO
cAMA
d
1H
NHCO
CH3
HH
HNHCO
CH3
Hminus2
533
minus0653
2H
NHCO
C 2H5
HH
HNHCO
C 2H5
Hminus2
204
minus0886
3H
NHCO
C 3H7
HH
HNHCO
C 3H7
Hminus1669
minus110
74
HNHCO
phH
HH
NHCO
PhH
minus0634
minus0041
5H
NHCO
-p-PhC
lH
HH
NHCO
-p-PhC
lH
mdash0699
6H
NHCO
-p-PhF
HH
HNHCO
-p-PhF
Hminus0
230
1523
7H
NHCO
-p-PhO
Me
HH
HNHCO
-p-PhO
Me
Hminus0
681
minus0041
8H
NHCO
-mp-Ph(OMe)2
HH
HNHCO
-mp-Ph(OMe)2
Hminus1207
0097
9H
NHCO
Me
HH
HNHCO
phH
minus2270
Toxa
10H
NHCO
Me
HH
HNHCO
-p-PhC
lH
minus1061
minus0462
11H
NHCO
Me
HH
HNHCO
-p-PhF
Hminus2
123
0174
12H
NHCO
Me
HH
HNHCO
-p-PhO
Me
Hminus1939
Toxa
13H
NHCO
Me
HH
HNHCO
-mp-Ph(OMe)2
Hmdash
minus0114
14H
HCH3
NH2
OMe
HH
0301
0398
15H
HCH2OH
NH2
OMe
HH
minus0732
minus0613
16H
HCH2Br
NH2
OMe
HH
minus0556
minus0663
17H
H(C
H2) 2OCO
OMe
NH2
OMe
HH
minus0380
minus0114
18H
H(C
H2) 2OCO
(CH2) 2CH3
NH2
OMe
HH
minus0491
minus039
819
HH
(CH2) 2OCO
CH2CH
(CH3) 2
NH2
OMe
HH
minus0342
Toxa
20H
H(C
H2) 2OCO
PhNH2
OMe
HH
0398
0699
21H
H(C
H2) 2OCO
PhF
NH2
OMe
HH
minus0114
0222
23H
H(C
H2) 2OCO
PhOMe
NH2
OMe
HH
minus0279
Toxa
24H
HCH3
ClOMe
HH
minus0041
minus0362
25H
HCH2OH
ClOMe
HH
minus2262
minus0959
26H
HCH2Br
ClOMe
HH
minus1703
minus117
027
HH
(CH2) 2OCO
OMe
ClOMe
HH
minus217
8minus1877
28H
H(C
H2) 2OCO
(CH2) 2CH3
ClOMe
HH
minus1707
minus1628
29H
H(C
H2) 2OCO
CH2CH
(CH3) 2
ClOMe
HH
minus1446
minus1561
30H
H(C
H2) 2OCO
PhCl
OMe
HH
minus213
5To
xa31
HH
(CH2) 2OCO
PhF
ClOMe
HH
minus2295
minus1645
32H
H(C
H2) 2OCO
PhCl
ClOMe
HH
minus219
4minus2
191
Advances in Physical Chemistry 5
Table1Con
tinued
Num
ber
R 1R 2
R 3R 4
R 5R 6
R 7pIC 50
b
PRO
cAMA
d
33H
H(C
H2) 2OCO
PhOMe
ClOMe
HH
minus2098
minus2088
34CH2NH2
HH
HH
HH
minus0230
minus053
135
CH2OH
HH
HH
HH
Toxa
minus1515
36CH2NHCO
(CH2) 3Cl
HH
HH
HH
minus1655
minus1427
37CH2OCO
(CH2) 3Cl
HH
HH
HH
minus2200
minus218
538
CH2NHCO
CH=C
H2
HH
HH
HH
minus2245
minus0833
39CH2OCO
CH=C
H2
HH
HH
HH
minus1464
mdash40
CH2NHCO
PhH
HH
HH
Hminus1654
minus112
141
CH2OCO
PhH
HH
HH
Hminus1885
Toxa
42CH2NHCO
-p-PhF
HH
HH
HH
minus1815
mdash43
CH2OCO
-p-PhF
HH
HH
HH
minus174
1To
xa44
CH2NHCO
-p-PhC
lH
HH
HH
HTo
xaminus1004
45CH2OCO
-p-PhC
lH
HH
HH
Hminus1790
minus1433
46CH2NHCO
-p-PhO
Me
HH
HH
HH
minus1513
minus0973
47CH2OCO
-p-PhO
Me
HH
HH
HH
minus1471
minus0869
48CH2NHCO
-p-PhN
Me 2
HH
HH
HH
minus153
9minus0
672
49CH2OCO
-p-PhN
Me 2
HH
HH
HH
minus1819
Toxa
50CH2NH2
HH
HH
HCH2NH2
minus0820
minus053
151
CH2OH
HH
HH
HCH2OH
Toxa
0222
52CH2NHCO
(CH2) 3Cl
HH
HH
HCH2NHCO
(CH2) 3Cl
minus0663
minus0813
54CH2NHCO
CH=C
H2
HH
HH
HCH2NHCO
CHCH2
minus1061
Toxa
56CH2NHCO
PhH
HH
HH
CH2NHCO
Phminus0
556
minus1236
57CH2OCO
PhH
HH
HH
CH2OCO
PhTo
xaminus0
716
58CH2NHCO
-p-PhF
HH
HH
HCH2NHCO
-p-PhF
minus0756
minus0255
60CH2NHCO
-p-PhC
lH
HH
HH
CH2NHCO
-p-PhC
lmdash
minus1562
62CH2NHCO
-p-PhO
Me
HH
HH
HCH2NHCO
-p-PhO
Me
minus1797
Toxa
63CH2OCO
-p-PhO
Me
HH
HH
HCH2OCO
-p-PhO
Me
Toxa
minus1825
64CH2NHCO
-p-PhN
Me 2
HH
HH
HCH2NHCO
-p-PhN
Me 2
minus0940
minus0724
65CH2OCO
-p-PhN
Me 2
HH
HH
HCH2OCO
-p-PhN
Me 2
Toxa
minus1667
a Toxictoxicity
observed
onhu
man
macroph
agesatconcentrations
thatdidno
tdisp
layantileishmanialactivity
-p-para-m
-meta
b pIC50=minuslog(
IC50)
c pIC50AMAantileish
manialactivity
againstamastig
otes
parasiteformdpIC 50PR
Oantileish
manialactivity
againstp
romastig
otes
parasiteform
6 Advances in Physical Chemistry
Inputw
b
w
b
+ +
Hidden layer Output layer
Output
Figure 2 The architecture used in our study of the artificial neural network
each model can predict the activity of any com-pound even with a completely different structurefrom those included in the study Therefore the ADis a tool to find out compounds that are outsidethe applicability domain of the built QSAR modeland it detects outliers present in the training setcompounds There are several methods for definingthe applicability domain (AD) of QSAR models [32]but themost common one is determining the leveragevalues ℎ119894 (ℎ119894 = 119909119879
119894(119883119879 119883)
minus1
119909119894 (119894 = 1 2 119899)) foreach compound where119909119894 is the descriptor row-vectorof query compound 119883 is 119899 119909 119896 minus 1 matrix of 119896model descriptor values for 119899 training set compoundsand the superscript 119879 refers to the transpose ofmatrixvector [32 33] In this study we use Williamsplot in this plot the applicability domain is estab-lished inside a squared area within standard deviationplusmn119909 (in this study 119909 = 3 ldquothree-sigma rulerdquo [34]) anda leverage threshold ℎlowast(ℎlowast = 3 lowast (119896 + 1)119899) [35]where 119899 is the number of training set compounds 119896is the number of model descriptors The leverage (ℎ)greater than the warning leverage (ℎlowast) suggested thatthe compoundwas very influential on themodel [36]The results of the leverage approach were comparedwith that of the simple approach introduced by Royet al [37]
231 Multiple Linear Regression (MLR) The descendentmultiple linear regression (MLR) analysis based on theelimination of aberrant descriptors (one by one) until a validmodel (including the critical probability 119875 value lt 005 forall descriptors and the model complete) was employed tofind a linear model of the activity of interest which takes thefollowing form
119884 = 1198860 +119899
sum119894=1
119886119894119909119894 (1)
where 119884 is the studied activity which is the dependent vari-able 1198860 is the intercept of the equation 119909119894 are the moleculardescriptors 119886119894 are the coefficients of those descriptors
This method is one of the most popular methods ofQSARQSPR thanks to its simplicity in operation repro-ducibility and ability to allow easy interpretation of thefeatures used The important advantage of the linear regres-sion analysis is that it is highly transparent therefore thealgorithm is available and predictions can bemade easily [38]It has served also to select the descriptors used as the inputparameters in the artificial neural network (ANN)
Table 2 Software packages used in this work
Drawing chemicalstructures
Marvin Sketch ACDChemSketchand ChemBioDraw
Generating 3D structures Gauss View 30 and ChemBio3DCalculating chemicaldescriptors
Gaussian 03 Marvin Sketch 62ChemSketch and ChemBio3D
Developing QSAR models XLSTAT 2009 and Matlab 790(version 2011)
232 Artificial Neural Networks (ANN) The artificial neuralnetworks (ANN) are used in order to increase the probabilityof characterizing the compounds and to generate a predic-tive QSAR model between the set of molecular descriptorsobtained from the MLR models and the observed activitiesvalues The ANN model is done on the MATLAB R2009bsoftware It consists of three layers of neurons called inputlayer hidden layer and output layer (Figure 2) The inputlayer formed by a number of neurons equal to the numberof descriptors obtained in the multiple linear regressionmodels and the output layer represents the calculated activityvalues For determination of the number of hidden neuronsin the hidden layer where all calculations of parameteroptimization of neural networks are made a parameter 120588has been proposed The parameter 120588 plays a major role indetermining the best artificial neural network architecture[39 40] 120588 is defined as follows
120588 =number of data points in the training set
sum of the number of connections in the ANN (2)
In order to avoid overfitting or underfitting it is recom-mended that the value of 120588 should be between 100 and 300if 120588 lt 1 the network simply memorizes the data whereas if120588 gt 3 the network is not able to generalize [41]
24 Software Packages Used in Our QSARDevelopment StudyThere are various free and commercial software available forQSAR development These include specialized software fordrawing chemical structures generating 3D structures cal-culating chemical descriptors and developing QSARmodelsThe software packages used in our works are represented inTable 2
3 Results and Discussions
31 Principal Component Analysis (PCA) PCA is a usefulstatistical technique for summarizing all the informationencoded in the structures of the compounds It is very helpful
Advances in Physical Chemistry 7
Table 3The absolute values of correlation coefficients with descrip-tors and activitiesDescriptors 119903Promastigotes
Henryrsquos law constant (119870H) 0130Ideal gas thermal capacity (IGTC) 0108Number of H-Bond acceptors (NHA) 0299Number of H-Bond donors (NHD) 0250Balaban index (119869) 0133Shape coefficient (119868) 0135Total valence connectivity (TVC) 0292N 0238H 0260Gibbs free energy (119866) 0119Dipole moment (120583) 0213Electronegativity (120594) 0599Electrophilicity index (120596) 0630Molecular topological index (MTI) 0116Polar surface area (PSA) 0282Total connectivity (TC) 0269Wiener index (119882) 0124119864HOMO 0554119864LUMO 0632119864Gap 0227
AmastigotesTotal energy 119864 0160119864HOMO 0286Electronegativity (120594) 0316Polar surface area (PSA) 0391O 0105Index of refraction (119899) 0548Density (119889) 0242Critical pressure (CP) 0176Gibbs free energy (119866) 0195Henryrsquos law constant (119870H) 0387Number of H-Bond acceptors (NHA) 0198Number of H-Bond donors (NHD) 0587Number of rotatable bonds (NRB) 0139Sum of valence degrees (SVD) 0146Dipole moment (120583) 0224119864LUMO 0338Electrophilicity index (120596) 0335N 0436C 0155Surface tension (120574) 0553Boiling point (TB) 0254Critical temperature (CT) 0276Heat of formation (119867∘) 0152log119875 0248Melting point (119879) 0314Partition coefficient (PC) 0163Shape coefficient (119868) 0176Winner index (119882) 0103
for understanding the distribution of the compounds Thisis an essentially descriptive statistical method which aimsto extract the maximum of information contained in thedataset compounds [42 43] In this work the PCA is usedto overview the examined compounds for similarities anddissimilarities and to select descriptors that show a high
Table 4 Descriptors selected by the principal component analysis(PCA) and software packages used in the calculation of descriptors
Software Descriptors
Gaussian 03
Highest occupied molecular orbital energy119864HOMO (eV) lowest unoccupied molecularorbital energy 119864LUMO (eV) hardness 120578 (eV)= (119864LUMO minus 119864HOMO)2 electronegativity 120594(eV) = minus(119864LUMO + 119864HOMO)2 electrophilicityindex 120596 (eV) = 12059422120578 total energy 119864 (eV)dipole moment 120583 (Debye) energy gapbetween 119864HOMO and 119864LUMO values 119864Gap (eV)
ChemOffice
Heat of formation119867∘ (kJmolminus1) Gibbs freeenergy 119866 (kJmolminus1) ideal gas thermalcapacity (IGTC) (Jmolminus1 Kminus1) melting point119879 (Kelvin) critical temperature (CT)(Kelvin) boiling point (TB) (Kelvin) criticalpressure (CP) (Bar) Henryrsquos law constant119870H total valence connectivity (TVC)partition coefficient (PC) moleculartopological index (MTI) number ofrotatable bonds (NRB) shape coefficient 119868sum of valence degrees (SVD) totalconnectivity (TC)
ChemSketchPercent ratios of nitrogen hydrogenoxygen and carbon atoms (N H Oand C) surface tension 120574 (dynecm) indexof refraction (119899) density (119889)
Marvin Sketchlog119875 Winner index (119882) number ofH-Bond acceptors (NHA) number ofH-Bond donors (NHD) Balaban index (119869)polar surface area (PSA) (119860∘)2
correlation with the response activity this one gives extraweight because it will be more effective at prediction [44]Tables 3 and 4 present the descriptors with a correlationcoefficient with the activity higher in absolute value than01 The absence of any serious multicollinearity betweenthe descriptors present in the model was confirmed by thecorrelation matrix
32 Univariate Partitioning (UP) The aim of the UP was therecognition of groups of objects based on their similarity thismethod is based on the criterion of partitioning proposedby Fisher [45] In this study the division of the dataset intotraining and test sets has been performed In this one fromeach obtained cluster one compound for the training setwas selected randomly to be used as test set compound Thepartitioning results are given in Table 5
33 Multiple Linear Regression (MLR) The QSAR analysiswas performed using the values of the chemical descriptorsselected by the PCA method and on the other hand theexperimental values of the antileishmanial activities for 60 ofthe acridines derivatives (effect) Table S1 in SupplementaryMaterial available online at httpdxdoiorg10115520165137289 shows the value of each molecular descriptor that isconfigured in established MLR models and the QSAR model
8 Advances in Physical Chemistry
Table 5 The partitioning results of compounds into training and test sets
ClassesThe test set compounds
Promastigotes Amastigotes3 6 10 14 17 21 37 38 42 64 5 6 26 27 29 36 37 44 46 50
1 1 8 37 40 56 1 10 15 16 34 48 50 57 642 2 14 29 34 48 2 25 38 44 47 51 523 3 18 19 3 26 40 564 4 24 25 27 30 42 58 4 7 8 11 13 17 21 465 6 26 5 14 206 7 9 12 21 33 39 41 47 67 10 11 16 31 32 43 45 18 29 588 15 17 23 28 36 49 52 54 27 639 20 38 46 62 24 28 31 35 36 45 60 6510 50 64 32 33 37
built is represented by the following equations and the valuesof the statistical parameters
MLRModel for pIC50 Promastigotes
pIC50(PRO) = 3030 minus 8214 10minus02N + 0239120583
minus 1264120596 minus 0233NHA + 0732NHD
+ 3311 10minus05MTI + 3184 104TVC
(3)
Statistical Parameters 1198772 = 0723 119877 = 0850 1198772adj = 0664MSE = 0189 P value lt 10minus4
MLRModel for pIC50 Amastigotes
pIC50(AMA) = minus7314 + 1191119864HOMO
+ 2543 10minus02CT minus 0350119870H
minus 1632 10minus02119879 + 0652NHA
+ 1262NHD
(4)
Statistical Parameters 1198772 = 0663 119877 = 0814 1198772adj = 0598MSE = 0208 119875 value lt 10minus4
(i) For the two models P value is lower than 00001it means that we would be taking a lower than001 risk in assuming that the null hypothesis (noeffect of the explanatory variables) is wrong and thatthe regressions equations have statistical significanceTherefore we can conclude with confidence that theselected variables do bring a significant amount ofinformation
(ii) Higher value of 1198772 and 1198772adj and lower mean squarederror (MSE) indicate that the two proposed modelsare predictive and reliable
(iii) Our obtained models were validated internally by theleave-one-out cross validation technique the crossvalidation coefficient 1198772cv for the two models was
determined based on the predictive ability of themodel The value of 1198772cv is higher than 05 (1198772cv =0536 for the MLR pIC
50(PRO) model and 1198772cv =
0525 for the MLR pIC50
(AMA) model) indicatinggood predictability of the model
(iv) True predictive power of these models is to test theirability to predict perfectly pIC
50of compounds from
an external test set (compounds that were not usedfor the developed model) pIC
50of the remaining set
of 10 compounds are deduced from the quantitativemodels proposed with the compounds used in train-ing set by MLR These models will be able to predictthe activities of test set molecules in agreement withthe experimentally determined value The observedand calculated pIC
50values are given in Table 6
The predictive capacity of the models was judgedthe higher value of 1198772test (1198772test = 0660 for theMLR pIC
50(PRO) model and 1198772test = 0718 for the
MLR pIC50
(AMA) model) indicates the improvedpredictability of the model
(v) Xternal Validation Plus indicates the absence of sys-tematic errors in the model and a moderate perfor-mance of prediction quality of a QSAR model basedon proposed MAE-based criteria (Table S2)
(vi) The values of calculated activities from (3) and (4) aregiven in Table S1 and the correlations of calculatedand observed activities values are illustrated in Fig-ure 3
In the first model (pIC50
PRO) the descriptors influencingnegatively the activities are the percent ratio of nitrogen(N) the number of H-Bond acceptors (NHA) and theelectrophilicity index (120596) and the parameters influencingpositively the activities are the dipole moment (120583) thenumber ofH-Bonddonors (NHD) themolecular topologicalindex (MTI) and the total valence connectivity (TVC)
(i) The number of H-Bond acceptors (NHA) has a nega-tive sign in the model which suggests that increasedactivity can be achieved by decreasing the number
Advances in Physical Chemistry 9
Table 6 The values of chemical descriptors and the observed and predicted activities using the MLR models for the test set
Test set (promastigotes) pIC50PRO
Number N 120583 120596 NHA NHD MTI TVC Obs MLR3 12030 4259 5024 3 2 14075 3891 10minus07 minus1669 minus20476 9270 5640 4654 5 2 28763 1716 10minus10 minus0230 minus101410 10780 5730 5267 3 2 16459 4902 10minus08 minus1061 minus183214 11760 2698 3062 3 1 4387 3050 10minus05 0301 minus001217 8580 4969 3431 4 1 10329 5188 10minus07 minus0380 minus066421 7180 5214 3498 5 1 18600 1334 10minus08 minus0114 minus055137 4460 6939 4779 2 0 8245 4076 10minus06 minus2200 minus178138 10680 5122 4374 2 1 6388 5083 10minus06 minus2245 minus151242 8480 5260 4457 3 1 12005 1307 10minus07 minus1815 minus160864 13170 5297 4388 5 2 47653 1201 10minus10 minus0940 minus0455
Test set (amastigotes) pIC50AMA
Number 119864HOMO 119870H 119879 NHA NHD CT Obs MLR5 minus5850 106461 1300866 3 2 19110 0699 minus07876 minus5785 100595 1247456 5 2 18715 1523 033226 minus5665 583700 926651 2 0 7815 minus1170 minus145627 minus5603 622270 930050 3 0 8827 minus1877 minus162829 minus5640 618850 938380 3 0 9007 minus1561 minus146736 minus5966 676650 1037977 2 1 12201 minus1427 minus077437 minus6025 552770 930967 2 0 8348 minus2185 minus145744 minus5919 749400 1083064 2 1 12828 minus1004 minus097846 minus5790 752980 1078419 3 1 13925 minus0973 minus073350 minus5461 625750 913563 3 2 13449 minus0531 minus1028
minus05
0
05
1
15
2
25
minus02 03 08 13 18 23
ActivesValidation
ActivesValidation
minus16
minus06
04
14
minus05 0 05 1 15
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 3 Correlations of observed and predicted activities (training set in black and test set in red) values calculated using MLR models
of heteroatoms (nitrogen or oxygen atoms) mostlythe nitrogen ones for decreasing the ratio of nitrogen(N) having also a negative sign in the model
(ii) The electrophilicity index 120596 has a negative sign inthe model which suggests that increased activity canbe achieved by decreasing the electrophilicity of the
acridine derivatives (a high value of electrophilicitydescribes a good electrophile while a small value ofelectrophilicity describes a good nucleophile)
(iii) The dipolemoment 120583 has a positive sign in themodelwhich suggests that increased activity can be achievedby increasing the polarity of the acridine derivatives
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
2 Advances in Physical Chemistry
strains the toxicity of current drugs severe side effects andhigh cost andor restricted therapeutic spectrum a need fordevelopment of new and safer drugs is warranted [2 3 10]A great number of natural and synthetic compounds havebeen tested in the past years in antileishmanial assays Theirstructures are diverse and often contain nitrogen heterocyclessuch as quinolines pyrimidines acridines phenothiazinesand indoles [11ndash13]
Many experiments have been performed with the com-pounds bearing the heterocyclic ring structures to exploretheir effectiveness against Leishmania These studies sug-gested their similar pharmacophoric feature of the hetero-cyclic scaffold as a potential target for drug discovery ofantileishmanial drugs [14]
Leishmania parasites exist in two forms one is promastig-otes and the other is amastigotes The promastigotes areflagellated and found in sand fly while the amastigotes areovoid and nonflagellated form of Leishmania [15] Antileish-manial activity is performed against promastigotes and thenamastigotes forms of parasites Heterocyclic system may alsobe formed by fusion with other rings either carbocyclic orheterocyclic
Since their discovery in the 1880s acridines family havedemonstrated a broad spectrum of pharmacological prop-erties [16] First employed as antibacterial agents duringthe beginning of the twentieth century [17] They havebeen rapidly revealing interesting antiproliferative activitiesagainst both protozoa and tumor cells [18 19] Consequentlythey have been extensively used in antiparasitic chemother-apy and a wide range of new acridines derivatives have beensynthesized and successfully assessed for their antileishma-nial activities [20 21]
In order to open a new way in antileishmanial drugresearch a series of sixty acridines derivatives were synthe-sized [22ndash24] and studied for their antileishmanial (againstpromastigotes and amastigotes form of parasites) activitiesThe aim of this study was to develop a QSAR model able tocorrelate the structural features of the acridines derivativeswith their biological activities
