Chinese Pharmaceutical Association
Institute of Materia Medica, Chinese Academy of Medical Sciences
Acta Pharmaceutica Sinica B
Acta Pharmaceutica Sinica B 2011;1(1):27–35
2211-3835 & 2011 In
hosting by Elsevier B
Peer review under res
doi:10.1016/j.apsb.20
$Project supportenCorresponding au
E-mail address: d
www.elsevier.com/locate/apsbwww.sciencedirect.com
ORIGINAL ARTICLE
Pharmacophore modeling and virtual screening for the
discovery of new fatty acid amide hydrolase inhibitors$
Dong-sheng Zhaoa, Hai-yan Wangb, Zhi-hui Liana, Da-xiong Hana,n, Xin Jina
aDepartment of Pharmacy, Medical College of Xiamen University, Xiamen 361005, ChinabThird Institute of Oceanography, State Oceanic Administration of China, Xiamen 361005, China
Received 18 November 2010; revised 21 January 2011; accepted 9 March 2011
KEY WORDS
FAAH inhibitors;
Pharmacophore;
Virtual screening
stitute of Materia M
.V. All rights rese
ponsibility of Insti
11.04.003
d by the Nationalthor. Tel.: þ86 592
Abstract A predictive pharmacophore model has been generated from a series of diverse fatty acid
amide hydrolase (FAAH) inhibitors and the optimal pharmacophore model applied in virtual
screening. The pharmacophore model was based on a training set of 21 compounds carefully
selected from the published literatures. The optimal model Hypo-1 included four features (two
hydrogen-bond acceptor units, one aromatic hydrophobic unit and one aromatic ring unit) and two
excluded volumes. Cross-validation of the model confirmed that Hypo-1 was not generated by
chance correlation. A large test set of 55 compounds showed that Hypo-1 performed well in
classifying highly active and less active FAAH inhibitors. Superimposition analysis of the FAAH
X-ray crystal structure and the pharmacophore Hypo-1 further validated the adequacy of the
model. Virtual screening generated a total of 976 hits from the Zinc Natural Products database, a
hit rate of 1.04% and enrichment of 83.89. The acceptable hit rate further supports the use of
Hypo-1 as a 3D query tool for virtual screening.
& 2011 Institute of Materia Medica, Chinese Academy of Medical Sciences and Chinese Pharmaceutical
Association. Production and hosting by Elsevier B.V. All rights reserved.
edica, Chinese Academy of Medical Sciences and Chinese Pharmaceutical Association. Production and
rved.
tute of Materia Medica, Chinese Academy of Medical Sciences and Chinese Pharmaceutical Association.
Science Foundation of China (Grant No. 40976050).2188681.
n (Da-xiong Han).
Dong-sheng Zhao et al.28
1. Introduction
Fatty acid amide hydrolase (FAAH) is an integral membrane
protein primarily responsible for the catabolism of the fatty
acid amide (FAA) family of endogenous signaling lipids.
Substrates of FAA include the endogenous cannabinoids
(endocannabinoid), N-arachidonoylethanolamine (ananda-
mide) and 2-arachidonoylglycerol (2-AG)1. The endocannabi-
noid system is composed of two G-protein-coupled receptors,
CB1 and CB22, which are activated by signaling lipids to
modulate a range of mammalian behaviors including pain,
inflammation, appetite, motility, sleep, thermoregulation, cog-
nition and emotional state3.
FAAH inhibition leads to elevated endogenous levels of
FAAs and a range of behavioral effects. FAAH inhibition
offers an attractive way to activate the endocannabinoid
system without many side effects associated with the use of
direct cannabinoid receptor agonists4. The past 10 years have
seen an explosion in the number and quality of FAAH
inhibitors available for pharmacological evaluation, including
URB880 URB597
URB602 JP83
BMS-1 OL
OL135 Training08
Training03 Trainin
ARC70484XX
Figure 1 Molecular structures of
the compounds that have shown promising in vivo activity,
selectivity and therapeutic effects5. Kimball et al.6 introduced
a large number of potential inhibitors based on a-ketohetero-cycles (e.g. OL135, Fig. 1) that reversibly form hemiketals with
the active site serine of FAAH; Mor et al.7 introduced another
class of inhibitors derived from N-alkylcarbamates (e.g.