In general the QSAR methods are based on the assump-tion that the activity of a certain chemical compound relatedto its structure through a certain mathematical algorithmThis relationship can be used in the prediction interpre-tation and assessment of new compounds with desiredactivities reducing and rationalizing time efforts and cost ofsynthesis and new product development
The basic assumption to drive a QSARmodel is presenteddue to a mathematical function of the chemical propertieswhich is related to the effect (activity) Therefore the effectis like the function ldquo119891rdquo of the chemical properties ldquo119909rdquo119910 = 119891(119909) To find this algorithm we use a number ofchemical compounds with known values of the studied effect(119910) For each chemical compound we calculate a series ofparameters (called chemical descriptors) Then we find analgorithm that provides a quite accurate value similar tothe real experimental value The final step is to check if theobtained algorithm is able to predict the activity values forother chemicals not used to build up the model (externalvalidation)
Indeed it is very important to generate a model whichworked not only for the chemical substances used withinthe training set but also for other similar chemicals Con-sequently the challenge is to define the correct statisticalproperties of the model
2 Materials and Methods
The current QSAR study investigates prediction and inter-pretation of the studied compounds and was also used fordesigning new proposed compounds by using linear andnonlinear methods It consists of four stages selection ofdataset and generation of molecular descriptors descriptiveanalysis statistical analysis (prediction and evaluation) andsuggestion of novel compounds
A flow chart for the development of the QSAR modelalong with the various validation methods used in this workis demonstrated in Figure 1
21 Selection of Dataset and Generation ofMolecular Descriptors
211 Selection of Dataset In this stage the datasets ofthe antileishmanial activities (against promastigotes andamastigotes forms of parasites) of various acridine derivatives(4-monosubstituted acridines 36-disubstituted acridines45-disubstituted acridines and 7-monosubstituted 9-chloroand 9-amino-2-methoxy acridines) were collected from pre-vious works [22ndash24] The molecular structures of the studiedmolecules with their antileishmanial activities are presentedin Table 1 All experimental IC50 antileishmanial activityvalues (120583M) were converted to the negative logarithm ofIC50 (pIC50 = minuslog
10(IC50))
212 Molecular Descriptors Generation A wide variety ofmolecular descriptors was calculated using Gaussian 03ACDChemSketch Marvin Sketch and ChemOffice pro-grams [25ndash28] to predict the correlation between thesedescriptors for the studied molecules with their antileishma-nial activities and to develop linear (multiple linear regression(MLR)) and nonlinear (artificial neural network (ANN))models Tables 3 and 4 show the selected descriptors (usingthe PCA method see more in descriptive analysis results) tobe used in this study
22 Descriptive Analysis In this stage the principal compo-nent analysis (PCA) was used to determine the nonlinearityand nonmulticollinearity among variables (descriptors) andto select descriptors that correlate with the activity Afterthat the univariate partitioning (UP) method was used toform dissimilar clusters of compounds to which the querycompounds would be compared for determining the degreeof similarity and dividing the dataset into training and testsets
23 Statistical Analysis (Models Development and Evaluation)In this stage linear and nonlinear QSAR models weredeveloped and evaluated to predict the activities of the testcompoundsThe study we conducted consists of the multiple
Advances in Physical Chemistry 3
N
1
Training set (promastigotes) N NHA NHD MTI
1 14330 4752 5141 3 2 82052 13080 4353 5023 3 2 108354 10070 4124 4469 3 2 255717 8800 0805 4167 5 2 355298 7820 0141 3989 7 2 45907
62 8310 3311 4511 5 2 39911
Obs25332204063406811207
1797Molecular descriptors
minus17minus12minus07minus02
0308131823
minus2 minus1 0 1 2
TrainingValidationTest
23
4
Activity = f (descriptors)
Preparation of the dataset
Calculation of molecular descriptors
Activity
Statistical analysis (QSAR models)
Internal and external validation of QSAR
models
Novel compounds with suggested
activities
N∘ 120583 120596 middot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middot
middot middot middot
pIC50
R5
R6
R7
R4
R1
R2
R3
Pred (pIC50 AMA)pIC50 AMA
Pred (pIC50 AMA)
pIC
50
AM
A
Figure 1 Flow chart of the methodology used in this work
linear regression (MLR) available in the XLSTAT softwareand the artificial neural network (ANN) available in theMatlab software
In order to propose models and to evaluate quanti-tatively the physicochemical effects of the substituents onthe activities of molecules we submitted the data matrixconstituted obviously from the used variables (descriptors)corresponding to the dataset molecules to a descendant MLRanalysis and to an ANN We use the coefficients 119877 1198772 1198772adjMSE and 119875 value to select the best regression performance[29] where119877 is the correlation coefficient1198772 is the coefficientof determination 1198772adj is the coefficient adjusted for degreesof freedom Mean squared error (MSE) is the standard errorof the coefficient of each descriptor and of the global modelwhich gives an indication of the valid inclusion of a descriptorin a QSAR model 119875 value is the probability (119875) of Fisherstatistics (119865) which gives an indication of the probability thata QSAR is a chance correlation
In order to assess the significance of the models andaccurate prediction ability for new compounds
(i) we use an internal validation procedure (leave-one-out cross validation) whereby one compound isremoved and the rebuilt model with the remainingmolecules is used to predict the response of theeliminated compound This one is then returned anda second is removed and the cycle is repeated and
so on until all compounds have been removed oneby one and an overall correlation coefficient 119877cv iscomputed
(ii) after the model is built an external prediction isnecessary In this one the obtained model was usedto predict the activities of a test set comprisingcompounds that are similar to those not used in thetraining set This is usually performed by splitting adataset into a training set and a test set typically ina 15 ratio Further before performing the externalvalidation of a model it is very important to check forthe presence of systematic error that violates the basicassumptions of the least squares regression model Ifhigh systematic error (bias) is present in the modelthen suchmodel should be discarded and performingany external validation test is of no use on such biasedmodel Xternal Validation Plus is a tool that checksthe presence of systematic errors in the model andfurther computes all the required external validationparameters while judging the performance of actualprediction quality of a QSARmodel based on recentlyproposed MAE-based criteria [30]
(iii) a model is valid only within its training domainand new molecules must be considered as belongingto the domain before the model is applied (OECDPrinciple 3 [31]) Without applicability domain (AD)
4 Advances in Physical Chemistry
Table1Ch
emicalstructurea
ndantileishmanialactivities
ofstu
died
compo
unds
(see
Figure
8)
Num
ber
R 1R 2
R 3R 4
R 5R 6
R 7pIC 50
b
PRO
cAMA
d
1H
NHCO
CH3
HH
HNHCO
CH3
Hminus2
533
minus0653
2H
NHCO
C 2H5
HH
HNHCO
C 2H5
Hminus2
204
minus0886
3H
NHCO
C 3H7
HH
HNHCO
C 3H7
Hminus1669
minus110
74
HNHCO
phH
HH
NHCO
PhH
minus0634
minus0041
5H
NHCO
-p-PhC
lH
HH
NHCO
-p-PhC
lH
mdash0699
6H
NHCO
-p-PhF
HH
HNHCO
-p-PhF
Hminus0
230
1523
7H
NHCO
-p-PhO
Me
HH
HNHCO
-p-PhO
Me
Hminus0
681
minus0041
8H
NHCO
-mp-Ph(OMe)2
HH
HNHCO
-mp-Ph(OMe)2
Hminus1207
0097
9H
NHCO
Me
HH
HNHCO
phH
minus2270
Toxa
10H
NHCO
Me
HH
HNHCO
-p-PhC
lH
minus1061
minus0462
11H
NHCO
Me
HH
HNHCO
-p-PhF
Hminus2
123
0174
12H
NHCO
Me
HH
HNHCO
-p-PhO
Me
Hminus1939
Toxa
13H
NHCO
Me
HH
HNHCO
-mp-Ph(OMe)2
Hmdash
minus0114
14H
HCH3
NH2
OMe
HH
0301
0398
15H
HCH2OH
NH2
OMe
HH
minus0732
minus0613
16H
HCH2Br
NH2
OMe
HH
minus0556
minus0663
17H
H(C
H2) 2OCO
OMe
NH2
OMe
HH
minus0380
minus0114
18H
H(C
H2) 2OCO
(CH2) 2CH3
NH2
OMe
HH
minus0491
minus039
819
HH
(CH2) 2OCO
CH2CH
(CH3) 2
NH2
OMe
HH
minus0342
Toxa
20H
H(C
H2) 2OCO
PhNH2
OMe
HH
0398
0699
21H
H(C
H2) 2OCO
PhF
NH2
OMe
HH
minus0114
0222
23H
H(C
H2) 2OCO
PhOMe
NH2
OMe
HH
minus0279
Toxa
24H
HCH3
ClOMe
HH
minus0041
minus0362
25H
HCH2OH
ClOMe
HH
minus2262
minus0959
26H
HCH2Br
ClOMe
HH
minus1703
minus117
027
HH
(CH2) 2OCO
OMe
ClOMe
HH
minus217
8minus1877
28H
H(C
H2) 2OCO
(CH2) 2CH3
ClOMe
HH
minus1707
minus1628
29H
H(C
H2) 2OCO
CH2CH
(CH3) 2
ClOMe
HH
minus1446
minus1561
30H
H(C
H2) 2OCO
PhCl
OMe
HH
minus213
5To
xa31
HH
(CH2) 2OCO
PhF
ClOMe
HH
minus2295
minus1645
32H
H(C
H2) 2OCO
PhCl
ClOMe
HH
minus219
4minus2
191
Advances in Physical Chemistry 5
Table1Con
tinued
Num
ber
R 1R 2
R 3R 4
R 5R 6
R 7pIC 50
b
PRO
cAMA
d
33H
H(C
H2) 2OCO
PhOMe
ClOMe
HH
minus2098
minus2088
34CH2NH2
HH
HH
HH
minus0230
minus053
135
CH2OH
HH
HH
HH
Toxa
minus1515
36CH2NHCO
(CH2) 3Cl
HH
HH
HH
minus1655
minus1427
37CH2OCO
(CH2) 3Cl
HH
HH
HH
minus2200
minus218
538
CH2NHCO
CH=C
H2
HH
HH
HH
minus2245
minus0833
39CH2OCO
CH=C
H2
HH
HH
HH
minus1464
mdash40
CH2NHCO
PhH
HH
HH
Hminus1654
minus112
141
CH2OCO
PhH
HH
HH
Hminus1885
Toxa
42CH2NHCO
-p-PhF
HH
HH
HH
minus1815
mdash43
CH2OCO
-p-PhF
HH
HH
HH
minus174
1To
xa44
CH2NHCO
-p-PhC
lH
HH
HH
HTo
xaminus1004
45CH2OCO
-p-PhC
lH
HH
HH
Hminus1790
minus1433
46CH2NHCO
-p-PhO
Me
HH
HH
HH
minus1513
minus0973
47CH2OCO
-p-PhO
Me
HH
HH
HH
minus1471
minus0869
48CH2NHCO
-p-PhN
Me 2
HH
HH
HH
minus153
9minus0
672
49CH2OCO
-p-PhN
Me 2
HH
HH
HH
minus1819
Toxa
50CH2NH2
HH
HH
HCH2NH2
minus0820
minus053
151
CH2OH
HH
HH
HCH2OH
Toxa
0222
52CH2NHCO
(CH2) 3Cl
HH
HH
HCH2NHCO
(CH2) 3Cl
minus0663
minus0813
54CH2NHCO
CH=C
H2
HH
HH
HCH2NHCO
CHCH2
minus1061
Toxa
56CH2NHCO
PhH
HH
HH
CH2NHCO
Phminus0
556
minus1236
57CH2OCO
PhH
HH
HH
CH2OCO
PhTo
xaminus0
716
58CH2NHCO
-p-PhF
HH
HH
HCH2NHCO
-p-PhF
minus0756
minus0255
60CH2NHCO
-p-PhC
lH
HH
HH
CH2NHCO
-p-PhC
lmdash
minus1562
62CH2NHCO
-p-PhO
Me
HH
HH
HCH2NHCO
-p-PhO
Me
minus1797
Toxa
63CH2OCO
-p-PhO
Me
HH
HH
HCH2OCO
-p-PhO
Me
Toxa
minus1825
64CH2NHCO
-p-PhN
Me 2
HH
HH
HCH2NHCO
-p-PhN
Me 2
minus0940
minus0724
65CH2OCO
-p-PhN
Me 2
HH
HH
HCH2OCO
-p-PhN
Me 2
Toxa
minus1667
a Toxictoxicity
observed
onhu
man
macroph
agesatconcentrations
thatdidno
tdisp
layantileishmanialactivity
-p-para-m
-meta
b pIC50=minuslog(
IC50)
c pIC50AMAantileish
manialactivity
againstamastig
otes
parasiteformdpIC 50PR
Oantileish
manialactivity
againstp
romastig
otes
parasiteform
6 Advances in Physical Chemistry
Inputw
b
w
b
+ +
Hidden layer Output layer
Output
Figure 2 The architecture used in our study of the artificial neural network
each model can predict the activity of any com-pound even with a completely different structurefrom those included in the study Therefore the ADis a tool to find out compounds that are outsidethe applicability domain of the built QSAR modeland it detects outliers present in the training setcompounds There are several methods for definingthe applicability domain (AD) of QSAR models [32]but themost common one is determining the leveragevalues ℎ119894 (ℎ119894 = 119909119879
119894(119883119879 119883)
minus1
119909119894 (119894 = 1 2 119899)) foreach compound where119909119894 is the descriptor row-vectorof query compound 119883 is 119899 119909 119896 minus 1 matrix of 119896model descriptor values for 119899 training set compoundsand the superscript 119879 refers to the transpose ofmatrixvector [32 33] In this study we use Williamsplot in this plot the applicability domain is estab-lished inside a squared area within standard deviationplusmn119909 (in this study 119909 = 3 ldquothree-sigma rulerdquo [34]) anda leverage threshold ℎlowast(ℎlowast = 3 lowast (119896 + 1)119899) [35]where 119899 is the number of training set compounds 119896is the number of model descriptors The leverage (ℎ)greater than the warning leverage (ℎlowast) suggested thatthe compoundwas very influential on themodel [36]The results of the leverage approach were comparedwith that of the simple approach introduced by Royet al [37]
231 Multiple Linear Regression (MLR) The descendentmultiple linear regression (MLR) analysis based on theelimination of aberrant descriptors (one by one) until a validmodel (including the critical probability 119875 value lt 005 forall descriptors and the model complete) was employed tofind a linear model of the activity of interest which takes thefollowing form
119884 = 1198860 +119899
sum119894=1
119886119894119909119894 (1)
where 119884 is the studied activity which is the dependent vari-able 1198860 is the intercept of the equation 119909119894 are the moleculardescriptors 119886119894 are the coefficients of those descriptors
This method is one of the most popular methods ofQSARQSPR thanks to its simplicity in operation repro-ducibility and ability to allow easy interpretation of thefeatures used The important advantage of the linear regres-sion analysis is that it is highly transparent therefore thealgorithm is available and predictions can bemade easily [38]It has served also to select the descriptors used as the inputparameters in the artificial neural network (ANN)
Table 2 Software packages used in this work
Drawing chemicalstructures
Marvin Sketch ACDChemSketchand ChemBioDraw
Generating 3D structures Gauss View 30 and ChemBio3DCalculating chemicaldescriptors
Gaussian 03 Marvin Sketch 62ChemSketch and ChemBio3D
Developing QSAR models XLSTAT 2009 and Matlab 790(version 2011)
232 Artificial Neural Networks (ANN) The artificial neuralnetworks (ANN) are used in order to increase the probabilityof characterizing the compounds and to generate a predic-tive QSAR model between the set of molecular descriptorsobtained from the MLR models and the observed activitiesvalues The ANN model is done on the MATLAB R2009bsoftware It consists of three layers of neurons called inputlayer hidden layer and output layer (Figure 2) The inputlayer formed by a number of neurons equal to the numberof descriptors obtained in the multiple linear regressionmodels and the output layer represents the calculated activityvalues For determination of the number of hidden neuronsin the hidden layer where all calculations of parameteroptimization of neural networks are made a parameter 120588has been proposed The parameter 120588 plays a major role indetermining the best artificial neural network architecture[39 40] 120588 is defined as follows
120588 =number of data points in the training set
sum of the number of connections in the ANN (2)
In order to avoid overfitting or underfitting it is recom-mended that the value of 120588 should be between 100 and 300if 120588 lt 1 the network simply memorizes the data whereas if120588 gt 3 the network is not able to generalize [41]
24 Software Packages Used in Our QSARDevelopment StudyThere are various free and commercial software available forQSAR development These include specialized software fordrawing chemical structures generating 3D structures cal-culating chemical descriptors and developing QSARmodelsThe software packages used in our works are represented inTable 2
3 Results and Discussions
31 Principal Component Analysis (PCA) PCA is a usefulstatistical technique for summarizing all the informationencoded in the structures of the compounds It is very helpful
Advances in Physical Chemistry 7
Table 3The absolute values of correlation coefficients with descrip-tors and activitiesDescriptors 119903Promastigotes
Henryrsquos law constant (119870H) 0130Ideal gas thermal capacity (IGTC) 0108Number of H-Bond acceptors (NHA) 0299Number of H-Bond donors (NHD) 0250Balaban index (119869) 0133Shape coefficient (119868) 0135Total valence connectivity (TVC) 0292N 0238H 0260Gibbs free energy (119866) 0119Dipole moment (120583) 0213Electronegativity (120594) 0599Electrophilicity index (120596) 0630Molecular topological index (MTI) 0116Polar surface area (PSA) 0282Total connectivity (TC) 0269Wiener index (119882) 0124119864HOMO 0554119864LUMO 0632119864Gap 0227
AmastigotesTotal energy 119864 0160119864HOMO 0286Electronegativity (120594) 0316Polar surface area (PSA) 0391O 0105Index of refraction (119899) 0548Density (119889) 0242Critical pressure (CP) 0176Gibbs free energy (119866) 0195Henryrsquos law constant (119870H) 0387Number of H-Bond acceptors (NHA) 0198Number of H-Bond donors (NHD) 0587Number of rotatable bonds (NRB) 0139Sum of valence degrees (SVD) 0146Dipole moment (120583) 0224119864LUMO 0338Electrophilicity index (120596) 0335N 0436C 0155Surface tension (120574) 0553Boiling point (TB) 0254Critical temperature (CT) 0276Heat of formation (119867∘) 0152log119875 0248Melting point (119879) 0314Partition coefficient (PC) 0163Shape coefficient (119868) 0176Winner index (119882) 0103
for understanding the distribution of the compounds Thisis an essentially descriptive statistical method which aimsto extract the maximum of information contained in thedataset compounds [42 43] In this work the PCA is usedto overview the examined compounds for similarities anddissimilarities and to select descriptors that show a high
Table 4 Descriptors selected by the principal component analysis(PCA) and software packages used in the calculation of descriptors
Software Descriptors
Gaussian 03
Highest occupied molecular orbital energy119864HOMO (eV) lowest unoccupied molecularorbital energy 119864LUMO (eV) hardness 120578 (eV)= (119864LUMO minus 119864HOMO)2 electronegativity 120594(eV) = minus(119864LUMO + 119864HOMO)2 electrophilicityindex 120596 (eV) = 12059422120578 total energy 119864 (eV)dipole moment 120583 (Debye) energy gapbetween 119864HOMO and 119864LUMO values 119864Gap (eV)
ChemOffice
Heat of formation119867∘ (kJmolminus1) Gibbs freeenergy 119866 (kJmolminus1) ideal gas thermalcapacity (IGTC) (Jmolminus1 Kminus1) melting point119879 (Kelvin) critical temperature (CT)(Kelvin) boiling point (TB) (Kelvin) criticalpressure (CP) (Bar) Henryrsquos law constant119870H total valence connectivity (TVC)partition coefficient (PC) moleculartopological index (MTI) number ofrotatable bonds (NRB) shape coefficient 119868sum of valence degrees (SVD) totalconnectivity (TC)
ChemSketchPercent ratios of nitrogen hydrogenoxygen and carbon atoms (N H Oand C) surface tension 120574 (dynecm) indexof refraction (119899) density (119889)
Marvin Sketchlog119875 Winner index (119882) number ofH-Bond acceptors (NHA) number ofH-Bond donors (NHD) Balaban index (119869)polar surface area (PSA) (119860∘)2
correlation with the response activity this one gives extraweight because it will be more effective at prediction [44]Tables 3 and 4 present the descriptors with a correlationcoefficient with the activity higher in absolute value than01 The absence of any serious multicollinearity betweenthe descriptors present in the model was confirmed by thecorrelation matrix
32 Univariate Partitioning (UP) The aim of the UP was therecognition of groups of objects based on their similarity thismethod is based on the criterion of partitioning proposedby Fisher [45] In this study the division of the dataset intotraining and test sets has been performed In this one fromeach obtained cluster one compound for the training setwas selected randomly to be used as test set compound Thepartitioning results are given in Table 5
33 Multiple Linear Regression (MLR) The QSAR analysiswas performed using the values of the chemical descriptorsselected by the PCA method and on the other hand theexperimental values of the antileishmanial activities for 60 ofthe acridines derivatives (effect) Table S1 in SupplementaryMaterial available online at httpdxdoiorg10115520165137289 shows the value of each molecular descriptor that isconfigured in established MLR models and the QSAR model
8 Advances in Physical Chemistry
Table 5 The partitioning results of compounds into training and test sets
ClassesThe test set compounds
Promastigotes Amastigotes3 6 10 14 17 21 37 38 42 64 5 6 26 27 29 36 37 44 46 50
1 1 8 37 40 56 1 10 15 16 34 48 50 57 642 2 14 29 34 48 2 25 38 44 47 51 523 3 18 19 3 26 40 564 4 24 25 27 30 42 58 4 7 8 11 13 17 21 465 6 26 5 14 206 7 9 12 21 33 39 41 47 67 10 11 16 31 32 43 45 18 29 588 15 17 23 28 36 49 52 54 27 639 20 38 46 62 24 28 31 35 36 45 60 6510 50 64 32 33 37
built is represented by the following equations and the valuesof the statistical parameters
MLRModel for pIC50 Promastigotes
pIC50(PRO) = 3030 minus 8214 10minus02N + 0239120583
minus 1264120596 minus 0233NHA + 0732NHD
+ 3311 10minus05MTI + 3184 104TVC
(3)
Statistical Parameters 1198772 = 0723 119877 = 0850 1198772adj = 0664MSE = 0189 P value lt 10minus4
MLRModel for pIC50 Amastigotes
pIC50(AMA) = minus7314 + 1191119864HOMO
+ 2543 10minus02CT minus 0350119870H
minus 1632 10minus02119879 + 0652NHA
+ 1262NHD
(4)
Statistical Parameters 1198772 = 0663 119877 = 0814 1198772adj = 0598MSE = 0208 119875 value lt 10minus4
(i) For the two models P value is lower than 00001it means that we would be taking a lower than001 risk in assuming that the null hypothesis (noeffect of the explanatory variables) is wrong and thatthe regressions equations have statistical significanceTherefore we can conclude with confidence that theselected variables do bring a significant amount ofinformation
(ii) Higher value of 1198772 and 1198772adj and lower mean squarederror (MSE) indicate that the two proposed modelsare predictive and reliable
(iii) Our obtained models were validated internally by theleave-one-out cross validation technique the crossvalidation coefficient 1198772cv for the two models was
determined based on the predictive ability of themodel The value of 1198772cv is higher than 05 (1198772cv =0536 for the MLR pIC
50(PRO) model and 1198772cv =
0525 for the MLR pIC50
(AMA) model) indicatinggood predictability of the model
(iv) True predictive power of these models is to test theirability to predict perfectly pIC
50of compounds from
an external test set (compounds that were not usedfor the developed model) pIC
50of the remaining set
of 10 compounds are deduced from the quantitativemodels proposed with the compounds used in train-ing set by MLR These models will be able to predictthe activities of test set molecules in agreement withthe experimentally determined value The observedand calculated pIC
50values are given in Table 6
The predictive capacity of the models was judgedthe higher value of 1198772test (1198772test = 0660 for theMLR pIC
50(PRO) model and 1198772test = 0718 for the
MLR pIC50
(AMA) model) indicates the improvedpredictability of the model
(v) Xternal Validation Plus indicates the absence of sys-tematic errors in the model and a moderate perfor-mance of prediction quality of a QSAR model basedon proposed MAE-based criteria (Table S2)
(vi) The values of calculated activities from (3) and (4) aregiven in Table S1 and the correlations of calculatedand observed activities values are illustrated in Fig-ure 3
In the first model (pIC50
PRO) the descriptors influencingnegatively the activities are the percent ratio of nitrogen(N) the number of H-Bond acceptors (NHA) and theelectrophilicity index (120596) and the parameters influencingpositively the activities are the dipole moment (120583) thenumber ofH-Bonddonors (NHD) themolecular topologicalindex (MTI) and the total valence connectivity (TVC)
(i) The number of H-Bond acceptors (NHA) has a nega-tive sign in the model which suggests that increasedactivity can be achieved by decreasing the number
Advances in Physical Chemistry 9
Table 6 The values of chemical descriptors and the observed and predicted activities using the MLR models for the test set
Test set (promastigotes) pIC50PRO
Number N 120583 120596 NHA NHD MTI TVC Obs MLR3 12030 4259 5024 3 2 14075 3891 10minus07 minus1669 minus20476 9270 5640 4654 5 2 28763 1716 10minus10 minus0230 minus101410 10780 5730 5267 3 2 16459 4902 10minus08 minus1061 minus183214 11760 2698 3062 3 1 4387 3050 10minus05 0301 minus001217 8580 4969 3431 4 1 10329 5188 10minus07 minus0380 minus066421 7180 5214 3498 5 1 18600 1334 10minus08 minus0114 minus055137 4460 6939 4779 2 0 8245 4076 10minus06 minus2200 minus178138 10680 5122 4374 2 1 6388 5083 10minus06 minus2245 minus151242 8480 5260 4457 3 1 12005 1307 10minus07 minus1815 minus160864 13170 5297 4388 5 2 47653 1201 10minus10 minus0940 minus0455
Test set (amastigotes) pIC50AMA
Number 119864HOMO 119870H 119879 NHA NHD CT Obs MLR5 minus5850 106461 1300866 3 2 19110 0699 minus07876 minus5785 100595 1247456 5 2 18715 1523 033226 minus5665 583700 926651 2 0 7815 minus1170 minus145627 minus5603 622270 930050 3 0 8827 minus1877 minus162829 minus5640 618850 938380 3 0 9007 minus1561 minus146736 minus5966 676650 1037977 2 1 12201 minus1427 minus077437 minus6025 552770 930967 2 0 8348 minus2185 minus145744 minus5919 749400 1083064 2 1 12828 minus1004 minus097846 minus5790 752980 1078419 3 1 13925 minus0973 minus073350 minus5461 625750 913563 3 2 13449 minus0531 minus1028
minus05
0
05
1
15
2
25
minus02 03 08 13 18 23
ActivesValidation
ActivesValidation
minus16
minus06
04
14
minus05 0 05 1 15
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 3 Correlations of observed and predicted activities (training set in black and test set in red) values calculated using MLR models
of heteroatoms (nitrogen or oxygen atoms) mostlythe nitrogen ones for decreasing the ratio of nitrogen(N) having also a negative sign in the model
(ii) The electrophilicity index 120596 has a negative sign inthe model which suggests that increased activity canbe achieved by decreasing the electrophilicity of the
acridine derivatives (a high value of electrophilicitydescribes a good electrophile while a small value ofelectrophilicity describes a good nucleophile)
(iii) The dipolemoment 120583 has a positive sign in themodelwhich suggests that increased activity can be achievedby increasing the polarity of the acridine derivatives
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
Advances in Physical Chemistry 3
N
1
Training set (promastigotes) N NHA NHD MTI
1 14330 4752 5141 3 2 82052 13080 4353 5023 3 2 108354 10070 4124 4469 3 2 255717 8800 0805 4167 5 2 355298 7820 0141 3989 7 2 45907
62 8310 3311 4511 5 2 39911
Obs25332204063406811207
1797Molecular descriptors
minus17minus12minus07minus02
0308131823
minus2 minus1 0 1 2
TrainingValidationTest
23
4
Activity = f (descriptors)
Preparation of the dataset
Calculation of molecular descriptors
Activity
Statistical analysis (QSAR models)
Internal and external validation of QSAR
models
Novel compounds with suggested
activities
N∘ 120583 120596 middot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middotmiddot middot middotmiddot middot middot
middot middot middotmiddot middot middotmiddot middot middot
middot middot middot
pIC50
R5
R6
R7
R4
R1
R2
R3
Pred (pIC50 AMA)pIC50 AMA
Pred (pIC50 AMA)
pIC
50
AM
A
Figure 1 Flow chart of the methodology used in this work
linear regression (MLR) available in the XLSTAT softwareand the artificial neural network (ANN) available in theMatlab software
In order to propose models and to evaluate quanti-tatively the physicochemical effects of the substituents onthe activities of molecules we submitted the data matrixconstituted obviously from the used variables (descriptors)corresponding to the dataset molecules to a descendant MLRanalysis and to an ANN We use the coefficients 119877 1198772 1198772adjMSE and 119875 value to select the best regression performance[29] where119877 is the correlation coefficient1198772 is the coefficientof determination 1198772adj is the coefficient adjusted for degreesof freedom Mean squared error (MSE) is the standard errorof the coefficient of each descriptor and of the global modelwhich gives an indication of the valid inclusion of a descriptorin a QSAR model 119875 value is the probability (119875) of Fisherstatistics (119865) which gives an indication of the probability thata QSAR is a chance correlation
In order to assess the significance of the models andaccurate prediction ability for new compounds
(i) we use an internal validation procedure (leave-one-out cross validation) whereby one compound isremoved and the rebuilt model with the remainingmolecules is used to predict the response of theeliminated compound This one is then returned anda second is removed and the cycle is repeated and
so on until all compounds have been removed oneby one and an overall correlation coefficient 119877cv iscomputed
(ii) after the model is built an external prediction isnecessary In this one the obtained model was usedto predict the activities of a test set comprisingcompounds that are similar to those not used in thetraining set This is usually performed by splitting adataset into a training set and a test set typically ina 15 ratio Further before performing the externalvalidation of a model it is very important to check forthe presence of systematic error that violates the basicassumptions of the least squares regression model Ifhigh systematic error (bias) is present in the modelthen suchmodel should be discarded and performingany external validation test is of no use on such biasedmodel Xternal Validation Plus is a tool that checksthe presence of systematic errors in the model andfurther computes all the required external validationparameters while judging the performance of actualprediction quality of a QSARmodel based on recentlyproposed MAE-based criteria [30]
(iii) a model is valid only within its training domainand new molecules must be considered as belongingto the domain before the model is applied (OECDPrinciple 3 [31]) Without applicability domain (AD)
4 Advances in Physical Chemistry
Table1Ch
emicalstructurea
ndantileishmanialactivities
ofstu
died
compo
unds
(see
Figure
8)
Num
ber
R 1R 2
R 3R 4
R 5R 6
R 7pIC 50
b
PRO
cAMA
d
1H
NHCO
CH3
HH
HNHCO
CH3
Hminus2
533
minus0653
2H
NHCO
C 2H5
HH
HNHCO
C 2H5
Hminus2
204
minus0886
3H
NHCO
C 3H7
HH
HNHCO
C 3H7
Hminus1669
minus110
74
HNHCO
phH
HH
NHCO
PhH
minus0634
minus0041
5H
NHCO
-p-PhC
lH
HH
NHCO
-p-PhC
lH
mdash0699
6H
NHCO
-p-PhF
HH
HNHCO
-p-PhF
Hminus0
230
1523
7H
NHCO
-p-PhO
Me
HH
HNHCO
-p-PhO
Me
Hminus0
681
minus0041
8H
NHCO
-mp-Ph(OMe)2
HH
HNHCO
-mp-Ph(OMe)2
Hminus1207
0097
9H
NHCO
Me
HH
HNHCO
phH
minus2270
Toxa
10H
NHCO
Me
HH
HNHCO
-p-PhC
lH
minus1061
minus0462
11H
NHCO
Me
HH
HNHCO
-p-PhF
Hminus2
123
0174
12H
NHCO
Me
HH
HNHCO
-p-PhO
Me
Hminus1939
Toxa
13H
NHCO
Me
HH
HNHCO
-mp-Ph(OMe)2
Hmdash
minus0114
14H
HCH3
NH2
OMe
HH
0301
0398
15H
HCH2OH
NH2
OMe
HH
minus0732
minus0613
16H
HCH2Br
NH2
OMe
HH
minus0556
minus0663
17H
H(C
H2) 2OCO
OMe
NH2
OMe
HH
minus0380
minus0114
18H
H(C
H2) 2OCO
(CH2) 2CH3
NH2
OMe
HH
minus0491
minus039
819
HH
(CH2) 2OCO
CH2CH
(CH3) 2
NH2
OMe
HH
minus0342
Toxa
20H
H(C
H2) 2OCO
PhNH2
OMe
HH
0398
0699
21H
H(C
H2) 2OCO
PhF
NH2
OMe
HH
minus0114
0222
23H
H(C
H2) 2OCO
PhOMe
NH2
OMe
HH
minus0279
Toxa
24H
HCH3
ClOMe
HH
minus0041
minus0362
25H
HCH2OH
ClOMe
HH
minus2262
minus0959
26H
HCH2Br
ClOMe
HH
minus1703
minus117
027
HH
(CH2) 2OCO
OMe
ClOMe
HH
minus217
8minus1877
28H
H(C
H2) 2OCO
(CH2) 2CH3
ClOMe
HH
minus1707
minus1628
29H
H(C
H2) 2OCO
CH2CH
(CH3) 2
ClOMe
HH
minus1446
minus1561
30H
H(C
H2) 2OCO
PhCl
OMe
HH
minus213
5To
xa31
HH
(CH2) 2OCO
PhF
ClOMe
HH
minus2295
minus1645
32H
H(C
H2) 2OCO
PhCl
ClOMe
HH
minus219
4minus2
191
Advances in Physical Chemistry 5
Table1Con
tinued
Num
ber
R 1R 2
R 3R 4
R 5R 6
R 7pIC 50
b
PRO
cAMA
d
33H
H(C
H2) 2OCO
PhOMe
ClOMe
HH
minus2098
minus2088
34CH2NH2
HH
HH
HH
minus0230
minus053
135
CH2OH
HH
HH
HH
Toxa
minus1515
36CH2NHCO
(CH2) 3Cl
HH
HH
HH
minus1655
minus1427
37CH2OCO
(CH2) 3Cl
HH
HH
HH
minus2200
minus218
538
CH2NHCO
CH=C
H2
HH
HH
HH
minus2245
minus0833
39CH2OCO
CH=C
H2
HH
HH
HH
minus1464
mdash40
CH2NHCO
PhH
HH
HH
Hminus1654
minus112
141
CH2OCO
PhH
HH
HH
Hminus1885
Toxa
42CH2NHCO
-p-PhF
HH
HH
HH
minus1815
mdash43
CH2OCO
-p-PhF
HH
HH
HH
minus174
1To
xa44
CH2NHCO
-p-PhC
lH
HH
HH
HTo
xaminus1004
45CH2OCO
-p-PhC
lH
HH
HH
Hminus1790
minus1433
46CH2NHCO
-p-PhO
Me
HH
HH
HH
minus1513
minus0973
47CH2OCO
-p-PhO
Me
HH
HH
HH
minus1471
minus0869
48CH2NHCO
-p-PhN
Me 2
HH
HH
HH
minus153
9minus0
672
49CH2OCO
-p-PhN
Me 2
HH
HH
HH
minus1819
Toxa
50CH2NH2
HH
HH
HCH2NH2
minus0820
minus053
151
CH2OH
HH
HH
HCH2OH
Toxa
0222
52CH2NHCO
(CH2) 3Cl
HH
HH
HCH2NHCO
(CH2) 3Cl
minus0663
minus0813
54CH2NHCO
CH=C
H2
HH
HH
HCH2NHCO
CHCH2
minus1061
Toxa
56CH2NHCO
PhH
HH
HH
CH2NHCO
Phminus0
556
minus1236
57CH2OCO
PhH
HH
HH
CH2OCO
PhTo
xaminus0
716
58CH2NHCO
-p-PhF
HH
HH
HCH2NHCO
-p-PhF
minus0756
minus0255
60CH2NHCO
-p-PhC
lH
HH
HH
CH2NHCO
-p-PhC
lmdash
minus1562
62CH2NHCO
-p-PhO
Me
HH
HH
HCH2NHCO
-p-PhO
Me
minus1797
Toxa
63CH2OCO
-p-PhO
Me
HH
HH
HCH2OCO
-p-PhO
Me
Toxa
minus1825
64CH2NHCO
-p-PhN
Me 2
HH
HH
HCH2NHCO
-p-PhN
Me 2
minus0940
minus0724
65CH2OCO
-p-PhN
Me 2
HH
HH
HCH2OCO
-p-PhN
Me 2
Toxa
minus1667
a Toxictoxicity
observed
onhu
man
macroph
agesatconcentrations
thatdidno
tdisp
layantileishmanialactivity
-p-para-m
-meta
b pIC50=minuslog(
IC50)
c pIC50AMAantileish
manialactivity
againstamastig
otes
parasiteformdpIC 50PR
Oantileish
manialactivity
againstp
romastig
otes
parasiteform
6 Advances in Physical Chemistry
Inputw
b
w
b
+ +
Hidden layer Output layer
Output
Figure 2 The architecture used in our study of the artificial neural network
each model can predict the activity of any com-pound even with a completely different structurefrom those included in the study Therefore the ADis a tool to find out compounds that are outsidethe applicability domain of the built QSAR modeland it detects outliers present in the training setcompounds There are several methods for definingthe applicability domain (AD) of QSAR models [32]but themost common one is determining the leveragevalues ℎ119894 (ℎ119894 = 119909119879
119894(119883119879 119883)
minus1
119909119894 (119894 = 1 2 119899)) foreach compound where119909119894 is the descriptor row-vectorof query compound 119883 is 119899 119909 119896 minus 1 matrix of 119896model descriptor values for 119899 training set compoundsand the superscript 119879 refers to the transpose ofmatrixvector [32 33] In this study we use Williamsplot in this plot the applicability domain is estab-lished inside a squared area within standard deviationplusmn119909 (in this study 119909 = 3 ldquothree-sigma rulerdquo [34]) anda leverage threshold ℎlowast(ℎlowast = 3 lowast (119896 + 1)119899) [35]where 119899 is the number of training set compounds 119896is the number of model descriptors The leverage (ℎ)greater than the warning leverage (ℎlowast) suggested thatthe compoundwas very influential on themodel [36]The results of the leverage approach were comparedwith that of the simple approach introduced by Royet al [37]
231 Multiple Linear Regression (MLR) The descendentmultiple linear regression (MLR) analysis based on theelimination of aberrant descriptors (one by one) until a validmodel (including the critical probability 119875 value lt 005 forall descriptors and the model complete) was employed tofind a linear model of the activity of interest which takes thefollowing form
119884 = 1198860 +119899
sum119894=1
119886119894119909119894 (1)
where 119884 is the studied activity which is the dependent vari-able 1198860 is the intercept of the equation 119909119894 are the moleculardescriptors 119886119894 are the coefficients of those descriptors
This method is one of the most popular methods ofQSARQSPR thanks to its simplicity in operation repro-ducibility and ability to allow easy interpretation of thefeatures used The important advantage of the linear regres-sion analysis is that it is highly transparent therefore thealgorithm is available and predictions can bemade easily [38]It has served also to select the descriptors used as the inputparameters in the artificial neural network (ANN)
Table 2 Software packages used in this work
Drawing chemicalstructures
Marvin Sketch ACDChemSketchand ChemBioDraw
Generating 3D structures Gauss View 30 and ChemBio3DCalculating chemicaldescriptors
Gaussian 03 Marvin Sketch 62ChemSketch and ChemBio3D
Developing QSAR models XLSTAT 2009 and Matlab 790(version 2011)
232 Artificial Neural Networks (ANN) The artificial neuralnetworks (ANN) are used in order to increase the probabilityof characterizing the compounds and to generate a predic-tive QSAR model between the set of molecular descriptorsobtained from the MLR models and the observed activitiesvalues The ANN model is done on the MATLAB R2009bsoftware It consists of three layers of neurons called inputlayer hidden layer and output layer (Figure 2) The inputlayer formed by a number of neurons equal to the numberof descriptors obtained in the multiple linear regressionmodels and the output layer represents the calculated activityvalues For determination of the number of hidden neuronsin the hidden layer where all calculations of parameteroptimization of neural networks are made a parameter 120588has been proposed The parameter 120588 plays a major role indetermining the best artificial neural network architecture[39 40] 120588 is defined as follows
120588 =number of data points in the training set
sum of the number of connections in the ANN (2)
In order to avoid overfitting or underfitting it is recom-mended that the value of 120588 should be between 100 and 300if 120588 lt 1 the network simply memorizes the data whereas if120588 gt 3 the network is not able to generalize [41]
24 Software Packages Used in Our QSARDevelopment StudyThere are various free and commercial software available forQSAR development These include specialized software fordrawing chemical structures generating 3D structures cal-culating chemical descriptors and developing QSARmodelsThe software packages used in our works are represented inTable 2
3 Results and Discussions
31 Principal Component Analysis (PCA) PCA is a usefulstatistical technique for summarizing all the informationencoded in the structures of the compounds It is very helpful
Advances in Physical Chemistry 7
Table 3The absolute values of correlation coefficients with descrip-tors and activitiesDescriptors 119903Promastigotes
Henryrsquos law constant (119870H) 0130Ideal gas thermal capacity (IGTC) 0108Number of H-Bond acceptors (NHA) 0299Number of H-Bond donors (NHD) 0250Balaban index (119869) 0133Shape coefficient (119868) 0135Total valence connectivity (TVC) 0292N 0238H 0260Gibbs free energy (119866) 0119Dipole moment (120583) 0213Electronegativity (120594) 0599Electrophilicity index (120596) 0630Molecular topological index (MTI) 0116Polar surface area (PSA) 0282Total connectivity (TC) 0269Wiener index (119882) 0124119864HOMO 0554119864LUMO 0632119864Gap 0227
AmastigotesTotal energy 119864 0160119864HOMO 0286Electronegativity (120594) 0316Polar surface area (PSA) 0391O 0105Index of refraction (119899) 0548Density (119889) 0242Critical pressure (CP) 0176Gibbs free energy (119866) 0195Henryrsquos law constant (119870H) 0387Number of H-Bond acceptors (NHA) 0198Number of H-Bond donors (NHD) 0587Number of rotatable bonds (NRB) 0139Sum of valence degrees (SVD) 0146Dipole moment (120583) 0224119864LUMO 0338Electrophilicity index (120596) 0335N 0436C 0155Surface tension (120574) 0553Boiling point (TB) 0254Critical temperature (CT) 0276Heat of formation (119867∘) 0152log119875 0248Melting point (119879) 0314Partition coefficient (PC) 0163Shape coefficient (119868) 0176Winner index (119882) 0103
for understanding the distribution of the compounds Thisis an essentially descriptive statistical method which aimsto extract the maximum of information contained in thedataset compounds [42 43] In this work the PCA is usedto overview the examined compounds for similarities anddissimilarities and to select descriptors that show a high
Table 4 Descriptors selected by the principal component analysis(PCA) and software packages used in the calculation of descriptors
Software Descriptors
Gaussian 03
Highest occupied molecular orbital energy119864HOMO (eV) lowest unoccupied molecularorbital energy 119864LUMO (eV) hardness 120578 (eV)= (119864LUMO minus 119864HOMO)2 electronegativity 120594(eV) = minus(119864LUMO + 119864HOMO)2 electrophilicityindex 120596 (eV) = 12059422120578 total energy 119864 (eV)dipole moment 120583 (Debye) energy gapbetween 119864HOMO and 119864LUMO values 119864Gap (eV)
ChemOffice
Heat of formation119867∘ (kJmolminus1) Gibbs freeenergy 119866 (kJmolminus1) ideal gas thermalcapacity (IGTC) (Jmolminus1 Kminus1) melting point119879 (Kelvin) critical temperature (CT)(Kelvin) boiling point (TB) (Kelvin) criticalpressure (CP) (Bar) Henryrsquos law constant119870H total valence connectivity (TVC)partition coefficient (PC) moleculartopological index (MTI) number ofrotatable bonds (NRB) shape coefficient 119868sum of valence degrees (SVD) totalconnectivity (TC)
ChemSketchPercent ratios of nitrogen hydrogenoxygen and carbon atoms (N H Oand C) surface tension 120574 (dynecm) indexof refraction (119899) density (119889)
Marvin Sketchlog119875 Winner index (119882) number ofH-Bond acceptors (NHA) number ofH-Bond donors (NHD) Balaban index (119869)polar surface area (PSA) (119860∘)2
correlation with the response activity this one gives extraweight because it will be more effective at prediction [44]Tables 3 and 4 present the descriptors with a correlationcoefficient with the activity higher in absolute value than01 The absence of any serious multicollinearity betweenthe descriptors present in the model was confirmed by thecorrelation matrix
32 Univariate Partitioning (UP) The aim of the UP was therecognition of groups of objects based on their similarity thismethod is based on the criterion of partitioning proposedby Fisher [45] In this study the division of the dataset intotraining and test sets has been performed In this one fromeach obtained cluster one compound for the training setwas selected randomly to be used as test set compound Thepartitioning results are given in Table 5