URB597, Fig. 1) that irreversibly acylate the active site serine
of FAAH. Other recently reported structures with inhibitory
activity against FAAH include piperidine and piperazine
ureas8, sulfonyl derivatives9, (thio)hydantoins10, boronic acid
derivatives11 and b-lactam derivatives12.
A chemical features based pharmacophore model may serve
as a guide in the screening of potential inhibitors. A pharma-
cophore represents the three-dimensional arrangement of
structural or chemical features of a drug essential for optimal
binding to the receptor13. In this report, we present the
generation and validation of a pharmacophore model of
FAAH inhibitors using the Discovery Studio 2.1/‘‘3D QSAR
Pharmacophore Generation’’ protocol. We also present the
results of virtual screening.
URB524 URB532
Training04 PF622
92 CAY10402
Training 01 Training02
g05 Training06
SPB03742 Training07
the 21 training set compounds.
Pharmacophore modeling and virtual screening for new FAAH inhibitors 29
2. Materials and methods
2.1. Molecular modeling
Molecular modeling was performed on a Silicon Graphics
work station (IBM Z-PRO 107) using Hyperchem 6.0 software
for molecular structure optimization and Discovery Studio 2.1
software (Accelrys Inc., San Diego, CA) for pharmacophore
models generation and virtual screening.
2.2. Training set and test set compound selection
Based on the published literatures, we collected 21 typical
compounds7,12,14–22 (Fig. 1) as a training set to generate
pharmacophore models of FAAH inhibitors according to a
three-dimensional quantitative structure–activity relationship.
Inhibitory activity as measured by IC50 values was in the range
0.4–5700 nmol/L spanning 5 orders of magnitude. To validate
the pharmacophore models obtained, 55 FAAH inhibi-
tors7,12,14–31 (Fig. 2) were selected to compose the test set.
For estimation purposes, activity was classified as follows:
IC50o100 nmol/L highly active (represented as þþ); 100rIC50o1000 nmol/L moderately active (represented as þ) and
IC50Z1000 nmol/L less active (represented as �).
2.3. Conformational analysis
Initially all training and test set structures were generated and
optimized using Hyperchem 6.0 software. All structures were
minimized using the Polak–Ribi�ere (conjugate-gradient) algo-
rithm with a RMS gradient set at 0.01 kcal/mol. To take into
account the molecular flexibility, the protocol ‘‘Diverse
Conformation Generation’’ in Discovery Studio 2.1 was
applied to both sets of molecules using the Poling algorithm
and selecting the ‘‘best’’ conformer generation option with an
energy threshold of 20 kcal/mol and a maximum conforma-
tion number of 300.
2.4. Generation of pharmacophores
An initial analysis of the training set using the ‘‘Create 3D
Fingerprints’’ protocol revealed that hydrogen-bond acceptor
(HBA), hydrophobic (HY) and aromatic ring (RA) features
effectively mapped the critical structural/chemical features of
all training set molecules. On this basis, we selected HBA, HY
(including hydrophobic aromatic (HY_ar), hydrophobic
aliphatic (HY_al)) and RA features for the ‘‘3D QSAR
Pharmacophore Generation’’ protocol. The minimum and
the maximum number of selected features were set at 0 and
5, respectively, the maximum number of excluded volumes
(EV) at 4, the uncertainty value at 2 (defined by software
Discovery Studio as a ratio of the reported value to the
minimum and maximum values) and the Fischer validation
at 95%.
3. Results and discussion
3.1. Pharmacophore generation
The 3D QSAR pharmacophore generation protocol generated
a set of 10 pharmacophore hypotheses from the 21 training set
compounds. All hypotheses showed high correlations indicat-
ing their stability. Three or four features (HBA, RA, HY and
HY_ar) and several excluded volumes (Table 1) were con-
sidered in all the hypotheses. The total cost of each hypothesis
is close to the fixed cost and very different to the null cost
value, indicating good hypotheses. The configuration value of
these hypotheses is 15.63, which, being less than 17, is within
the allowed range.