33 Multiple Linear Regression (MLR) The QSAR analysiswas performed using the values of the chemical descriptorsselected by the PCA method and on the other hand theexperimental values of the antileishmanial activities for 60 ofthe acridines derivatives (effect) Table S1 in SupplementaryMaterial available online at httpdxdoiorg10115520165137289 shows the value of each molecular descriptor that isconfigured in established MLR models and the QSAR model
8 Advances in Physical Chemistry
Table 5 The partitioning results of compounds into training and test sets
ClassesThe test set compounds
Promastigotes Amastigotes3 6 10 14 17 21 37 38 42 64 5 6 26 27 29 36 37 44 46 50
1 1 8 37 40 56 1 10 15 16 34 48 50 57 642 2 14 29 34 48 2 25 38 44 47 51 523 3 18 19 3 26 40 564 4 24 25 27 30 42 58 4 7 8 11 13 17 21 465 6 26 5 14 206 7 9 12 21 33 39 41 47 67 10 11 16 31 32 43 45 18 29 588 15 17 23 28 36 49 52 54 27 639 20 38 46 62 24 28 31 35 36 45 60 6510 50 64 32 33 37
built is represented by the following equations and the valuesof the statistical parameters
MLRModel for pIC50 Promastigotes
pIC50(PRO) = 3030 minus 8214 10minus02N + 0239120583
minus 1264120596 minus 0233NHA + 0732NHD
+ 3311 10minus05MTI + 3184 104TVC
(3)
Statistical Parameters 1198772 = 0723 119877 = 0850 1198772adj = 0664MSE = 0189 P value lt 10minus4
MLRModel for pIC50 Amastigotes
pIC50(AMA) = minus7314 + 1191119864HOMO
+ 2543 10minus02CT minus 0350119870H
minus 1632 10minus02119879 + 0652NHA
+ 1262NHD
(4)
Statistical Parameters 1198772 = 0663 119877 = 0814 1198772adj = 0598MSE = 0208 119875 value lt 10minus4
(i) For the two models P value is lower than 00001it means that we would be taking a lower than001 risk in assuming that the null hypothesis (noeffect of the explanatory variables) is wrong and thatthe regressions equations have statistical significanceTherefore we can conclude with confidence that theselected variables do bring a significant amount ofinformation
(ii) Higher value of 1198772 and 1198772adj and lower mean squarederror (MSE) indicate that the two proposed modelsare predictive and reliable
(iii) Our obtained models were validated internally by theleave-one-out cross validation technique the crossvalidation coefficient 1198772cv for the two models was
determined based on the predictive ability of themodel The value of 1198772cv is higher than 05 (1198772cv =0536 for the MLR pIC
50(PRO) model and 1198772cv =
0525 for the MLR pIC50
(AMA) model) indicatinggood predictability of the model
(iv) True predictive power of these models is to test theirability to predict perfectly pIC
50of compounds from
an external test set (compounds that were not usedfor the developed model) pIC
50of the remaining set
of 10 compounds are deduced from the quantitativemodels proposed with the compounds used in train-ing set by MLR These models will be able to predictthe activities of test set molecules in agreement withthe experimentally determined value The observedand calculated pIC
50values are given in Table 6
The predictive capacity of the models was judgedthe higher value of 1198772test (1198772test = 0660 for theMLR pIC
50(PRO) model and 1198772test = 0718 for the
MLR pIC50
(AMA) model) indicates the improvedpredictability of the model
(v) Xternal Validation Plus indicates the absence of sys-tematic errors in the model and a moderate perfor-mance of prediction quality of a QSAR model basedon proposed MAE-based criteria (Table S2)
(vi) The values of calculated activities from (3) and (4) aregiven in Table S1 and the correlations of calculatedand observed activities values are illustrated in Fig-ure 3
In the first model (pIC50
PRO) the descriptors influencingnegatively the activities are the percent ratio of nitrogen(N) the number of H-Bond acceptors (NHA) and theelectrophilicity index (120596) and the parameters influencingpositively the activities are the dipole moment (120583) thenumber ofH-Bonddonors (NHD) themolecular topologicalindex (MTI) and the total valence connectivity (TVC)
(i) The number of H-Bond acceptors (NHA) has a nega-tive sign in the model which suggests that increasedactivity can be achieved by decreasing the number
Advances in Physical Chemistry 9
Table 6 The values of chemical descriptors and the observed and predicted activities using the MLR models for the test set
Test set (promastigotes) pIC50PRO
Number N 120583 120596 NHA NHD MTI TVC Obs MLR3 12030 4259 5024 3 2 14075 3891 10minus07 minus1669 minus20476 9270 5640 4654 5 2 28763 1716 10minus10 minus0230 minus101410 10780 5730 5267 3 2 16459 4902 10minus08 minus1061 minus183214 11760 2698 3062 3 1 4387 3050 10minus05 0301 minus001217 8580 4969 3431 4 1 10329 5188 10minus07 minus0380 minus066421 7180 5214 3498 5 1 18600 1334 10minus08 minus0114 minus055137 4460 6939 4779 2 0 8245 4076 10minus06 minus2200 minus178138 10680 5122 4374 2 1 6388 5083 10minus06 minus2245 minus151242 8480 5260 4457 3 1 12005 1307 10minus07 minus1815 minus160864 13170 5297 4388 5 2 47653 1201 10minus10 minus0940 minus0455
Test set (amastigotes) pIC50AMA
Number 119864HOMO 119870H 119879 NHA NHD CT Obs MLR5 minus5850 106461 1300866 3 2 19110 0699 minus07876 minus5785 100595 1247456 5 2 18715 1523 033226 minus5665 583700 926651 2 0 7815 minus1170 minus145627 minus5603 622270 930050 3 0 8827 minus1877 minus162829 minus5640 618850 938380 3 0 9007 minus1561 minus146736 minus5966 676650 1037977 2 1 12201 minus1427 minus077437 minus6025 552770 930967 2 0 8348 minus2185 minus145744 minus5919 749400 1083064 2 1 12828 minus1004 minus097846 minus5790 752980 1078419 3 1 13925 minus0973 minus073350 minus5461 625750 913563 3 2 13449 minus0531 minus1028
minus05
0
05
1
15
2
25
minus02 03 08 13 18 23
ActivesValidation
ActivesValidation
minus16
minus06
04
14
minus05 0 05 1 15
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 3 Correlations of observed and predicted activities (training set in black and test set in red) values calculated using MLR models
of heteroatoms (nitrogen or oxygen atoms) mostlythe nitrogen ones for decreasing the ratio of nitrogen(N) having also a negative sign in the model
(ii) The electrophilicity index 120596 has a negative sign inthe model which suggests that increased activity canbe achieved by decreasing the electrophilicity of the
acridine derivatives (a high value of electrophilicitydescribes a good electrophile while a small value ofelectrophilicity describes a good nucleophile)
(iii) The dipolemoment 120583 has a positive sign in themodelwhich suggests that increased activity can be achievedby increasing the polarity of the acridine derivatives
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
4 Advances in Physical Chemistry
Table1Ch
emicalstructurea
ndantileishmanialactivities
ofstu
died
compo
unds
(see
Figure
8)
Num
ber
R 1R 2
R 3R 4
R 5R 6
R 7pIC 50
b
PRO
cAMA
d
1H
NHCO
CH3
HH
HNHCO
CH3
Hminus2
533
minus0653
2H
NHCO
C 2H5
HH
HNHCO
C 2H5
Hminus2
204
minus0886
3H
NHCO
C 3H7
HH
HNHCO
C 3H7
Hminus1669
minus110
74
HNHCO
phH
HH
NHCO
PhH
minus0634
minus0041
5H
NHCO
-p-PhC
lH
HH
NHCO
-p-PhC
lH
mdash0699
6H
NHCO
-p-PhF
HH
HNHCO
-p-PhF
Hminus0
230
1523
7H
NHCO
-p-PhO
Me
HH
HNHCO
-p-PhO
Me
Hminus0
681
minus0041
8H
NHCO
-mp-Ph(OMe)2
HH
HNHCO
-mp-Ph(OMe)2
Hminus1207
0097
9H
NHCO
Me
HH
HNHCO
phH
minus2270
Toxa
10H
NHCO
Me
HH
HNHCO
-p-PhC
lH
minus1061
minus0462
11H
NHCO
Me
HH
HNHCO
-p-PhF
Hminus2
123
0174
12H
NHCO
Me
HH
HNHCO
-p-PhO
Me
Hminus1939
Toxa
13H
NHCO
Me
HH
HNHCO
-mp-Ph(OMe)2
Hmdash
minus0114
14H
HCH3
NH2
OMe
HH
0301
0398
15H
HCH2OH
NH2
OMe
HH
minus0732
minus0613
16H
HCH2Br
NH2
OMe
HH
minus0556
minus0663
17H
H(C
H2) 2OCO
OMe
NH2
OMe
HH
minus0380
minus0114
18H
H(C
H2) 2OCO
(CH2) 2CH3
NH2
OMe
HH
minus0491
minus039
819
HH
(CH2) 2OCO
CH2CH
(CH3) 2
NH2
OMe
HH
minus0342
Toxa
20H
H(C
H2) 2OCO
PhNH2
OMe
HH
0398
0699
21H
H(C
H2) 2OCO
PhF
NH2
OMe
HH
minus0114
0222
23H
H(C
H2) 2OCO
PhOMe
NH2
OMe
HH
minus0279
Toxa
24H
HCH3
ClOMe
HH
minus0041
minus0362
25H
HCH2OH
ClOMe
HH
minus2262
minus0959
26H
HCH2Br
ClOMe
HH
minus1703
minus117
027
HH
(CH2) 2OCO
OMe
ClOMe
HH
minus217
8minus1877
28H
H(C
H2) 2OCO
(CH2) 2CH3
ClOMe
HH
minus1707
minus1628
29H
H(C
H2) 2OCO
CH2CH
(CH3) 2
ClOMe
HH
minus1446
minus1561
30H
H(C
H2) 2OCO
PhCl
OMe
HH
minus213
5To
xa31
HH
(CH2) 2OCO
PhF
ClOMe
HH
minus2295
minus1645
32H
H(C
H2) 2OCO
PhCl
ClOMe
HH
minus219
4minus2
191
Advances in Physical Chemistry 5
Table1Con
tinued
Num
ber
R 1R 2
R 3R 4
R 5R 6
R 7pIC 50
b
PRO
cAMA
d
33H
H(C
H2) 2OCO
PhOMe
ClOMe
HH
minus2098
minus2088
34CH2NH2
HH
HH
HH
minus0230
minus053
135
CH2OH
HH
HH
HH
Toxa
minus1515
36CH2NHCO
(CH2) 3Cl
HH
HH
HH
minus1655
minus1427
37CH2OCO
(CH2) 3Cl
HH
HH
HH
minus2200
minus218
538
CH2NHCO
CH=C
H2
HH
HH
HH
minus2245
minus0833
39CH2OCO
CH=C
H2
HH
HH
HH
minus1464
mdash40
CH2NHCO
PhH
HH
HH
Hminus1654
minus112
141
CH2OCO
PhH
HH
HH
Hminus1885
Toxa
42CH2NHCO
-p-PhF
HH
HH
HH
minus1815
mdash43
CH2OCO
-p-PhF
HH
HH
HH
minus174
1To
xa44
CH2NHCO
-p-PhC
lH
HH
HH
HTo
xaminus1004
45CH2OCO
-p-PhC
lH
HH
HH
Hminus1790
minus1433
46CH2NHCO
-p-PhO
Me
HH
HH
HH
minus1513
minus0973
47CH2OCO
-p-PhO
Me
HH
HH
HH
minus1471
minus0869
48CH2NHCO
-p-PhN
Me 2
HH
HH
HH
minus153
9minus0
672
49CH2OCO
-p-PhN
Me 2
HH
HH
HH
minus1819
Toxa
50CH2NH2
HH
HH
HCH2NH2
minus0820
minus053
151
CH2OH
HH
HH
HCH2OH
Toxa
0222
52CH2NHCO
(CH2) 3Cl
HH
HH
HCH2NHCO
(CH2) 3Cl
minus0663
minus0813
54CH2NHCO
CH=C
H2
HH
HH
HCH2NHCO
CHCH2
minus1061
Toxa
56CH2NHCO
PhH
HH
HH
CH2NHCO
Phminus0
556
minus1236
57CH2OCO
PhH
HH
HH
CH2OCO
PhTo
xaminus0
716
58CH2NHCO
-p-PhF
HH
HH
HCH2NHCO
-p-PhF
minus0756
minus0255
60CH2NHCO
-p-PhC
lH
HH
HH
CH2NHCO
-p-PhC
lmdash
minus1562
62CH2NHCO
-p-PhO
Me
HH
HH
HCH2NHCO
-p-PhO
Me
minus1797
Toxa
63CH2OCO
-p-PhO
Me
HH
HH
HCH2OCO
-p-PhO
Me
Toxa
minus1825
64CH2NHCO
-p-PhN
Me 2
HH
HH
HCH2NHCO
-p-PhN
Me 2
minus0940
minus0724
65CH2OCO
-p-PhN
Me 2
HH
HH
HCH2OCO
-p-PhN
Me 2
Toxa
minus1667
a Toxictoxicity
observed
onhu
man
macroph
agesatconcentrations
thatdidno
tdisp
layantileishmanialactivity
-p-para-m
-meta
b pIC50=minuslog(
IC50)
c pIC50AMAantileish
manialactivity
againstamastig
otes
parasiteformdpIC 50PR
Oantileish
manialactivity
againstp
romastig
otes
parasiteform
6 Advances in Physical Chemistry
Inputw
b
w
b
+ +
Hidden layer Output layer
Output
Figure 2 The architecture used in our study of the artificial neural network
each model can predict the activity of any com-pound even with a completely different structurefrom those included in the study Therefore the ADis a tool to find out compounds that are outsidethe applicability domain of the built QSAR modeland it detects outliers present in the training setcompounds There are several methods for definingthe applicability domain (AD) of QSAR models [32]but themost common one is determining the leveragevalues ℎ119894 (ℎ119894 = 119909119879
119894(119883119879 119883)
minus1
119909119894 (119894 = 1 2 119899)) foreach compound where119909119894 is the descriptor row-vectorof query compound 119883 is 119899 119909 119896 minus 1 matrix of 119896model descriptor values for 119899 training set compoundsand the superscript 119879 refers to the transpose ofmatrixvector [32 33] In this study we use Williamsplot in this plot the applicability domain is estab-lished inside a squared area within standard deviationplusmn119909 (in this study 119909 = 3 ldquothree-sigma rulerdquo [34]) anda leverage threshold ℎlowast(ℎlowast = 3 lowast (119896 + 1)119899) [35]where 119899 is the number of training set compounds 119896is the number of model descriptors The leverage (ℎ)greater than the warning leverage (ℎlowast) suggested thatthe compoundwas very influential on themodel [36]The results of the leverage approach were comparedwith that of the simple approach introduced by Royet al [37]
231 Multiple Linear Regression (MLR) The descendentmultiple linear regression (MLR) analysis based on theelimination of aberrant descriptors (one by one) until a validmodel (including the critical probability 119875 value lt 005 forall descriptors and the model complete) was employed tofind a linear model of the activity of interest which takes thefollowing form
119884 = 1198860 +119899
sum119894=1
119886119894119909119894 (1)
where 119884 is the studied activity which is the dependent vari-able 1198860 is the intercept of the equation 119909119894 are the moleculardescriptors 119886119894 are the coefficients of those descriptors
This method is one of the most popular methods ofQSARQSPR thanks to its simplicity in operation repro-ducibility and ability to allow easy interpretation of thefeatures used The important advantage of the linear regres-sion analysis is that it is highly transparent therefore thealgorithm is available and predictions can bemade easily [38]It has served also to select the descriptors used as the inputparameters in the artificial neural network (ANN)
Table 2 Software packages used in this work
Drawing chemicalstructures
Marvin Sketch ACDChemSketchand ChemBioDraw
Generating 3D structures Gauss View 30 and ChemBio3DCalculating chemicaldescriptors
Gaussian 03 Marvin Sketch 62ChemSketch and ChemBio3D
Developing QSAR models XLSTAT 2009 and Matlab 790(version 2011)
232 Artificial Neural Networks (ANN) The artificial neuralnetworks (ANN) are used in order to increase the probabilityof characterizing the compounds and to generate a predic-tive QSAR model between the set of molecular descriptorsobtained from the MLR models and the observed activitiesvalues The ANN model is done on the MATLAB R2009bsoftware It consists of three layers of neurons called inputlayer hidden layer and output layer (Figure 2) The inputlayer formed by a number of neurons equal to the numberof descriptors obtained in the multiple linear regressionmodels and the output layer represents the calculated activityvalues For determination of the number of hidden neuronsin the hidden layer where all calculations of parameteroptimization of neural networks are made a parameter 120588has been proposed The parameter 120588 plays a major role indetermining the best artificial neural network architecture[39 40] 120588 is defined as follows
120588 =number of data points in the training set
sum of the number of connections in the ANN (2)
In order to avoid overfitting or underfitting it is recom-mended that the value of 120588 should be between 100 and 300if 120588 lt 1 the network simply memorizes the data whereas if120588 gt 3 the network is not able to generalize [41]
24 Software Packages Used in Our QSARDevelopment StudyThere are various free and commercial software available forQSAR development These include specialized software fordrawing chemical structures generating 3D structures cal-culating chemical descriptors and developing QSARmodelsThe software packages used in our works are represented inTable 2
3 Results and Discussions
31 Principal Component Analysis (PCA) PCA is a usefulstatistical technique for summarizing all the informationencoded in the structures of the compounds It is very helpful
Advances in Physical Chemistry 7
Table 3The absolute values of correlation coefficients with descrip-tors and activitiesDescriptors 119903Promastigotes
Henryrsquos law constant (119870H) 0130Ideal gas thermal capacity (IGTC) 0108Number of H-Bond acceptors (NHA) 0299Number of H-Bond donors (NHD) 0250Balaban index (119869) 0133Shape coefficient (119868) 0135Total valence connectivity (TVC) 0292N 0238H 0260Gibbs free energy (119866) 0119Dipole moment (120583) 0213Electronegativity (120594) 0599Electrophilicity index (120596) 0630Molecular topological index (MTI) 0116Polar surface area (PSA) 0282Total connectivity (TC) 0269Wiener index (119882) 0124119864HOMO 0554119864LUMO 0632119864Gap 0227
AmastigotesTotal energy 119864 0160119864HOMO 0286Electronegativity (120594) 0316Polar surface area (PSA) 0391O 0105Index of refraction (119899) 0548Density (119889) 0242Critical pressure (CP) 0176Gibbs free energy (119866) 0195Henryrsquos law constant (119870H) 0387Number of H-Bond acceptors (NHA) 0198Number of H-Bond donors (NHD) 0587Number of rotatable bonds (NRB) 0139Sum of valence degrees (SVD) 0146Dipole moment (120583) 0224119864LUMO 0338Electrophilicity index (120596) 0335N 0436C 0155Surface tension (120574) 0553Boiling point (TB) 0254Critical temperature (CT) 0276Heat of formation (119867∘) 0152log119875 0248Melting point (119879) 0314Partition coefficient (PC) 0163Shape coefficient (119868) 0176Winner index (119882) 0103
for understanding the distribution of the compounds Thisis an essentially descriptive statistical method which aimsto extract the maximum of information contained in thedataset compounds [42 43] In this work the PCA is usedto overview the examined compounds for similarities anddissimilarities and to select descriptors that show a high
Table 4 Descriptors selected by the principal component analysis(PCA) and software packages used in the calculation of descriptors
Software Descriptors
Gaussian 03
Highest occupied molecular orbital energy119864HOMO (eV) lowest unoccupied molecularorbital energy 119864LUMO (eV) hardness 120578 (eV)= (119864LUMO minus 119864HOMO)2 electronegativity 120594(eV) = minus(119864LUMO + 119864HOMO)2 electrophilicityindex 120596 (eV) = 12059422120578 total energy 119864 (eV)dipole moment 120583 (Debye) energy gapbetween 119864HOMO and 119864LUMO values 119864Gap (eV)
ChemOffice
Heat of formation119867∘ (kJmolminus1) Gibbs freeenergy 119866 (kJmolminus1) ideal gas thermalcapacity (IGTC) (Jmolminus1 Kminus1) melting point119879 (Kelvin) critical temperature (CT)(Kelvin) boiling point (TB) (Kelvin) criticalpressure (CP) (Bar) Henryrsquos law constant119870H total valence connectivity (TVC)partition coefficient (PC) moleculartopological index (MTI) number ofrotatable bonds (NRB) shape coefficient 119868sum of valence degrees (SVD) totalconnectivity (TC)
ChemSketchPercent ratios of nitrogen hydrogenoxygen and carbon atoms (N H Oand C) surface tension 120574 (dynecm) indexof refraction (119899) density (119889)
Marvin Sketchlog119875 Winner index (119882) number ofH-Bond acceptors (NHA) number ofH-Bond donors (NHD) Balaban index (119869)polar surface area (PSA) (119860∘)2
correlation with the response activity this one gives extraweight because it will be more effective at prediction [44]Tables 3 and 4 present the descriptors with a correlationcoefficient with the activity higher in absolute value than01 The absence of any serious multicollinearity betweenthe descriptors present in the model was confirmed by thecorrelation matrix
32 Univariate Partitioning (UP) The aim of the UP was therecognition of groups of objects based on their similarity thismethod is based on the criterion of partitioning proposedby Fisher [45] In this study the division of the dataset intotraining and test sets has been performed In this one fromeach obtained cluster one compound for the training setwas selected randomly to be used as test set compound Thepartitioning results are given in Table 5
33 Multiple Linear Regression (MLR) The QSAR analysiswas performed using the values of the chemical descriptorsselected by the PCA method and on the other hand theexperimental values of the antileishmanial activities for 60 ofthe acridines derivatives (effect) Table S1 in SupplementaryMaterial available online at httpdxdoiorg10115520165137289 shows the value of each molecular descriptor that isconfigured in established MLR models and the QSAR model
8 Advances in Physical Chemistry
Table 5 The partitioning results of compounds into training and test sets
ClassesThe test set compounds
Promastigotes Amastigotes3 6 10 14 17 21 37 38 42 64 5 6 26 27 29 36 37 44 46 50
1 1 8 37 40 56 1 10 15 16 34 48 50 57 642 2 14 29 34 48 2 25 38 44 47 51 523 3 18 19 3 26 40 564 4 24 25 27 30 42 58 4 7 8 11 13 17 21 465 6 26 5 14 206 7 9 12 21 33 39 41 47 67 10 11 16 31 32 43 45 18 29 588 15 17 23 28 36 49 52 54 27 639 20 38 46 62 24 28 31 35 36 45 60 6510 50 64 32 33 37
built is represented by the following equations and the valuesof the statistical parameters
MLRModel for pIC50 Promastigotes
pIC50(PRO) = 3030 minus 8214 10minus02N + 0239120583
minus 1264120596 minus 0233NHA + 0732NHD
+ 3311 10minus05MTI + 3184 104TVC
(3)
Statistical Parameters 1198772 = 0723 119877 = 0850 1198772adj = 0664MSE = 0189 P value lt 10minus4
MLRModel for pIC50 Amastigotes
pIC50(AMA) = minus7314 + 1191119864HOMO
+ 2543 10minus02CT minus 0350119870H
minus 1632 10minus02119879 + 0652NHA
+ 1262NHD
(4)
Statistical Parameters 1198772 = 0663 119877 = 0814 1198772adj = 0598MSE = 0208 119875 value lt 10minus4
(i) For the two models P value is lower than 00001it means that we would be taking a lower than001 risk in assuming that the null hypothesis (noeffect of the explanatory variables) is wrong and thatthe regressions equations have statistical significanceTherefore we can conclude with confidence that theselected variables do bring a significant amount ofinformation
(ii) Higher value of 1198772 and 1198772adj and lower mean squarederror (MSE) indicate that the two proposed modelsare predictive and reliable
(iii) Our obtained models were validated internally by theleave-one-out cross validation technique the crossvalidation coefficient 1198772cv for the two models was
determined based on the predictive ability of themodel The value of 1198772cv is higher than 05 (1198772cv =0536 for the MLR pIC
50(PRO) model and 1198772cv =
0525 for the MLR pIC50
(AMA) model) indicatinggood predictability of the model
(iv) True predictive power of these models is to test theirability to predict perfectly pIC
50of compounds from
an external test set (compounds that were not usedfor the developed model) pIC
50of the remaining set
of 10 compounds are deduced from the quantitativemodels proposed with the compounds used in train-ing set by MLR These models will be able to predictthe activities of test set molecules in agreement withthe experimentally determined value The observedand calculated pIC
50values are given in Table 6
The predictive capacity of the models was judgedthe higher value of 1198772test (1198772test = 0660 for theMLR pIC
50(PRO) model and 1198772test = 0718 for the
MLR pIC50
(AMA) model) indicates the improvedpredictability of the model
(v) Xternal Validation Plus indicates the absence of sys-tematic errors in the model and a moderate perfor-mance of prediction quality of a QSAR model basedon proposed MAE-based criteria (Table S2)
(vi) The values of calculated activities from (3) and (4) aregiven in Table S1 and the correlations of calculatedand observed activities values are illustrated in Fig-ure 3
In the first model (pIC50
PRO) the descriptors influencingnegatively the activities are the percent ratio of nitrogen(N) the number of H-Bond acceptors (NHA) and theelectrophilicity index (120596) and the parameters influencingpositively the activities are the dipole moment (120583) thenumber ofH-Bonddonors (NHD) themolecular topologicalindex (MTI) and the total valence connectivity (TVC)
(i) The number of H-Bond acceptors (NHA) has a nega-tive sign in the model which suggests that increasedactivity can be achieved by decreasing the number
Advances in Physical Chemistry 9
Table 6 The values of chemical descriptors and the observed and predicted activities using the MLR models for the test set
Test set (promastigotes) pIC50PRO
Number N 120583 120596 NHA NHD MTI TVC Obs MLR3 12030 4259 5024 3 2 14075 3891 10minus07 minus1669 minus20476 9270 5640 4654 5 2 28763 1716 10minus10 minus0230 minus101410 10780 5730 5267 3 2 16459 4902 10minus08 minus1061 minus183214 11760 2698 3062 3 1 4387 3050 10minus05 0301 minus001217 8580 4969 3431 4 1 10329 5188 10minus07 minus0380 minus066421 7180 5214 3498 5 1 18600 1334 10minus08 minus0114 minus055137 4460 6939 4779 2 0 8245 4076 10minus06 minus2200 minus178138 10680 5122 4374 2 1 6388 5083 10minus06 minus2245 minus151242 8480 5260 4457 3 1 12005 1307 10minus07 minus1815 minus160864 13170 5297 4388 5 2 47653 1201 10minus10 minus0940 minus0455
Test set (amastigotes) pIC50AMA
Number 119864HOMO 119870H 119879 NHA NHD CT Obs MLR5 minus5850 106461 1300866 3 2 19110 0699 minus07876 minus5785 100595 1247456 5 2 18715 1523 033226 minus5665 583700 926651 2 0 7815 minus1170 minus145627 minus5603 622270 930050 3 0 8827 minus1877 minus162829 minus5640 618850 938380 3 0 9007 minus1561 minus146736 minus5966 676650 1037977 2 1 12201 minus1427 minus077437 minus6025 552770 930967 2 0 8348 minus2185 minus145744 minus5919 749400 1083064 2 1 12828 minus1004 minus097846 minus5790 752980 1078419 3 1 13925 minus0973 minus073350 minus5461 625750 913563 3 2 13449 minus0531 minus1028
minus05
0
05
1
15
2
25
minus02 03 08 13 18 23
ActivesValidation
ActivesValidation
minus16
minus06
04
14
minus05 0 05 1 15
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 3 Correlations of observed and predicted activities (training set in black and test set in red) values calculated using MLR models
of heteroatoms (nitrogen or oxygen atoms) mostlythe nitrogen ones for decreasing the ratio of nitrogen(N) having also a negative sign in the model
(ii) The electrophilicity index 120596 has a negative sign inthe model which suggests that increased activity canbe achieved by decreasing the electrophilicity of the
acridine derivatives (a high value of electrophilicitydescribes a good electrophile while a small value ofelectrophilicity describes a good nucleophile)
(iii) The dipolemoment 120583 has a positive sign in themodelwhich suggests that increased activity can be achievedby increasing the polarity of the acridine derivatives
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
Advances in Physical Chemistry 5
Table1Con
tinued
Num
ber
R 1R 2
R 3R 4
R 5R 6
R 7pIC 50
b
PRO
cAMA
d
33H
H(C
H2) 2OCO
PhOMe
ClOMe
HH
minus2098
minus2088
34CH2NH2
HH
HH
HH
minus0230
minus053
135
CH2OH
HH
HH
HH
Toxa
minus1515
36CH2NHCO
(CH2) 3Cl
HH
HH
HH
minus1655
minus1427
37CH2OCO
(CH2) 3Cl
HH
HH
HH
minus2200
minus218
538
CH2NHCO
CH=C
H2
HH
HH
HH
minus2245
minus0833
39CH2OCO
CH=C
H2
HH
HH
HH
minus1464
mdash40
CH2NHCO
PhH
HH
HH
Hminus1654
minus112
141
CH2OCO
PhH
HH
HH
Hminus1885
Toxa
42CH2NHCO
-p-PhF
HH
HH
HH
minus1815
mdash43
CH2OCO
-p-PhF
HH
HH
HH
minus174
1To
xa44
CH2NHCO
-p-PhC
lH
HH
HH
HTo
xaminus1004
45CH2OCO
-p-PhC
lH
HH
HH
Hminus1790
minus1433
46CH2NHCO
-p-PhO
Me
HH
HH
HH
minus1513
minus0973
47CH2OCO
-p-PhO
Me
HH
HH
HH
minus1471
minus0869
48CH2NHCO
-p-PhN
Me 2
HH
HH
HH
minus153
9minus0
672
49CH2OCO
-p-PhN
Me 2
HH
HH
HH
minus1819
Toxa
50CH2NH2
HH
HH
HCH2NH2
minus0820
minus053
151
CH2OH
HH
HH
HCH2OH
Toxa
0222
52CH2NHCO
(CH2) 3Cl
HH
HH
HCH2NHCO
(CH2) 3Cl
minus0663
minus0813
54CH2NHCO
CH=C
H2
HH
HH
HCH2NHCO
CHCH2
minus1061
Toxa
56CH2NHCO
PhH
HH
HH
CH2NHCO
Phminus0
556
minus1236
57CH2OCO
PhH
HH
HH
CH2OCO
PhTo
xaminus0
716
58CH2NHCO
-p-PhF
HH
HH
HCH2NHCO
-p-PhF
minus0756
minus0255
60CH2NHCO
-p-PhC
lH
HH
HH
CH2NHCO
-p-PhC
lmdash
minus1562
62CH2NHCO
-p-PhO
Me
HH
HH
HCH2NHCO
-p-PhO
Me
minus1797
Toxa
63CH2OCO
-p-PhO
Me
HH
HH
HCH2OCO
-p-PhO
Me
Toxa
minus1825
64CH2NHCO
-p-PhN
Me 2
HH
HH
HCH2NHCO
-p-PhN
Me 2
minus0940
minus0724
65CH2OCO
-p-PhN
Me 2
HH
HH
HCH2OCO
-p-PhN
Me 2
Toxa
minus1667
a Toxictoxicity
observed
onhu
man
macroph
agesatconcentrations
thatdidno
tdisp
layantileishmanialactivity
-p-para-m
-meta
b pIC50=minuslog(
IC50)
c pIC50AMAantileish
manialactivity
againstamastig
otes
parasiteformdpIC 50PR
Oantileish
manialactivity
againstp
romastig
otes
parasiteform
6 Advances in Physical Chemistry
Inputw
b
w
b
+ +
Hidden layer Output layer
Output
Figure 2 The architecture used in our study of the artificial neural network
each model can predict the activity of any com-pound even with a completely different structurefrom those included in the study Therefore the ADis a tool to find out compounds that are outsidethe applicability domain of the built QSAR modeland it detects outliers present in the training setcompounds There are several methods for definingthe applicability domain (AD) of QSAR models [32]but themost common one is determining the leveragevalues ℎ119894 (ℎ119894 = 119909119879
119894(119883119879 119883)
minus1
119909119894 (119894 = 1 2 119899)) foreach compound where119909119894 is the descriptor row-vectorof query compound 119883 is 119899 119909 119896 minus 1 matrix of 119896model descriptor values for 119899 training set compoundsand the superscript 119879 refers to the transpose ofmatrixvector [32 33] In this study we use Williamsplot in this plot the applicability domain is estab-lished inside a squared area within standard deviationplusmn119909 (in this study 119909 = 3 ldquothree-sigma rulerdquo [34]) anda leverage threshold ℎlowast(ℎlowast = 3 lowast (119896 + 1)119899) [35]where 119899 is the number of training set compounds 119896is the number of model descriptors The leverage (ℎ)greater than the warning leverage (ℎlowast) suggested thatthe compoundwas very influential on themodel [36]The results of the leverage approach were comparedwith that of the simple approach introduced by Royet al [37]
231 Multiple Linear Regression (MLR) The descendentmultiple linear regression (MLR) analysis based on theelimination of aberrant descriptors (one by one) until a validmodel (including the critical probability 119875 value lt 005 forall descriptors and the model complete) was employed tofind a linear model of the activity of interest which takes thefollowing form
119884 = 1198860 +119899
sum119894=1
119886119894119909119894 (1)
where 119884 is the studied activity which is the dependent vari-able 1198860 is the intercept of the equation 119909119894 are the moleculardescriptors 119886119894 are the coefficients of those descriptors
This method is one of the most popular methods ofQSARQSPR thanks to its simplicity in operation repro-ducibility and ability to allow easy interpretation of thefeatures used The important advantage of the linear regres-sion analysis is that it is highly transparent therefore thealgorithm is available and predictions can bemade easily [38]It has served also to select the descriptors used as the inputparameters in the artificial neural network (ANN)
Table 2 Software packages used in this work
Drawing chemicalstructures
Marvin Sketch ACDChemSketchand ChemBioDraw
Generating 3D structures Gauss View 30 and ChemBio3DCalculating chemicaldescriptors
Gaussian 03 Marvin Sketch 62ChemSketch and ChemBio3D
Developing QSAR models XLSTAT 2009 and Matlab 790(version 2011)
232 Artificial Neural Networks (ANN) The artificial neuralnetworks (ANN) are used in order to increase the probabilityof characterizing the compounds and to generate a predic-tive QSAR model between the set of molecular descriptorsobtained from the MLR models and the observed activitiesvalues The ANN model is done on the MATLAB R2009bsoftware It consists of three layers of neurons called inputlayer hidden layer and output layer (Figure 2) The inputlayer formed by a number of neurons equal to the numberof descriptors obtained in the multiple linear regressionmodels and the output layer represents the calculated activityvalues For determination of the number of hidden neuronsin the hidden layer where all calculations of parameteroptimization of neural networks are made a parameter 120588has been proposed The parameter 120588 plays a major role indetermining the best artificial neural network architecture[39 40] 120588 is defined as follows
120588 =number of data points in the training set
sum of the number of connections in the ANN (2)
In order to avoid overfitting or underfitting it is recom-mended that the value of 120588 should be between 100 and 300if 120588 lt 1 the network simply memorizes the data whereas if120588 gt 3 the network is not able to generalize [41]
24 Software Packages Used in Our QSARDevelopment StudyThere are various free and commercial software available forQSAR development These include specialized software fordrawing chemical structures generating 3D structures cal-culating chemical descriptors and developing QSARmodelsThe software packages used in our works are represented inTable 2
3 Results and Discussions
31 Principal Component Analysis (PCA) PCA is a usefulstatistical technique for summarizing all the informationencoded in the structures of the compounds It is very helpful
Advances in Physical Chemistry 7
Table 3The absolute values of correlation coefficients with descrip-tors and activitiesDescriptors 119903Promastigotes
Henryrsquos law constant (119870H) 0130Ideal gas thermal capacity (IGTC) 0108Number of H-Bond acceptors (NHA) 0299Number of H-Bond donors (NHD) 0250Balaban index (119869) 0133Shape coefficient (119868) 0135Total valence connectivity (TVC) 0292N 0238H 0260Gibbs free energy (119866) 0119Dipole moment (120583) 0213Electronegativity (120594) 0599Electrophilicity index (120596) 0630Molecular topological index (MTI) 0116Polar surface area (PSA) 0282Total connectivity (TC) 0269Wiener index (119882) 0124119864HOMO 0554119864LUMO 0632119864Gap 0227
AmastigotesTotal energy 119864 0160119864HOMO 0286Electronegativity (120594) 0316Polar surface area (PSA) 0391O 0105Index of refraction (119899) 0548Density (119889) 0242Critical pressure (CP) 0176Gibbs free energy (119866) 0195Henryrsquos law constant (119870H) 0387Number of H-Bond acceptors (NHA) 0198Number of H-Bond donors (NHD) 0587Number of rotatable bonds (NRB) 0139Sum of valence degrees (SVD) 0146Dipole moment (120583) 0224119864LUMO 0338Electrophilicity index (120596) 0335N 0436C 0155Surface tension (120574) 0553Boiling point (TB) 0254Critical temperature (CT) 0276Heat of formation (119867∘) 0152log119875 0248Melting point (119879) 0314Partition coefficient (PC) 0163Shape coefficient (119868) 0176Winner index (119882) 0103
for understanding the distribution of the compounds Thisis an essentially descriptive statistical method which aimsto extract the maximum of information contained in thedataset compounds [42 43] In this work the PCA is usedto overview the examined compounds for similarities anddissimilarities and to select descriptors that show a high
Table 4 Descriptors selected by the principal component analysis(PCA) and software packages used in the calculation of descriptors
Software Descriptors
Gaussian 03
Highest occupied molecular orbital energy119864HOMO (eV) lowest unoccupied molecularorbital energy 119864LUMO (eV) hardness 120578 (eV)= (119864LUMO minus 119864HOMO)2 electronegativity 120594(eV) = minus(119864LUMO + 119864HOMO)2 electrophilicityindex 120596 (eV) = 12059422120578 total energy 119864 (eV)dipole moment 120583 (Debye) energy gapbetween 119864HOMO and 119864LUMO values 119864Gap (eV)
ChemOffice
Heat of formation119867∘ (kJmolminus1) Gibbs freeenergy 119866 (kJmolminus1) ideal gas thermalcapacity (IGTC) (Jmolminus1 Kminus1) melting point119879 (Kelvin) critical temperature (CT)(Kelvin) boiling point (TB) (Kelvin) criticalpressure (CP) (Bar) Henryrsquos law constant119870H total valence connectivity (TVC)partition coefficient (PC) moleculartopological index (MTI) number ofrotatable bonds (NRB) shape coefficient 119868sum of valence degrees (SVD) totalconnectivity (TC)
ChemSketchPercent ratios of nitrogen hydrogenoxygen and carbon atoms (N H Oand C) surface tension 120574 (dynecm) indexof refraction (119899) density (119889)
Marvin Sketchlog119875 Winner index (119882) number ofH-Bond acceptors (NHA) number ofH-Bond donors (NHD) Balaban index (119869)polar surface area (PSA) (119860∘)2
correlation with the response activity this one gives extraweight because it will be more effective at prediction [44]Tables 3 and 4 present the descriptors with a correlationcoefficient with the activity higher in absolute value than01 The absence of any serious multicollinearity betweenthe descriptors present in the model was confirmed by thecorrelation matrix
32 Univariate Partitioning (UP) The aim of the UP was therecognition of groups of objects based on their similarity thismethod is based on the criterion of partitioning proposedby Fisher [45] In this study the division of the dataset intotraining and test sets has been performed In this one fromeach obtained cluster one compound for the training setwas selected randomly to be used as test set compound Thepartitioning results are given in Table 5
33 Multiple Linear Regression (MLR) The QSAR analysiswas performed using the values of the chemical descriptorsselected by the PCA method and on the other hand theexperimental values of the antileishmanial activities for 60 ofthe acridines derivatives (effect) Table S1 in SupplementaryMaterial available online at httpdxdoiorg10115520165137289 shows the value of each molecular descriptor that isconfigured in established MLR models and the QSAR model
8 Advances in Physical Chemistry
Table 5 The partitioning results of compounds into training and test sets
ClassesThe test set compounds
Promastigotes Amastigotes3 6 10 14 17 21 37 38 42 64 5 6 26 27 29 36 37 44 46 50
1 1 8 37 40 56 1 10 15 16 34 48 50 57 642 2 14 29 34 48 2 25 38 44 47 51 523 3 18 19 3 26 40 564 4 24 25 27 30 42 58 4 7 8 11 13 17 21 465 6 26 5 14 206 7 9 12 21 33 39 41 47 67 10 11 16 31 32 43 45 18 29 588 15 17 23 28 36 49 52 54 27 639 20 38 46 62 24 28 31 35 36 45 60 6510 50 64 32 33 37
built is represented by the following equations and the valuesof the statistical parameters
MLRModel for pIC50 Promastigotes
pIC50(PRO) = 3030 minus 8214 10minus02N + 0239120583
minus 1264120596 minus 0233NHA + 0732NHD
+ 3311 10minus05MTI + 3184 104TVC
(3)
Statistical Parameters 1198772 = 0723 119877 = 0850 1198772adj = 0664MSE = 0189 P value lt 10minus4
MLRModel for pIC50 Amastigotes
pIC50(AMA) = minus7314 + 1191119864HOMO
+ 2543 10minus02CT minus 0350119870H
minus 1632 10minus02119879 + 0652NHA
+ 1262NHD
(4)
Statistical Parameters 1198772 = 0663 119877 = 0814 1198772adj = 0598MSE = 0208 119875 value lt 10minus4
(i) For the two models P value is lower than 00001it means that we would be taking a lower than001 risk in assuming that the null hypothesis (noeffect of the explanatory variables) is wrong and thatthe regressions equations have statistical significanceTherefore we can conclude with confidence that theselected variables do bring a significant amount ofinformation
(ii) Higher value of 1198772 and 1198772adj and lower mean squarederror (MSE) indicate that the two proposed modelsare predictive and reliable
(iii) Our obtained models were validated internally by theleave-one-out cross validation technique the crossvalidation coefficient 1198772cv for the two models was
determined based on the predictive ability of themodel The value of 1198772cv is higher than 05 (1198772cv =0536 for the MLR pIC
50(PRO) model and 1198772cv =
0525 for the MLR pIC50
(AMA) model) indicatinggood predictability of the model
(iv) True predictive power of these models is to test theirability to predict perfectly pIC
50of compounds from
an external test set (compounds that were not usedfor the developed model) pIC
50of the remaining set
of 10 compounds are deduced from the quantitativemodels proposed with the compounds used in train-ing set by MLR These models will be able to predictthe activities of test set molecules in agreement withthe experimentally determined value The observedand calculated pIC
50values are given in Table 6
The predictive capacity of the models was judgedthe higher value of 1198772test (1198772test = 0660 for theMLR pIC
50(PRO) model and 1198772test = 0718 for the
MLR pIC50
(AMA) model) indicates the improvedpredictability of the model
(v) Xternal Validation Plus indicates the absence of sys-tematic errors in the model and a moderate perfor-mance of prediction quality of a QSAR model basedon proposed MAE-based criteria (Table S2)
(vi) The values of calculated activities from (3) and (4) aregiven in Table S1 and the correlations of calculatedand observed activities values are illustrated in Fig-ure 3
In the first model (pIC50
PRO) the descriptors influencingnegatively the activities are the percent ratio of nitrogen(N) the number of H-Bond acceptors (NHA) and theelectrophilicity index (120596) and the parameters influencingpositively the activities are the dipole moment (120583) thenumber ofH-Bonddonors (NHD) themolecular topologicalindex (MTI) and the total valence connectivity (TVC)
(i) The number of H-Bond acceptors (NHA) has a nega-tive sign in the model which suggests that increasedactivity can be achieved by decreasing the number
Advances in Physical Chemistry 9
Table 6 The values of chemical descriptors and the observed and predicted activities using the MLR models for the test set
Test set (promastigotes) pIC50PRO
Number N 120583 120596 NHA NHD MTI TVC Obs MLR3 12030 4259 5024 3 2 14075 3891 10minus07 minus1669 minus20476 9270 5640 4654 5 2 28763 1716 10minus10 minus0230 minus101410 10780 5730 5267 3 2 16459 4902 10minus08 minus1061 minus183214 11760 2698 3062 3 1 4387 3050 10minus05 0301 minus001217 8580 4969 3431 4 1 10329 5188 10minus07 minus0380 minus066421 7180 5214 3498 5 1 18600 1334 10minus08 minus0114 minus055137 4460 6939 4779 2 0 8245 4076 10minus06 minus2200 minus178138 10680 5122 4374 2 1 6388 5083 10minus06 minus2245 minus151242 8480 5260 4457 3 1 12005 1307 10minus07 minus1815 minus160864 13170 5297 4388 5 2 47653 1201 10minus10 minus0940 minus0455
Test set (amastigotes) pIC50AMA
Number 119864HOMO 119870H 119879 NHA NHD CT Obs MLR5 minus5850 106461 1300866 3 2 19110 0699 minus07876 minus5785 100595 1247456 5 2 18715 1523 033226 minus5665 583700 926651 2 0 7815 minus1170 minus145627 minus5603 622270 930050 3 0 8827 minus1877 minus162829 minus5640 618850 938380 3 0 9007 minus1561 minus146736 minus5966 676650 1037977 2 1 12201 minus1427 minus077437 minus6025 552770 930967 2 0 8348 minus2185 minus145744 minus5919 749400 1083064 2 1 12828 minus1004 minus097846 minus5790 752980 1078419 3 1 13925 minus0973 minus073350 minus5461 625750 913563 3 2 13449 minus0531 minus1028
minus05
0
05
1
15
2
25
minus02 03 08 13 18 23
ActivesValidation
ActivesValidation
minus16
minus06