The optimal pharmacophore hypothesis, i.e. the one with
the highest correlation, the lowest RMS and the lowest total
cost, was named Hypo-1. It consisted of four features (HBA1,
HBA2, HY_ar and RA) and two excluded volumes (EV1 and
EV2) (Fig. 3). The RMS value of 1.47, correlation coefficient
of 0.926 and cost difference of 112.09 (Table 1) demonstrate
that there is 490% probability of accurately predicting the
activities of compounds in the training set. The reason for the
rather high value of RMS is that the uncertainty value was set
at 3 not at 2.
Table 2 shows the actual and estimated IC50 values of
training set molecules with error values, fit values and
matching degrees with the pharmacophore Hypo-1. On the
basis of the activity scale mentioned above, only one highly
active (þþ) compound was underestimated as moderate
activity (þ), one less active (�) compound was overestimated
as moderate activity (þ) and no highly active (þþ) compound
was estimated as less active (�) and vice versa. This indicates
that the pharmacophore Hypo-1 classifies the training set
compounds correctly. Fig. 3 shows the highly active com-
pound OL92 and the less active compound URB602 mapped
on Hypo-1; OL92 mapped well but URB602 missed two
features. Overall, the mapping results suggest that the phar-
macophore Hypo-1 provides a reasonable representation of
the pharmacophore of FAAH inhibitors.
3.2. Cross-validation study
The quality of the best pharmacophore hypothesis was
assessed using the Fischer validation (CatScramble test) in
Discovery Studio 2.1. This technique randomly reassigns the
activity values of the training set compounds and then creates
new spreadsheets to validate the strong correlation between
chemical structure and biological activity of training set
compounds. A total of 19 randomizations are created when
the Fischer validation is set at 95%.
The results of the Fischer validation are shown in Table 3.
The results clearly indicate that in the 19 randomizations, all
correlations are less than 0.76 and all cost values are more
than that of Hypo-1. These cross-validation results verify that
there is a 95% chance that Hypo-1 represents a true correla-
tion in the training set activity data.
3.3. Validation of Hypo-1 in the test set
The test set of 55 reported FAAH inhibitors was analyzed
to check the predictive capability of Hypo-1. The ‘‘Ligand
Pharmacophore Mapping’’ protocol with Maximum Omitted
Features set at 2 and the Fitting Method of Rigid in Discovery
Studio 2.1 was used to estimate the activities of the test set
compounds. The estimated activities and actual activities are
compared in Table 4. Of the 30 highly active (þþ) com-
pounds, none was classified as less activity (�) and only one
test01 test02 test03 test04
test05 test06 test07 test08
test09 test10 test11 test12
test13 test14 test15 test16
test17 test18 test19 test20
test21 test22 test23 test24
test25 test26 test27 test28
test29 test30 test31 test32
test33 test34 test35
test36 test37 test38 test39
test40 test41 test42 test43
test44 test45 test46 test47
test48 test49 test50 test51
test52 test53 test54 test55
Figure 2 Molecular structures of the 55 test set compounds.
Dong-sheng Zhao et al.30
Table 1 Statistical significance, predictive power and features of the top-10 pharmacophore hypotheses derived from the training
set of molecule.