04
14
minus05 0 05 1 15
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 3 Correlations of observed and predicted activities (training set in black and test set in red) values calculated using MLR models
of heteroatoms (nitrogen or oxygen atoms) mostlythe nitrogen ones for decreasing the ratio of nitrogen(N) having also a negative sign in the model
(ii) The electrophilicity index 120596 has a negative sign inthe model which suggests that increased activity canbe achieved by decreasing the electrophilicity of the
acridine derivatives (a high value of electrophilicitydescribes a good electrophile while a small value ofelectrophilicity describes a good nucleophile)
(iii) The dipolemoment 120583 has a positive sign in themodelwhich suggests that increased activity can be achievedby increasing the polarity of the acridine derivatives
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
6 Advances in Physical Chemistry
Inputw
b
w
b
+ +
Hidden layer Output layer
Output
Figure 2 The architecture used in our study of the artificial neural network
each model can predict the activity of any com-pound even with a completely different structurefrom those included in the study Therefore the ADis a tool to find out compounds that are outsidethe applicability domain of the built QSAR modeland it detects outliers present in the training setcompounds There are several methods for definingthe applicability domain (AD) of QSAR models [32]but themost common one is determining the leveragevalues ℎ119894 (ℎ119894 = 119909119879
119894(119883119879 119883)
minus1
119909119894 (119894 = 1 2 119899)) foreach compound where119909119894 is the descriptor row-vectorof query compound 119883 is 119899 119909 119896 minus 1 matrix of 119896model descriptor values for 119899 training set compoundsand the superscript 119879 refers to the transpose ofmatrixvector [32 33] In this study we use Williamsplot in this plot the applicability domain is estab-lished inside a squared area within standard deviationplusmn119909 (in this study 119909 = 3 ldquothree-sigma rulerdquo [34]) anda leverage threshold ℎlowast(ℎlowast = 3 lowast (119896 + 1)119899) [35]where 119899 is the number of training set compounds 119896is the number of model descriptors The leverage (ℎ)greater than the warning leverage (ℎlowast) suggested thatthe compoundwas very influential on themodel [36]The results of the leverage approach were comparedwith that of the simple approach introduced by Royet al [37]
231 Multiple Linear Regression (MLR) The descendentmultiple linear regression (MLR) analysis based on theelimination of aberrant descriptors (one by one) until a validmodel (including the critical probability 119875 value lt 005 forall descriptors and the model complete) was employed tofind a linear model of the activity of interest which takes thefollowing form
119884 = 1198860 +119899
sum119894=1
119886119894119909119894 (1)
where 119884 is the studied activity which is the dependent vari-able 1198860 is the intercept of the equation 119909119894 are the moleculardescriptors 119886119894 are the coefficients of those descriptors
This method is one of the most popular methods ofQSARQSPR thanks to its simplicity in operation repro-ducibility and ability to allow easy interpretation of thefeatures used The important advantage of the linear regres-sion analysis is that it is highly transparent therefore thealgorithm is available and predictions can bemade easily [38]It has served also to select the descriptors used as the inputparameters in the artificial neural network (ANN)
Table 2 Software packages used in this work
Drawing chemicalstructures
Marvin Sketch ACDChemSketchand ChemBioDraw
Generating 3D structures Gauss View 30 and ChemBio3DCalculating chemicaldescriptors
Gaussian 03 Marvin Sketch 62ChemSketch and ChemBio3D
Developing QSAR models XLSTAT 2009 and Matlab 790(version 2011)
232 Artificial Neural Networks (ANN) The artificial neuralnetworks (ANN) are used in order to increase the probabilityof characterizing the compounds and to generate a predic-tive QSAR model between the set of molecular descriptorsobtained from the MLR models and the observed activitiesvalues The ANN model is done on the MATLAB R2009bsoftware It consists of three layers of neurons called inputlayer hidden layer and output layer (Figure 2) The inputlayer formed by a number of neurons equal to the numberof descriptors obtained in the multiple linear regressionmodels and the output layer represents the calculated activityvalues For determination of the number of hidden neuronsin the hidden layer where all calculations of parameteroptimization of neural networks are made a parameter 120588has been proposed The parameter 120588 plays a major role indetermining the best artificial neural network architecture[39 40] 120588 is defined as follows
120588 =number of data points in the training set
sum of the number of connections in the ANN (2)
In order to avoid overfitting or underfitting it is recom-mended that the value of 120588 should be between 100 and 300if 120588 lt 1 the network simply memorizes the data whereas if120588 gt 3 the network is not able to generalize [41]
24 Software Packages Used in Our QSARDevelopment StudyThere are various free and commercial software available forQSAR development These include specialized software fordrawing chemical structures generating 3D structures cal-culating chemical descriptors and developing QSARmodelsThe software packages used in our works are represented inTable 2
3 Results and Discussions
31 Principal Component Analysis (PCA) PCA is a usefulstatistical technique for summarizing all the informationencoded in the structures of the compounds It is very helpful
Advances in Physical Chemistry 7
Table 3The absolute values of correlation coefficients with descrip-tors and activitiesDescriptors 119903Promastigotes
Henryrsquos law constant (119870H) 0130Ideal gas thermal capacity (IGTC) 0108Number of H-Bond acceptors (NHA) 0299Number of H-Bond donors (NHD) 0250Balaban index (119869) 0133Shape coefficient (119868) 0135Total valence connectivity (TVC) 0292N 0238H 0260Gibbs free energy (119866) 0119Dipole moment (120583) 0213Electronegativity (120594) 0599Electrophilicity index (120596) 0630Molecular topological index (MTI) 0116Polar surface area (PSA) 0282Total connectivity (TC) 0269Wiener index (119882) 0124119864HOMO 0554119864LUMO 0632119864Gap 0227
AmastigotesTotal energy 119864 0160119864HOMO 0286Electronegativity (120594) 0316Polar surface area (PSA) 0391O 0105Index of refraction (119899) 0548Density (119889) 0242Critical pressure (CP) 0176Gibbs free energy (119866) 0195Henryrsquos law constant (119870H) 0387Number of H-Bond acceptors (NHA) 0198Number of H-Bond donors (NHD) 0587Number of rotatable bonds (NRB) 0139Sum of valence degrees (SVD) 0146Dipole moment (120583) 0224119864LUMO 0338Electrophilicity index (120596) 0335N 0436C 0155Surface tension (120574) 0553Boiling point (TB) 0254Critical temperature (CT) 0276Heat of formation (119867∘) 0152log119875 0248Melting point (119879) 0314Partition coefficient (PC) 0163Shape coefficient (119868) 0176Winner index (119882) 0103
for understanding the distribution of the compounds Thisis an essentially descriptive statistical method which aimsto extract the maximum of information contained in thedataset compounds [42 43] In this work the PCA is usedto overview the examined compounds for similarities anddissimilarities and to select descriptors that show a high
Table 4 Descriptors selected by the principal component analysis(PCA) and software packages used in the calculation of descriptors
Software Descriptors
Gaussian 03
Highest occupied molecular orbital energy119864HOMO (eV) lowest unoccupied molecularorbital energy 119864LUMO (eV) hardness 120578 (eV)= (119864LUMO minus 119864HOMO)2 electronegativity 120594(eV) = minus(119864LUMO + 119864HOMO)2 electrophilicityindex 120596 (eV) = 12059422120578 total energy 119864 (eV)dipole moment 120583 (Debye) energy gapbetween 119864HOMO and 119864LUMO values 119864Gap (eV)
ChemOffice
Heat of formation119867∘ (kJmolminus1) Gibbs freeenergy 119866 (kJmolminus1) ideal gas thermalcapacity (IGTC) (Jmolminus1 Kminus1) melting point119879 (Kelvin) critical temperature (CT)(Kelvin) boiling point (TB) (Kelvin) criticalpressure (CP) (Bar) Henryrsquos law constant119870H total valence connectivity (TVC)partition coefficient (PC) moleculartopological index (MTI) number ofrotatable bonds (NRB) shape coefficient 119868sum of valence degrees (SVD) totalconnectivity (TC)
ChemSketchPercent ratios of nitrogen hydrogenoxygen and carbon atoms (N H Oand C) surface tension 120574 (dynecm) indexof refraction (119899) density (119889)
Marvin Sketchlog119875 Winner index (119882) number ofH-Bond acceptors (NHA) number ofH-Bond donors (NHD) Balaban index (119869)polar surface area (PSA) (119860∘)2
correlation with the response activity this one gives extraweight because it will be more effective at prediction [44]Tables 3 and 4 present the descriptors with a correlationcoefficient with the activity higher in absolute value than01 The absence of any serious multicollinearity betweenthe descriptors present in the model was confirmed by thecorrelation matrix
32 Univariate Partitioning (UP) The aim of the UP was therecognition of groups of objects based on their similarity thismethod is based on the criterion of partitioning proposedby Fisher [45] In this study the division of the dataset intotraining and test sets has been performed In this one fromeach obtained cluster one compound for the training setwas selected randomly to be used as test set compound Thepartitioning results are given in Table 5
33 Multiple Linear Regression (MLR) The QSAR analysiswas performed using the values of the chemical descriptorsselected by the PCA method and on the other hand theexperimental values of the antileishmanial activities for 60 ofthe acridines derivatives (effect) Table S1 in SupplementaryMaterial available online at httpdxdoiorg10115520165137289 shows the value of each molecular descriptor that isconfigured in established MLR models and the QSAR model
8 Advances in Physical Chemistry
Table 5 The partitioning results of compounds into training and test sets
ClassesThe test set compounds
Promastigotes Amastigotes3 6 10 14 17 21 37 38 42 64 5 6 26 27 29 36 37 44 46 50
1 1 8 37 40 56 1 10 15 16 34 48 50 57 642 2 14 29 34 48 2 25 38 44 47 51 523 3 18 19 3 26 40 564 4 24 25 27 30 42 58 4 7 8 11 13 17 21 465 6 26 5 14 206 7 9 12 21 33 39 41 47 67 10 11 16 31 32 43 45 18 29 588 15 17 23 28 36 49 52 54 27 639 20 38 46 62 24 28 31 35 36 45 60 6510 50 64 32 33 37
built is represented by the following equations and the valuesof the statistical parameters
MLRModel for pIC50 Promastigotes
pIC50(PRO) = 3030 minus 8214 10minus02N + 0239120583
minus 1264120596 minus 0233NHA + 0732NHD
+ 3311 10minus05MTI + 3184 104TVC
(3)
Statistical Parameters 1198772 = 0723 119877 = 0850 1198772adj = 0664MSE = 0189 P value lt 10minus4
MLRModel for pIC50 Amastigotes
pIC50(AMA) = minus7314 + 1191119864HOMO
+ 2543 10minus02CT minus 0350119870H
minus 1632 10minus02119879 + 0652NHA
+ 1262NHD
(4)
Statistical Parameters 1198772 = 0663 119877 = 0814 1198772adj = 0598MSE = 0208 119875 value lt 10minus4
(i) For the two models P value is lower than 00001it means that we would be taking a lower than001 risk in assuming that the null hypothesis (noeffect of the explanatory variables) is wrong and thatthe regressions equations have statistical significanceTherefore we can conclude with confidence that theselected variables do bring a significant amount ofinformation
(ii) Higher value of 1198772 and 1198772adj and lower mean squarederror (MSE) indicate that the two proposed modelsare predictive and reliable
(iii) Our obtained models were validated internally by theleave-one-out cross validation technique the crossvalidation coefficient 1198772cv for the two models was
determined based on the predictive ability of themodel The value of 1198772cv is higher than 05 (1198772cv =0536 for the MLR pIC
50(PRO) model and 1198772cv =
0525 for the MLR pIC50
(AMA) model) indicatinggood predictability of the model
(iv) True predictive power of these models is to test theirability to predict perfectly pIC
50of compounds from
an external test set (compounds that were not usedfor the developed model) pIC
50of the remaining set
of 10 compounds are deduced from the quantitativemodels proposed with the compounds used in train-ing set by MLR These models will be able to predictthe activities of test set molecules in agreement withthe experimentally determined value The observedand calculated pIC
50values are given in Table 6
The predictive capacity of the models was judgedthe higher value of 1198772test (1198772test = 0660 for theMLR pIC
50(PRO) model and 1198772test = 0718 for the
MLR pIC50
(AMA) model) indicates the improvedpredictability of the model
(v) Xternal Validation Plus indicates the absence of sys-tematic errors in the model and a moderate perfor-mance of prediction quality of a QSAR model basedon proposed MAE-based criteria (Table S2)
(vi) The values of calculated activities from (3) and (4) aregiven in Table S1 and the correlations of calculatedand observed activities values are illustrated in Fig-ure 3
In the first model (pIC50
PRO) the descriptors influencingnegatively the activities are the percent ratio of nitrogen(N) the number of H-Bond acceptors (NHA) and theelectrophilicity index (120596) and the parameters influencingpositively the activities are the dipole moment (120583) thenumber ofH-Bonddonors (NHD) themolecular topologicalindex (MTI) and the total valence connectivity (TVC)
(i) The number of H-Bond acceptors (NHA) has a nega-tive sign in the model which suggests that increasedactivity can be achieved by decreasing the number
Advances in Physical Chemistry 9
Table 6 The values of chemical descriptors and the observed and predicted activities using the MLR models for the test set
Test set (promastigotes) pIC50PRO
Number N 120583 120596 NHA NHD MTI TVC Obs MLR3 12030 4259 5024 3 2 14075 3891 10minus07 minus1669 minus20476 9270 5640 4654 5 2 28763 1716 10minus10 minus0230 minus101410 10780 5730 5267 3 2 16459 4902 10minus08 minus1061 minus183214 11760 2698 3062 3 1 4387 3050 10minus05 0301 minus001217 8580 4969 3431 4 1 10329 5188 10minus07 minus0380 minus066421 7180 5214 3498 5 1 18600 1334 10minus08 minus0114 minus055137 4460 6939 4779 2 0 8245 4076 10minus06 minus2200 minus178138 10680 5122 4374 2 1 6388 5083 10minus06 minus2245 minus151242 8480 5260 4457 3 1 12005 1307 10minus07 minus1815 minus160864 13170 5297 4388 5 2 47653 1201 10minus10 minus0940 minus0455
Test set (amastigotes) pIC50AMA
Number 119864HOMO 119870H 119879 NHA NHD CT Obs MLR5 minus5850 106461 1300866 3 2 19110 0699 minus07876 minus5785 100595 1247456 5 2 18715 1523 033226 minus5665 583700 926651 2 0 7815 minus1170 minus145627 minus5603 622270 930050 3 0 8827 minus1877 minus162829 minus5640 618850 938380 3 0 9007 minus1561 minus146736 minus5966 676650 1037977 2 1 12201 minus1427 minus077437 minus6025 552770 930967 2 0 8348 minus2185 minus145744 minus5919 749400 1083064 2 1 12828 minus1004 minus097846 minus5790 752980 1078419 3 1 13925 minus0973 minus073350 minus5461 625750 913563 3 2 13449 minus0531 minus1028
minus05
0
05
1
15
2
25
minus02 03 08 13 18 23
ActivesValidation
ActivesValidation
minus16
minus06
04
14
minus05 0 05 1 15
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 3 Correlations of observed and predicted activities (training set in black and test set in red) values calculated using MLR models
of heteroatoms (nitrogen or oxygen atoms) mostlythe nitrogen ones for decreasing the ratio of nitrogen(N) having also a negative sign in the model
(ii) The electrophilicity index 120596 has a negative sign inthe model which suggests that increased activity canbe achieved by decreasing the electrophilicity of the
acridine derivatives (a high value of electrophilicitydescribes a good electrophile while a small value ofelectrophilicity describes a good nucleophile)
(iii) The dipolemoment 120583 has a positive sign in themodelwhich suggests that increased activity can be achievedby increasing the polarity of the acridine derivatives
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
Advances in Physical Chemistry 7
Table 3The absolute values of correlation coefficients with descrip-tors and activitiesDescriptors 119903Promastigotes
Henryrsquos law constant (119870H) 0130Ideal gas thermal capacity (IGTC) 0108Number of H-Bond acceptors (NHA) 0299Number of H-Bond donors (NHD) 0250Balaban index (119869) 0133Shape coefficient (119868) 0135Total valence connectivity (TVC) 0292N 0238H 0260Gibbs free energy (119866) 0119Dipole moment (120583) 0213Electronegativity (120594) 0599Electrophilicity index (120596) 0630Molecular topological index (MTI) 0116Polar surface area (PSA) 0282Total connectivity (TC) 0269Wiener index (119882) 0124119864HOMO 0554119864LUMO 0632119864Gap 0227
AmastigotesTotal energy 119864 0160119864HOMO 0286Electronegativity (120594) 0316Polar surface area (PSA) 0391O 0105Index of refraction (119899) 0548Density (119889) 0242Critical pressure (CP) 0176Gibbs free energy (119866) 0195Henryrsquos law constant (119870H) 0387Number of H-Bond acceptors (NHA) 0198Number of H-Bond donors (NHD) 0587Number of rotatable bonds (NRB) 0139Sum of valence degrees (SVD) 0146Dipole moment (120583) 0224119864LUMO 0338Electrophilicity index (120596) 0335N 0436C 0155Surface tension (120574) 0553Boiling point (TB) 0254Critical temperature (CT) 0276Heat of formation (119867∘) 0152log119875 0248Melting point (119879) 0314Partition coefficient (PC) 0163Shape coefficient (119868) 0176Winner index (119882) 0103
for understanding the distribution of the compounds Thisis an essentially descriptive statistical method which aimsto extract the maximum of information contained in thedataset compounds [42 43] In this work the PCA is usedto overview the examined compounds for similarities anddissimilarities and to select descriptors that show a high
Table 4 Descriptors selected by the principal component analysis(PCA) and software packages used in the calculation of descriptors
Software Descriptors
Gaussian 03
Highest occupied molecular orbital energy119864HOMO (eV) lowest unoccupied molecularorbital energy 119864LUMO (eV) hardness 120578 (eV)= (119864LUMO minus 119864HOMO)2 electronegativity 120594(eV) = minus(119864LUMO + 119864HOMO)2 electrophilicityindex 120596 (eV) = 12059422120578 total energy 119864 (eV)dipole moment 120583 (Debye) energy gapbetween 119864HOMO and 119864LUMO values 119864Gap (eV)
ChemOffice
Heat of formation119867∘ (kJmolminus1) Gibbs freeenergy 119866 (kJmolminus1) ideal gas thermalcapacity (IGTC) (Jmolminus1 Kminus1) melting point119879 (Kelvin) critical temperature (CT)(Kelvin) boiling point (TB) (Kelvin) criticalpressure (CP) (Bar) Henryrsquos law constant119870H total valence connectivity (TVC)partition coefficient (PC) moleculartopological index (MTI) number ofrotatable bonds (NRB) shape coefficient 119868sum of valence degrees (SVD) totalconnectivity (TC)
ChemSketchPercent ratios of nitrogen hydrogenoxygen and carbon atoms (N H Oand C) surface tension 120574 (dynecm) indexof refraction (119899) density (119889)
Marvin Sketchlog119875 Winner index (119882) number ofH-Bond acceptors (NHA) number ofH-Bond donors (NHD) Balaban index (119869)polar surface area (PSA) (119860∘)2
correlation with the response activity this one gives extraweight because it will be more effective at prediction [44]Tables 3 and 4 present the descriptors with a correlationcoefficient with the activity higher in absolute value than01 The absence of any serious multicollinearity betweenthe descriptors present in the model was confirmed by thecorrelation matrix
32 Univariate Partitioning (UP) The aim of the UP was therecognition of groups of objects based on their similarity thismethod is based on the criterion of partitioning proposedby Fisher [45] In this study the division of the dataset intotraining and test sets has been performed In this one fromeach obtained cluster one compound for the training setwas selected randomly to be used as test set compound Thepartitioning results are given in Table 5
33 Multiple Linear Regression (MLR) The QSAR analysiswas performed using the values of the chemical descriptorsselected by the PCA method and on the other hand theexperimental values of the antileishmanial activities for 60 ofthe acridines derivatives (effect) Table S1 in SupplementaryMaterial available online at httpdxdoiorg10115520165137289 shows the value of each molecular descriptor that isconfigured in established MLR models and the QSAR model
8 Advances in Physical Chemistry
Table 5 The partitioning results of compounds into training and test sets
ClassesThe test set compounds
Promastigotes Amastigotes3 6 10 14 17 21 37 38 42 64 5 6 26 27 29 36 37 44 46 50
1 1 8 37 40 56 1 10 15 16 34 48 50 57 642 2 14 29 34 48 2 25 38 44 47 51 523 3 18 19 3 26 40 564 4 24 25 27 30 42 58 4 7 8 11 13 17 21 465 6 26 5 14 206 7 9 12 21 33 39 41 47 67 10 11 16 31 32 43 45 18 29 588 15 17 23 28 36 49 52 54 27 639 20 38 46 62 24 28 31 35 36 45 60 6510 50 64 32 33 37
built is represented by the following equations and the valuesof the statistical parameters
MLRModel for pIC50 Promastigotes
pIC50(PRO) = 3030 minus 8214 10minus02N + 0239120583
minus 1264120596 minus 0233NHA + 0732NHD
+ 3311 10minus05MTI + 3184 104TVC
(3)
Statistical Parameters 1198772 = 0723 119877 = 0850 1198772adj = 0664MSE = 0189 P value lt 10minus4
MLRModel for pIC50 Amastigotes
pIC50(AMA) = minus7314 + 1191119864HOMO
+ 2543 10minus02CT minus 0350119870H
minus 1632 10minus02119879 + 0652NHA
+ 1262NHD
(4)
Statistical Parameters 1198772 = 0663 119877 = 0814 1198772adj = 0598MSE = 0208 119875 value lt 10minus4
(i) For the two models P value is lower than 00001it means that we would be taking a lower than001 risk in assuming that the null hypothesis (noeffect of the explanatory variables) is wrong and thatthe regressions equations have statistical significanceTherefore we can conclude with confidence that theselected variables do bring a significant amount ofinformation
(ii) Higher value of 1198772 and 1198772adj and lower mean squarederror (MSE) indicate that the two proposed modelsare predictive and reliable
(iii) Our obtained models were validated internally by theleave-one-out cross validation technique the crossvalidation coefficient 1198772cv for the two models was
determined based on the predictive ability of themodel The value of 1198772cv is higher than 05 (1198772cv =0536 for the MLR pIC
50(PRO) model and 1198772cv =
0525 for the MLR pIC50
(AMA) model) indicatinggood predictability of the model
(iv) True predictive power of these models is to test theirability to predict perfectly pIC
50of compounds from
an external test set (compounds that were not usedfor the developed model) pIC
50of the remaining set
of 10 compounds are deduced from the quantitativemodels proposed with the compounds used in train-ing set by MLR These models will be able to predictthe activities of test set molecules in agreement withthe experimentally determined value The observedand calculated pIC
50values are given in Table 6
The predictive capacity of the models was judgedthe higher value of 1198772test (1198772test = 0660 for theMLR pIC
50(PRO) model and 1198772test = 0718 for the
MLR pIC50
(AMA) model) indicates the improvedpredictability of the model
(v) Xternal Validation Plus indicates the absence of sys-tematic errors in the model and a moderate perfor-mance of prediction quality of a QSAR model basedon proposed MAE-based criteria (Table S2)
(vi) The values of calculated activities from (3) and (4) aregiven in Table S1 and the correlations of calculatedand observed activities values are illustrated in Fig-ure 3
In the first model (pIC50
PRO) the descriptors influencingnegatively the activities are the percent ratio of nitrogen(N) the number of H-Bond acceptors (NHA) and theelectrophilicity index (120596) and the parameters influencingpositively the activities are the dipole moment (120583) thenumber ofH-Bonddonors (NHD) themolecular topologicalindex (MTI) and the total valence connectivity (TVC)
(i) The number of H-Bond acceptors (NHA) has a nega-tive sign in the model which suggests that increasedactivity can be achieved by decreasing the number
Advances in Physical Chemistry 9
Table 6 The values of chemical descriptors and the observed and predicted activities using the MLR models for the test set
Test set (promastigotes) pIC50PRO
Number N 120583 120596 NHA NHD MTI TVC Obs MLR3 12030 4259 5024 3 2 14075 3891 10minus07 minus1669 minus20476 9270 5640 4654 5 2 28763 1716 10minus10 minus0230 minus101410 10780 5730 5267 3 2 16459 4902 10minus08 minus1061 minus183214 11760 2698 3062 3 1 4387 3050 10minus05 0301 minus001217 8580 4969 3431 4 1 10329 5188 10minus07 minus0380 minus066421 7180 5214 3498 5 1 18600 1334 10minus08 minus0114 minus055137 4460 6939 4779 2 0 8245 4076 10minus06 minus2200 minus178138 10680 5122 4374 2 1 6388 5083 10minus06 minus2245 minus151242 8480 5260 4457 3 1 12005 1307 10minus07 minus1815 minus160864 13170 5297 4388 5 2 47653 1201 10minus10 minus0940 minus0455
Test set (amastigotes) pIC50AMA
Number 119864HOMO 119870H 119879 NHA NHD CT Obs MLR5 minus5850 106461 1300866 3 2 19110 0699 minus07876 minus5785 100595 1247456 5 2 18715 1523 033226 minus5665 583700 926651 2 0 7815 minus1170 minus145627 minus5603 622270 930050 3 0 8827 minus1877 minus162829 minus5640 618850 938380 3 0 9007 minus1561 minus146736 minus5966 676650 1037977 2 1 12201 minus1427 minus077437 minus6025 552770 930967 2 0 8348 minus2185 minus145744 minus5919 749400 1083064 2 1 12828 minus1004 minus097846 minus5790 752980 1078419 3 1 13925 minus0973 minus073350 minus5461 625750 913563 3 2 13449 minus0531 minus1028
minus05
0
05
1
15
2
25
minus02 03 08 13 18 23
ActivesValidation
ActivesValidation
minus16
minus06
04
14
minus05 0 05 1 15
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 3 Correlations of observed and predicted activities (training set in black and test set in red) values calculated using MLR models
of heteroatoms (nitrogen or oxygen atoms) mostlythe nitrogen ones for decreasing the ratio of nitrogen(N) having also a negative sign in the model
(ii) The electrophilicity index 120596 has a negative sign inthe model which suggests that increased activity canbe achieved by decreasing the electrophilicity of the
acridine derivatives (a high value of electrophilicitydescribes a good electrophile while a small value ofelectrophilicity describes a good nucleophile)
(iii) The dipolemoment 120583 has a positive sign in themodelwhich suggests that increased activity can be achievedby increasing the polarity of the acridine derivatives
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
8 Advances in Physical Chemistry
Table 5 The partitioning results of compounds into training and test sets
ClassesThe test set compounds
Promastigotes Amastigotes3 6 10 14 17 21 37 38 42 64 5 6 26 27 29 36 37 44 46 50
1 1 8 37 40 56 1 10 15 16 34 48 50 57 642 2 14 29 34 48 2 25 38 44 47 51 523 3 18 19 3 26 40 564 4 24 25 27 30 42 58 4 7 8 11 13 17 21 465 6 26 5 14 206 7 9 12 21 33 39 41 47 67 10 11 16 31 32 43 45 18 29 588 15 17 23 28 36 49 52 54 27 639 20 38 46 62 24 28 31 35 36 45 60 6510 50 64 32 33 37
built is represented by the following equations and the valuesof the statistical parameters
MLRModel for pIC50 Promastigotes
pIC50(PRO) = 3030 minus 8214 10minus02N + 0239120583
minus 1264120596 minus 0233NHA + 0732NHD
+ 3311 10minus05MTI + 3184 104TVC
(3)
Statistical Parameters 1198772 = 0723 119877 = 0850 1198772adj = 0664MSE = 0189 P value lt 10minus4
MLRModel for pIC50 Amastigotes
pIC50(AMA) = minus7314 + 1191119864HOMO
+ 2543 10minus02CT minus 0350119870H
minus 1632 10minus02119879 + 0652NHA
+ 1262NHD
(4)
Statistical Parameters 1198772 = 0663 119877 = 0814 1198772adj = 0598MSE = 0208 119875 value lt 10minus4
(i) For the two models P value is lower than 00001it means that we would be taking a lower than001 risk in assuming that the null hypothesis (noeffect of the explanatory variables) is wrong and thatthe regressions equations have statistical significanceTherefore we can conclude with confidence that theselected variables do bring a significant amount ofinformation
(ii) Higher value of 1198772 and 1198772adj and lower mean squarederror (MSE) indicate that the two proposed modelsare predictive and reliable
(iii) Our obtained models were validated internally by theleave-one-out cross validation technique the crossvalidation coefficient 1198772cv for the two models was
determined based on the predictive ability of themodel The value of 1198772cv is higher than 05 (1198772cv =0536 for the MLR pIC
50(PRO) model and 1198772cv =
0525 for the MLR pIC50
(AMA) model) indicatinggood predictability of the model
(iv) True predictive power of these models is to test theirability to predict perfectly pIC
50of compounds from
an external test set (compounds that were not usedfor the developed model) pIC
50of the remaining set
of 10 compounds are deduced from the quantitativemodels proposed with the compounds used in train-ing set by MLR These models will be able to predictthe activities of test set molecules in agreement withthe experimentally determined value The observedand calculated pIC
50values are given in Table 6
The predictive capacity of the models was judgedthe higher value of 1198772test (1198772test = 0660 for theMLR pIC
50(PRO) model and 1198772test = 0718 for the
MLR pIC50
(AMA) model) indicates the improvedpredictability of the model
(v) Xternal Validation Plus indicates the absence of sys-tematic errors in the model and a moderate perfor-mance of prediction quality of a QSAR model basedon proposed MAE-based criteria (Table S2)
(vi) The values of calculated activities from (3) and (4) aregiven in Table S1 and the correlations of calculatedand observed activities values are illustrated in Fig-ure 3
In the first model (pIC50
PRO) the descriptors influencingnegatively the activities are the percent ratio of nitrogen(N) the number of H-Bond acceptors (NHA) and theelectrophilicity index (120596) and the parameters influencingpositively the activities are the dipole moment (120583) thenumber ofH-Bonddonors (NHD) themolecular topologicalindex (MTI) and the total valence connectivity (TVC)
(i) The number of H-Bond acceptors (NHA) has a nega-tive sign in the model which suggests that increasedactivity can be achieved by decreasing the number
Advances in Physical Chemistry 9
Table 6 The values of chemical descriptors and the observed and predicted activities using the MLR models for the test set
Test set (promastigotes) pIC50PRO
Number N 120583 120596 NHA NHD MTI TVC Obs MLR3 12030 4259 5024 3 2 14075 3891 10minus07 minus1669 minus20476 9270 5640 4654 5 2 28763 1716 10minus10 minus0230 minus101410 10780 5730 5267 3 2 16459 4902 10minus08 minus1061 minus183214 11760 2698 3062 3 1 4387 3050 10minus05 0301 minus001217 8580 4969 3431 4 1 10329 5188 10minus07 minus0380 minus066421 7180 5214 3498 5 1 18600 1334 10minus08 minus0114 minus055137 4460 6939 4779 2 0 8245 4076 10minus06 minus2200 minus178138 10680 5122 4374 2 1 6388 5083 10minus06 minus2245 minus151242 8480 5260 4457 3 1 12005 1307 10minus07 minus1815 minus160864 13170 5297 4388 5 2 47653 1201 10minus10 minus0940 minus0455
Test set (amastigotes) pIC50AMA
Number 119864HOMO 119870H 119879 NHA NHD CT Obs MLR5 minus5850 106461 1300866 3 2 19110 0699 minus07876 minus5785 100595 1247456 5 2 18715 1523 033226 minus5665 583700 926651 2 0 7815 minus1170 minus145627 minus5603 622270 930050 3 0 8827 minus1877 minus162829 minus5640 618850 938380 3 0 9007 minus1561 minus146736 minus5966 676650 1037977 2 1 12201 minus1427 minus077437 minus6025 552770 930967 2 0 8348 minus2185 minus145744 minus5919 749400 1083064 2 1 12828 minus1004 minus097846 minus5790 752980 1078419 3 1 13925 minus0973 minus073350 minus5461 625750 913563 3 2 13449 minus0531 minus1028
minus05
0
05
1
15
2
25
minus02 03 08 13 18 23
ActivesValidation
ActivesValidation
minus16
minus06
04
14
minus05 0 05 1 15
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 3 Correlations of observed and predicted activities (training set in black and test set in red) values calculated using MLR models
of heteroatoms (nitrogen or oxygen atoms) mostlythe nitrogen ones for decreasing the ratio of nitrogen(N) having also a negative sign in the model
(ii) The electrophilicity index 120596 has a negative sign inthe model which suggests that increased activity canbe achieved by decreasing the electrophilicity of the
acridine derivatives (a high value of electrophilicitydescribes a good electrophile while a small value ofelectrophilicity describes a good nucleophile)
(iii) The dipolemoment 120583 has a positive sign in themodelwhich suggests that increased activity can be achievedby increasing the polarity of the acridine derivatives
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
Advances in Physical Chemistry 9
Table 6 The values of chemical descriptors and the observed and predicted activities using the MLR models for the test set
Test set (promastigotes) pIC50PRO
Number N 120583 120596 NHA NHD MTI TVC Obs MLR3 12030 4259 5024 3 2 14075 3891 10minus07 minus1669 minus20476 9270 5640 4654 5 2 28763 1716 10minus10 minus0230 minus101410 10780 5730 5267 3 2 16459 4902 10minus08 minus1061 minus183214 11760 2698 3062 3 1 4387 3050 10minus05 0301 minus001217 8580 4969 3431 4 1 10329 5188 10minus07 minus0380 minus066421 7180 5214 3498 5 1 18600 1334 10minus08 minus0114 minus055137 4460 6939 4779 2 0 8245 4076 10minus06 minus2200 minus178138 10680 5122 4374 2 1 6388 5083 10minus06 minus2245 minus151242 8480 5260 4457 3 1 12005 1307 10minus07 minus1815 minus160864 13170 5297 4388 5 2 47653 1201 10minus10 minus0940 minus0455
Test set (amastigotes) pIC50AMA
Number 119864HOMO 119870H 119879 NHA NHD CT Obs MLR5 minus5850 106461 1300866 3 2 19110 0699 minus07876 minus5785 100595 1247456 5 2 18715 1523 033226 minus5665 583700 926651 2 0 7815 minus1170 minus145627 minus5603 622270 930050 3 0 8827 minus1877 minus162829 minus5640 618850 938380 3 0 9007 minus1561 minus146736 minus5966 676650 1037977 2 1 12201 minus1427 minus077437 minus6025 552770 930967 2 0 8348 minus2185 minus145744 minus5919 749400 1083064 2 1 12828 minus1004 minus097846 minus5790 752980 1078419 3 1 13925 minus0973 minus073350 minus5461 625750 913563 3 2 13449 minus0531 minus1028
minus05
0
05
1
15
2
25
minus02 03 08 13 18 23
ActivesValidation
ActivesValidation
minus16
minus06
04
14
minus05 0 05 1 15
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 3 Correlations of observed and predicted activities (training set in black and test set in red) values calculated using MLR models
of heteroatoms (nitrogen or oxygen atoms) mostlythe nitrogen ones for decreasing the ratio of nitrogen(N) having also a negative sign in the model
(ii) The electrophilicity index 120596 has a negative sign inthe model which suggests that increased activity canbe achieved by decreasing the electrophilicity of the
acridine derivatives (a high value of electrophilicitydescribes a good electrophile while a small value ofelectrophilicity describes a good nucleophile)
(iii) The dipolemoment 120583 has a positive sign in themodelwhich suggests that increased activity can be achievedby increasing the polarity of the acridine derivatives
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
10 Advances in Physical Chemistry
(iv) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomwith one or more hydrogen atoms
(v) The molecular topological index (MTI) coefficient(topological parameter) has a negative sign in themodel which suggests that increased activity can beachieved by increasing the flexibility of the substituentside chain
(vi) The total valence connectivity (TVC) (topologicalparameter) has a positive sign in the model whichsuggests that increased activity can be achieved byincreasing in branching of the acridine derivativesbecause the overall connectivity increases with bothmolecule size and complexity as expressed in branch-ing and cyclicity of molecular skeleton [46]
In the second model (pIC50
AMA) the descriptors influ-encing negatively the activity are Henryrsquos law constant (119870H)and the melting point (119879) and the parameters influencingpositively the activities are 119864HOMO energy the critical tem-perature (CT) the number of H-Bond acceptors (NHA) andthe number of H-Bond donors (NHD)
(i) Henryrsquos law constant (119870H) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing solubility consequently bydecreasing the polarity 119870H = 119888119901 (119888 solubility 119901pressure)
(ii) The melting point (119879) has a negative sign in themodel which suggests that increased activity can beachieved by decreasing the polarity and the branch-ing of molecule (increasing branching makes themolecule less compact As the surface area of themolecule decreases it will becomemore compact andthus easier to pack)
(iii) The highest occupied molecular orbital energy119864HOMO (negative values) has a positive sign in themodel which suggests that the higher of 119864HOMO theweaker donating electron ability is showing the factthat the nucleophilic reaction occurs more easily andthe activity of the compound is higher
(iv) The critical temperature (CT) has a positive sign inthe model which suggests that increased activity canbe achieved by increasing the critical temperature
(v) The number of H-Bond acceptors (NHA) has a posi-tive sign in the model which suggests that increasedactivity can be achieved by increasing the number ofheteroatoms (nitrogen or oxygen atoms)
(vi) The number of H-Bond donors (NHD) has a positivesign in themodel which suggests that increased activ-ity can be achieved by increasing of the heteroatomswith one or more hydrogen atoms
Comparing the importance of each descriptor on pIC50
ofacridines we must know the standardized coefficient and the119905-test values of them in the MLR equations The bigger the
absolute value of the 119905-test value is the greater the influenceof the descriptor is
In (3) the 119905-test values are 1773 minus4035 7408 1720minus3361 minus2429 and minus4570 for N 120583 120596 NHA NHD MTIand TVC respectively Moreover in (4) the 119905-test values are3196 3556minus3764minus3802minus3838 andminus6012 for119879119870H CT119864HOMO NHA and NHD respectively
This meant that the 119905-test values of 120596 120583 NHD andTVC are both larger than those of the other descriptorswhich indicates that in this model the influence of thesedescriptors on activity is stronger than that of the others Itshows also the importance of electrophilicity index in theprediction of pIC
50PRO Moreover the 119905-test value of NHD
is larger than that of the other descriptors which indicatesthat the influence of this descriptor in this model on activityis stronger than that of the others
In the conclusion these results illustrate that to increasethe antileishmanial activity against promastigotes parasiteswe will decrease the electrophilicity and increase the branch-ing the polarity and the number of hydrogen atoms attachedin the heteroatom of the acridine derivatives Moreoverto increase the antileishmanial activity against amastigotesparasites we will decrease the solubility the polarity and thebranching and increase the electrophilicity and the numberof heteroatoms mostly attached in the hydrogen atoms andthe critical temperature of the acridine derivatives
(i) Detection of Outliners Applicability Domain (AD) Theapplicability domain (AD) of these models was evaluated byleverage analysis expressed as Williams plot (Figures 4 and5) in which the standardized residuals (119903) and the leveragethreshold values (ℎlowast = 0585 and 0553 for pIC
50(PRO)
and pIC50
(AMA) resp) were plotted Any new value ofpredicted pIC
50data must be considered reliable only for
those compounds that fall within this ADonwhich themodelwas constructed
From these figures it is obvious that there is no responseoutlier in training set and no response outside in test set Onlyone chemical is identified as outside forMLRmodel for pIC
50
PROand two chemicals forMLRmodel for pIC50AMA these
outsides compounds are given as follows
For pIC50 PRO Compound number 64 in test set has higherleverage which is greater than ℎlowast value of 0585 and all thecompounds have a standard deviation into plusmn119909 interval (119909 =3)
For pIC50 AMA Compounds numbers 5 and 6 in test set havea standard deviation value greater than the plusmn119909 interval (119909 =3)
These results are confirmed using the simple approach(Tables S3 and S4) the chemical number 8 is identified asoutlier and the chemical number 64 is identified as outsideforMLRmodel for pIC
50PRO and any chemical is identified
as outside or outlier for MLR model for pIC50AMA
These erroneous predictions could probably be attributedto wrong experimental data or to the structure of theseoutsides Cl F or NMe2 substitutes the tree compoundsat the para position of the phenyl ring (Figure 6) maybe
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
Advances in Physical Chemistry 11
0 02 04 06
minus2
minus15
minus1
minus05
0
05
1
15
2
12
4
7
8
9
11
12
15
16
18
19
20
23
24
25
26
27
28
29
30
3132
3334
36
39
40
41
43
45
46
47
48
49
50
52
54
56
58
62
3
6 10
1417
21
37
38
42
64
William plot train samples in black test samples in red
Leverage
Stan
dard
ized
resid
uals
Figure 4 Williams plot of standardized residual versus leverage forpIC50(PRO)MLRmodel (with ℎlowast = 0585 and residual limits= plusmn3)
the selected descriptors do not pay much attention to thesespecial substructures The predictions of tree compounds areextrapolations of themodel but fortunately they are all ldquogoodleveragerdquo chemicals