Hypothesis Correlationa RMSb Total cost DCostc Featuresd EV
Hypo-1 0.926 1.47 106.63 112.09 HBA HBA RA HY_ar 2
Hypo-2 0.922 1.55 117.08 101.64 HBA HBA HY HY_ar 3
Hypo-3 0.918 1.54 106.84 111.88 HBA HBA RA HY_ar 3
Hypo-4 0.906 1.67 117.34 101.38 HBA HBA HY_ar 2
Hypo-5 0.899 1.74 122.9 95.82 HBA HBA HY HY_ar 2
Hypo-6 0.896 1.73 118.01 100.71 HBA HBA HY RA 4
Hypo-7 0.892 1.77 118.94 99.78 HBA HBA HY HY_ar 3
Hypo-8 0.881 1.84 120.7 98.02 HBA HBA HY HY_ar 4
Hypo-9 0.877 1.87 120.95 99.77 HBA HBA HY 3
Hypo-10 0.865 1.95 124.4 94.32 HBA HBA HY RA 3
Fixed cost¼80.3, null cost¼218.72 and configuration¼15.63.aCorrelation coefficient based on linear regression derived from the geometric fit index.bRMS, the deviation of log (estimated activity) from log (measured activity) normalized to log (uncertainty).cDCost¼null cost�total cost.dHBA, hydrogen-bond accepter; HY, hydrophobic; HY_ar, hydrophobic_aromatic; EV, excluded volume.
Figure 3 Mapping of top scoring pharmacophore Hypo-1 on (a) highly active training set compound OL92; (b) less active training set
compound URB602; (c) highly active test set compound test 55 and (d) less active test set compound test 52. Cyan sphere, HBA; green
sphere, HY_ar; yellow/brown sphere, RA; black sphere, EV.
Pharmacophore modeling and virtual screening for new FAAH inhibitors 31
was underestimated as moderately active (þ); of the 16
moderately active (þ) compounds, all were predicted cor-
rectly; of the other 9 less active (�) compounds, none was
overestimated as highly active (þþ) but 7 were classified as
moderately active (þ) with most of their error values close to
1. On this basis, it can be concluded that Hypo-1 is clearly able
to correctly classify the compounds outside the training set.
Mapping of one highly active and one less active compound in
the test set on Hypo-1 is shown in Fig. 3(c) and (d).
3.4. X-ray crystallographic analysis
Mileni et al.4 presented the X-ray crystal structure of the complex
of OL135 and FAAH (PDB ID: 2WJ2) (Fig. 4). Superposition of
Hypo-1 onto FAAH reveals a good agreement of the critical
interacting chemical features. As shown in Fig. 4, OL135 in the
crystal structure and in the training set adopts the similar
conformations, the pyridine group, oxazole group and benzene
ring separately mapping onto the HY_ar, HBA1 and RA groups
correctly. In the X-ray crystal structure, the hydrophobic tail and
the phenyl group of OL135 are compressed by the hydrophobic
and aromatic Phe381, Phe432, Met436 and Met495 residues to
form a complementary cavity with a corresponding RA feature;
FAAH has an unusual Ser241–Ser217–Lys142 catalytic triad, the
carbonyl group of OL135 undergoing nucleophilic attack by the
Ser241 residue. Corresponding to Hypo-1, there is an HBA
feature indicating that potential hits must have a hydrogen-bond
accepter group at this position. The HY_ar feature of Hypo-1 is
also compressed by the hydrophobic Phe381, Leu278 and
Table 2 Actual IC50 values and their estimated IC50 of 21 training set molecules calculated on the basis of the pharmacophore
Hypo-1.