34 Artificial Neural Network (ANN) In this study we used afeed forward network with two layers with a sigmoid transferfunction in the hidden layer and a linear transfer function inthe output layer (Figure 2) The architecture of the artificialneural network used in this work was 7-3-1 and 120588 = 182 forpIC50(PRO) and 6-3-1 and 120588 = 2 for pIC
50(AMA)
The output layer represents the calculated activity values(predicted pIC
50)The calculation result and the performance
of established models are recorded in the output layerTo justify the predictive quality of models total data
are distributed randomly into three groups The first group(70 of the total data) used to drive the system The secondgroup (15 of the total data) will be used to validate thenetwork and the remaining 15 that did not participate inthe learning models will be used as an independent test ofnetwork generalizationThe distribution of the total data intotraining validation and test sets is shown in Table 7
The correlation between the experimental and calculatedvalues using the artificial neural network models is highlysignificant as illustrated in Figure 7 and as indicated bybetter 119877 and 1198772 and the small MSE values for all threephases training validation and test (Table 7) The predictedactivities calculated with the artificial neural network and theobserved values are given in Table S5
The results obtained byMLR and ANN are very sufficientto conclude the performance of the models A comparison
0 01 02 03 04 05
minus4
minus3
minus2
minus1
0
1
2
12
3
47
8
1011
13
14
15
16
17
18
20
21
24
2528
31 32
33
34
35
38
40
45
47
48
5152
56
57
58
60
63
6465
5
6
26
2729
36 37
44
46
50
William plot train samples in black test samples in red
LeverageSt
anda
rdiz
ed re
sidua
ls
Figure 5 Williams plot of standardized residual versus leverage forpIC50
(AMA) MLR model (with ℎlowast = 0553 and residual limits =plusmn3)
of the quality of the statistical terms of these models showsthat the ANN has substantially better predictive capabilityANN was able to establish a more satisfactory relationshipbetween the molecular descriptors and the activity of thestudied compounds compared toMLR but themost negativeside of this method is the fact that it is poorly transparentwhereas transparency of MLR approach gives the most inter-pretable results and gives a good explanation of activities withdescriptors Consequently we can design new compoundswith improved values of activity compared to the studiedcompounds using the MLR models Taking into accountthe above results we added suitable substitutions and thenwe moved to calculate their activities using the proposedmodels equations ((3) and (4)) Therefore the suggestedmodels will reduce the time and cost of synthesis as wellas the determination of the antileishmanial activities againstpromastigotes and amastigotes forms of parasites for theacridine derivatives
35 Design New Compounds with Higher AntileishmanialActivities According to the above discussions the MLRmodels could be applied to other acridines derivativesaccording to Table 1 and could add further knowledge in theimprovement of new way in antileishmanial drug researchIf we develop a new compound with better values than theexisting ones it may give rise to the development of moreactive compounds than those currently in use
In this way we carried out structural modification start-ing from compounds having the highest pIC
50values as
template (number 14 and number 20) The structures of thedesigned compounds and their parameter values calculated
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
12 Advances in Physical Chemistry
N
OO
Cl Cl
N
OO
F F
N
HN NH
O O
N N
Me
Me
Me
Me
Compound number 5 Compound number 6
Compound number 64
Figure 6 Chemical structures of the outsides compounds
minus05
0
05
1
15
2
25
minus05 0 05 1 15 2
TrainingValidationTest
TrainingValidationTest
minus17
minus12
minus07
minus02
03
08
13
18
23
minus2 minus1 0 1 2
Pred (pIC50 PRO)pIC50 PRO Pred (pIC50 AMA)pIC50 AMA
pIC
50
PRO
Pred (pIC50 PRO)
pIC
50
AM
A
Pred (pIC50 AMA)
Figure 7 Correlations between observed and predicted activities values calculated using ANN models (training set in black validation setin red and test set in blue)
Table 7The coefficient of determination the coe1fficient of correlation and the MSE obtained by the model established by the ANN for thethree phases Training validation and test
pIC50PRO
Model 119899 = 51 MSE = 0079 1198772 = 0870 119877 = 0933
Training set [1 3 4 5 6 7 11 12 13 14 16 19 21 25 26 27 28 30 3234 35 36 37 38 39 40 41 43 44 46 47 48 49 50 51] MSE = 0044 1198772 = 0914 119877 = 0956
Validation set [8 17 18 22 23 24 29 31] MSE = 0125 1198772 = 0922 119877 = 0960Test set [2 9 10 15 20 33 42 45] MSE = 0186 1198772 = 0725 119877 = 0851pIC50AMA
Model 119899 = 48 MSE = 0104 1198772 = 0918 119877 = 0843
Training set [2 3 4 5 6 7 10 11 13 16 18 19 20 21 23 27 28 2930 31 32 33 34 35 36 37 38 39 41 43 44 45 46 47] MSE = 0060 1198772 = 0897 119877 = 0947
Validation set [1 8 14 15 17 24 25] MSE = 0198 1198772 = 0710 119877 = 0843Test set [9 12 22 26 40 42 48] MSE = 0228 1198772 = 0791 119877 = 0890
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
Advances in Physical Chemistry 13Ta
ble8Va
lues
ofdescrip
torsantileish
manialactivity
pIC 50and
leverages(ℎ)
forthe
newdesig
nedcompo
unds
Designedcompo
unds
Num
ber
Descriptorsvalues
pIC 50(PRO
)Num
ber
Descriptorsvalues
pIC 50(A
MA)
N
R 2R 1
NH
2
1198831
R 1=OHR2=CH3
N=12490120583
=2232120596=
3099
NHA=3NHD=2MTI
=3651
TVC=334
10minus05
0571
119884 1
R 1=OCH3R2=CF3
119864 HOMO=minus5
235C
T=885365
119870H=93
02
119879=5814
30N
HA=6NHD=1
1395
119884 2
R 1=OHR2=CF3
119864 HOMO=minus5
312CT
=903801
119870H=12056
119879=647130N
HA=6NHD=2
099
8
119884 3
R 1=OCH3R2=F
119864 HOMO=minus5
034C
T=893897
119870H=10174
119879=566560NHA=4NHD=1
048
5
1198832
R 1=OHR2=OH
N=12380120583
=10
80120596
=3131
NHA=4NHD=3MTI
=3497
TVC=14
910minus05
0170
119884 4
R 1=R 2
=OC 2
H5
119864 HOMO=minus4
696C
T=9272
41
119870H=11222
119879=622010NHA=4NHD=1
046
4
119884 5
R 1=R 2
=OCH3
119864 HOMO=minus4
731C
T=916967
119870H=11469
119879=5994
70N
HA=4NHD=1
044
2
119884 6
R 1=OCH3R2=OC(
CH3) 3
119864 HOMO=minus4
872C
T=928656
119870H=11099
119879=635700NHA=4NHD=1
0109
119884 7
R 1=OCH3R2=CH2NH2
119864 HOMO=minus4
801C
T=934561
119870H=1418
5119879=66
0500NHA=4NHD=2
0122
N
O
O
R 2R 1
NH
2
1198833
R 1=OHR2=meta-NH2
N=11250120583
=5944120596=3392
NHA=5NHD=3MTI
=17128
TVC=223
10minus08
083
8119884 8
R 1=OCH3R2=param
eta-
F119864 H
OMO=minus5
081C
T=1006097
119870H=13267119879
=762370NHA=
6NHD=1
030
8
1198834
R 1=OCH3R2=para-N
H2
N=10850120583
=7114120596
=3295
NHA=5NHD=2MTI
=19138
TVC=204
10minus08
060
7119884 9
R 1=OCH3R2=para
NH2
119864 HOMO=minus4
966C
T=1066417
119870H=16853119879
=8319
3NHA=
5NHD=2
0199
1198835
R 1=OCH3R2=meta-NH2
N=10850120583
=6193120596=3353
NHA=5NHD=2MTI
=18992
TVC=204
10minus08
030
9119884 10
R 1=OCH3R2=metaN
H2
119864 HOMO=minus4
993C
T=1066417
119870H=16853119879
=8319
30N
HA=
5NHD=2
0166
1198836
R 1=OHR2=H
N=78
20120583
=5631120596=3483
NHA=4NHD=2MTI
=15585
TVC=447
10minus08
038
1
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
14 Advances in Physical Chemistry
N
R1
R2
R3
R4R5
R6
R7
Figure 8
by the same methods as well as pIC50
values theoreticallypredicted by the MLR models are listed in Table 8
From the predicted activities it has been observed thatthe designed compounds have higher pIC
50values than the
existing compounds in the case of the 60 studied compounds(Table 1)
The leverage values (ℎ) for the new designed com-pounds are 0268 0602 0775 0527 0386 and 0369 for1198831 1198832 1198836 respectively and 0649 0797 0201 02510235 0166 0176 0339 0169 and 0166 for 1198841 1198842 11988410respectively (Table 8) Only compounds 1198832 1198833 1198841 and1198842 are defined as outliers and consequently they are notbe considered because they have higher leverage which isgreater than ℎlowast (0585 for pIC
50(PRO) and 0553 for pIC
50
(AMA)) we suggest all other compounds as candidates thatwill be synthetized and evaluated as antileishmanial drugs
4 Conclusion
Following the five principles of the Organisation for Eco-nomic Co-operation and Development (OECD) for thevalidation of QSAR models two different modelling meth-ods multiple linear regression (MLR) and artificial neuralnetwork (ANN) were used in the construction of QSARmodels for the antileishmanial activities against two formsof parasites (promastigotes and amastigotes) of acridinesderivatives The accuracy and predictability of the proposedmodels were proven by the comparison of key statisticalterms of models The good results obtained with the internaland external validations show that the proposed models inthis paper are able to predict activities with a great perfor-mance and that the selected descriptors are pertinent Theapplicability domain (AD) of the MLR model was definedThe resulting models have shown that we have establisheda relationship between some descriptors and the activitiesin satisfactory manners The ANN results have substantiallybetter predictive capability than the MLR but the latter givesthe most important interpretable results
The obtained results show that to increase antileishma-nial activity against promastigotes parasites we will increaseelectrophilicity anddecrease branching polarity andnumberof hydrogen atoms attached in the heteroatom of the acridinederivatives Moreover to increase antileishmanial activityagainst amastigotes parasites we will increase solubilitypolarity and branching and decrease electrophilicity andnumber of heteroatoms mostly attached in the hydrogenatoms and critical temperature of acridine derivatives Com-paring 119905-test and standardized coefficient values of descrip-tors indicates that the influences of the electrophilicity index
(120596) on pIC50
(PRO) and of the number of H-Bond donors(NHD) on pIC
50(AMA) are stronger than those of the others
The most important finding from this research is that wehave designed and proposed new compounds with highervalues of activities compared to existing ones by addingsuitable substituents and calculating their activity using theregression equations Consequently the proposed modelswill reduce the time and cost of synthesis as well as thedetermination of the antileishmanial activities against pro-mastigotes and amastigotes forms of parasites of acridinederivatives
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
The authors are grateful to the ldquoAssociation Marocainedes Chimistes Theoriciensrdquo (AMCT) for its pertinent helpconcerning the programs
References
[1] M R Moein R S Pawar S I Khan B L Tekwani and I AKhan ldquoAntileishmanial antiplasmodial amp cytotoxic activities of12 16-dideoxyaegyptinone B from Zhumeria majdae Rech f ampWendelbordquo Phytotherapy Research vol 22 pp 283ndash285 2008
[2] L G Rocha J R G S Almeida R O Macedo and J MBarbosa-Filho ldquoA review of natural products with antileishma-nial activityrdquo Phytomedicine vol 12 no 6-7 pp 514ndash535 2005
[3] O Kayser A F Kiderlen and S L Croft ldquoNatural productsas potential antiparasitic drugsrdquo Studies in Natural ProductsChemistry vol 26 pp 779ndash848 2002
[4] WHOmdashWorld Health Organization 2011 httpwwwwhointleishmaniasisdisease epidemiologyenindexhtml
[5] S M B Jeronimo P Duggal R F S Braz et al ldquoAn emergingperi-urban pattern of infection with Leishmania chagasi theprotozoan causing visceral leishmaniasis in northeast BrazilrdquoScandinavian Journal of Infectious Diseases vol 36 no 6-7 pp443ndash449 2004
[6] M V L Marlet D K Sang K Ritmeijer R O Muga JOnsongo and R N Davidson ldquoEmergence or re-emergence ofvisceral leishmaniasis in areas of Somalia northeastern Kenyaand south-eastern Ethiopia in 2000-2001rdquo Transactions of theRoyal Society of Tropical Medicine amp Hygiene vol 97 no 5 pp515ndash518 2003
[7] J Querido ldquoEmergency initiative to reduce leishmaniasis inAfghanistanrdquo The Lancet Infectious Diseases vol 4 no 10 p599 2004
[8] B L Herwaldt ldquoLeishmaniasisrdquo The Lancet vol 354 no 9185pp 1191ndash1199 1999
[9] Control of the leishmaniases Control of the leishmaniasesControl of the leishmaniases 2010 Geneva Switzerland httpappswhointirisbitstream10665444121WHO TRS 949 engpdf
[10] P Desjeux ldquoThe increase in risk factors for Leishmania-sis worldwiderdquo Transactions of the Royal Society of TropicalMedicine amp Hygiene vol 95 no 3 pp 239ndash243 2001
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
Advances in Physical Chemistry 15
[11] G Chakrabarti A Basu P P Manna S B Mahato NB Mandal and S Bandyopadhyay ldquoIndolylquinoline deriva-tives are cytotoxic to Leishmania donovani promastigotes andamastigotes in vitro and are effective in treating murine visceralleishmaniasisrdquo Journal of Antimicrobial Chemotherapy vol 43no 3 pp 359ndash366 1999
[12] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 1rsquo-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[13] M O F Khan S E Austin C Chan et al ldquoUse of anadditional hydrophobic binding site the Z site in the rationaldrug design of a new class of stronger trypanothione reductaseinhibitor quaternary alkylammonium phenothiazinesrdquo Journalof Medicinal Chemistry vol 43 no 16 pp 3148ndash3156 2000
[14] D Pathak M Yadav N Siddiqui and S Kushawah ldquoAntileish-manial agents an updated reviewrdquoDer Pharma Chemica vol 3no 1 pp 39ndash249 2011
[15] L Gupta A Talwar Nishi S Palne S Gupta and P M SChauhan ldquoSynthesis of marine alkaloid 89-Dihydrocoscin-amide B and its analogues as novel class of antileishmanialagentsrdquo Bioorganic and Medicinal Chemistry Letters vol 17 no14 pp 4075ndash4079 2007
[16] WADenny ldquoAcridine derivatives as chemotherapeutic agentsrdquoCurrent Medicinal Chemistry vol 9 no 18 pp 1655ndash1665 2002
[17] M Wainwright ldquoAcridinemdasha neglected antibacterial chro-mophorerdquo Journal of Antimicrobial Chemotherapy vol 47 no1 pp 1ndash13 2001
[18] I Antonini ldquoDNA-binding antitumor agents From pyrim-ido[561-de]acridines to other intriguing classes of acridinederivativesrdquo Current Medicinal Chemistry vol 9 no 18 pp1701ndash1716 2002
[19] M Demeunynck F Charmantray and A Martelli ldquoInterest ofacridine derivatives in the anticancer chemotherapyrdquo CurrentPharmaceutical Design vol 7 no 17 pp 1703ndash1724 2001
[20] S A Gamage D P Figgitt S J Wojcik et al ldquoStructure-activity relationships for the antileishmanial and antitrypanoso-mal activities of 11015840-substituted 9-anilinoacridinesrdquo Journal ofMedicinal Chemistry vol 40 no 16 pp 2634ndash2642 1997
[21] C M Mesa-Valle J Castilla-Calvente M Sanchez-Moreno VMoraleda-Lindez J Barbe and A Osuna ldquoActivity and modeof action of acridine compounds against Leishmania donovanirdquoAntimicrobial Agents and Chemotherapy vol 40 no 3 pp 684ndash690 1996
[22] C Di Giorgio K Shimi G Boyer F Delmas and J-P GalyldquoSynthesis and antileishmanial activity of 6-mono-substitutedand 36-di-substituted acridines obtained by acylation ofproflavinerdquo European Journal of Medicinal Chemistry vol 42no 10 pp 1277ndash1284 2007
[23] C Di Giorgio F Delmas N Filloux et al ldquoIn vitro activitiesof 7-substituted 9-chloro and 9-amino-2-methoxyacridines andtheir bis- and tetra-acridine complexes against Leishmaniainfantumrdquo Antimicrobial Agents and Chemotherapy vol 47 no1 pp 174ndash180 2003
[24] GCarole DMMichel C Julien et al ldquoSynthesis and antileish-manial activities of 45-di-substituted acridines as compared totheir 4-mono-substituted homologuesrdquo Bioorganic ampMedicinalChemistry vol 13 no 19 pp 5560ndash5568 2005
[25] Adamoand Baron (2000) Parac ampGrimme (2003) Gaussian 03[26] ACDLABS 10 ldquoAdvanced Chemistry Development Inc
Toronto ON Canadardquo 2015 httpwwwacdlabscom
[27] MarvinSketch 5114 ldquoChem Axonrdquo 2012 httpwwwche-maxoncom
[28] ChemBioOffice PerkinElmer Informatics 2010 httpwwwcambridgesoftcom
[29] J C Dearden M T D Cronin and K L E Kaiser ldquoHow not todevelop a quantitative structure-activity or structure-propertyrelationship (QSARQSPR)rdquo SAR and QSAR in EnvironmentalResearch vol 20 no 3-4 pp 241ndash266 2009
[30] K Roy R N Das P Ambure and R B Aher ldquoBe aware of errormeasures Further studies on validation of predictive QSARmodelsrdquo Chemometrics and Intelligent Laboratory Systems vol152 pp 18ndash33 2016
[31] OECD Guidance Document on the Validation of QSAR ModelsOrganization for Economic Co-Operation amp DevelopmentParis France 2007
[32] L Eriksson J Jaworska A P Worth M T D Cronin RM McDowell and P Gramatica ldquoMethods for reliability anduncertainty assessment and for applicability evaluations ofclassification- and regression-based QSARsrdquo EnvironmentalHealth Perspectives vol 111 no 10 pp 1361ndash1375 2003
[33] P Gramatica ldquoPrinciples of QSAR models validation internaland externalrdquo QSAR and Combinatorial Science vol 26 no 5pp 694ndash701 2007
[34] G E Batista and D F Silva ldquoHow k-nearest neighbor param-eters affect its performancerdquo in Proceedings of the ArgentineSymposium on Artificial Intelligence pp 1ndash12 Instituto deCiencias Matematicase de Computa cao Sao Carlos Brazil2009
[35] T I Netzeva A P Worth T Aldenberg et al ldquoCurrentstatus of methods for defining the applicability domain of(quantitative) structure-activity relationshipsrdquo Alternatives toLaboratory Animals vol 33 no 2 pp 155ndash173 2005
[36] J C Dearden ldquoThe history and development of quantitativestructure-activity relationships (QSARs)rdquo International Journalof Quantitative Structure-Property Relationships vol 1 no 1 pp1ndash44 2016
[37] K Roy S Kar and P Ambure ldquoOn a simple approach for deter-mining applicability domain of QSAR modelsrdquo Chemometricsand Intelligent Laboratory Systems vol 145 pp 22ndash29 2015
[38] K Roy S Kar and R Das Narayan ldquoUnderstanding the Basicsof QSAR for Applications in Pharmaceutical Sciences amp RiskAssessmentrdquo in Chapter 6-Selected Statistical Methods in QSARpp 191ndash229 Academic Press Boston Mass USA 2015
[39] S-S So and W Graham Richards ldquoApplication of neuralnetworks quantitative structure-activity relationships of thederivatives of 24-diamino-5-(substituted-benzyl)pyrimidinesas DHFR inhibitorsrdquo Journal ofMedicinal Chemistry vol 35 no17 pp 3201ndash3207 1992
[40] T A Andrea and H Kalayeh ldquoApplications of neural networksin quantitative structure-activity relationships of dihydrofolatereductase inhibitorsrdquo Journal of Medicinal Chemistry vol 34no 9 pp 2824ndash2836 1991
[41] A Golbraikh and A Tropsha ldquoBeware of q2rdquo Journal ofMolecular Graphics and Modelling vol 20 no 4 pp 269ndash2762002
[42] S Chtita RHmamouchiM LarifMGhamaliM Bouachrineand T Lakhlifi ldquoQSPR studies of 9-aniliioacridine derivativesfor their DNA drug binding properties based on densityfunctional theory using statistical methods model validationand influencing factorsrdquo Journal of TaibahUniversity for Sciencevol 10 no 6 pp 868ndash876 2016
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
16 Advances in Physical Chemistry
[43] S Chtita M Larif M Ghamali M Bouachrine and TLakhlifi ldquoQuantitative structurendashactivity relationship studies ofdibenzo[ad]cycloalkenimine derivatives for non-competitiveantagonists of N-methyl-d-aspartate based on density func-tional theory with electronic and topological descriptorsrdquoJournal of Taibah University for Science vol 9 no 2 pp 143ndash154 2015
[44] R G Brereton Chemometrics Data Analysis for the Laboratoryand Chemical Plant John Wiley amp Sons Chichester UK 2003
[45] WD Fisher ldquoOn grouping formaximumhomogeneityrdquo Journalof the American Statistical Association vol 53 no 284 pp 789ndash798 1958
[46] A N Choudhary A Kumar and V Juyal ldquoQuantitativestructure activity relationship (QSAR) analysis of substituted4-oxothiazolidines and 5-arylidines as lipoxygenase inhibitorsrdquoMini-Reviews inMedicinal Chemistry vol 10 no 8 pp 705ndash7142010
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Inorganic ChemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal ofPhotoenergy
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Carbohydrate Chemistry
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Physical Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom
Analytical Methods in Chemistry
Journal of
Volume 2014
Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
SpectroscopyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Medicinal ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chromatography Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Theoretical ChemistryJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Spectroscopy
Analytical ChemistryInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Quantum Chemistry
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Organic Chemistry International
ElectrochemistryInternational Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CatalystsJournal of