No. Name Fita Mappingb Est.c (nmol/L) Est. scale Act. scale Act.d (nmol/L) Errore
HBA1 HBA2 HY_ar RA
1 Training05 8.94 16 24 28 1� �
0.34 þþ þþ 0.4 �1.2
2 OL92 8.05 � 17 20 1� �
2.6 þþ þþ 0.44 þ5.9
3 URB880 8.75 20 6 21 12� �
0.52 þþ þþ 0.63 �1.2
4 BMS1 7.88 15 26 31 �� �
3.9 þþ þþ 2 þ1.9
5 Training06 7.64 � 19 22 1� �
6.6 þþ þþ 2 þ3.3
6 URB597 7.46 25 11 � 17� �
10 þþ þþ 4.6 þ2.2
7 Training01 7.53 23 3 1 �� �
8.5 þþ þþ 5 þ1.7
8 CAY10402 7.88 � 15 19 1� �
3.8 þþ þþ 9 �2.4
9 Training04 7.54 27 � 2 19� �
8.3 þþ þþ 12 �1.4
10 JP83 7.56 15 � 29 1� �
7.9 þþ þþ 14 �1.8
11 OL135 7.24 � 17 23 1� �
17 þþ þþ 15 þ1.1
12 PF622 6.9 � 26 23 4� �
36 þþ þþ 33 þ1.1
13 Training08 7.08 � 12 4 14� �
24 þþ þþ 43 �1.8
14 ARC70484xx 6.92 32 18 16 �� �
35 þþ þþ 59 �1.7
15 URB524 5.62 � 4 8 �� �
600 þ þþ 63 þ9.5
16 Training07 7.14 31 15 1 �� �
21 þþ þþ 64 �3
17 URB532 5.62 � 15 10 �� �
700 þ þ 390 þ1.8
18 Training02 5.61 � � 20 1� �
710 þ þ 400 þ1.8
19 SPB03742 6.12 � 10 12 22� �
220 þ þ 690 �3.2
20 Training03 5.65 24 5 � 17� �
660 þ � 4000 �6.1
21 URB602 5.3 � 11 13 �� �
1500 � � 5700 �3.9
aFit value calculated by geometry fitting between the hypothesis and the compound; the higher the value, the better the fit.bMatching degree of pharmacophore element with the specific conformation of the compound; n means no matching; HBA, hydrogen-
bond accepter; HY, hydrophobic; HY_ar, hydrophobic_aromatic; EV, excluded volume.cEstimated IC50 value.dActual IC50 value.eError factor calculated as the ratio of the measured activity to the estimated activity or the inverse if the estimated activity is greater than
the measured activity. If measured activity is greater than the estimated activity, the error factor is negative and vice versa.
Table 3 Results from cross-validation run using CatScramble technique.
Validation Cost Correlation Validation Cost Correlation
Unscrambled 106.63 0.926 Random10 162.8 0.7
Random1 170.69 0.685 Random11 176.68 0.613
Random2 161.6 0.689 Random12 151.72 0.737
Random3 175.09 0.626 Random13 154.88 0.729
Random4 174.43 0.649 Random14 153.37 0.733
Random5 166.85 0.676 Random15 172.49 0.64
Random6 165.35 0.667 Random16 154.96 0.736
Random7 176.33 0.634 Random17 155.72 0.752
Random8 156.89 0.73 Random18 153.05 0.737
Random9 165.58 0.689 Random19 163.5 0.691
Dong-sheng Zhao et al.32
MET191 residues. In conclusion, the superposition studies
provide strong evidence that Hypo-1 is a good representation
of the pharmacophore of FAAH inhibitors.
3.5. Virtual screening
In order to further validate the discriminatory power of
Hypo-1 and to discover new potential FAAH inhibitors, 30
active FAAH inhibitors and 93,527 molecules from the Zinc
Natural Products database (2008.05)32 (to serve as decoy) were
screened. FAAH inhibitor enrichment of the database com-
pounds was expressed as enrichment factor (EF) calculated
using the following equation33:
EF¼TP=n
A=N
where TP is the number of FAAH inhibitors matched by
Hypo-1, n is the number of FAAH inhibitors and decoy
Table 4 Actual IC50 values and their estimated IC50 of 55 test set molecules calculated on the basis of the pharmacophore
Hypo-1.
Name Fita Est.b
(nmol/L)
Est.
scale
Act.
scale
Act.c
(nmol/L)
Errord Name Fita Est.b
(nmol/L)
Est.
scale
Act.
scale
Act.c
(nmol/L)
Errord
Test01 5.62 694.6 þ þ 360 þ1.93 Test29 6.82 43.99 þþ þþ 64 �1.45
Test02 6.86 39.85 þþ þþ 11 þ3.62 Test30 7.09 23.53 þþ þþ 4 þ5.88
Test03 5.62 688.81 þ � 1857 �2.7 Test31 7 28.66 þþ þþ 26.7 þ1.07
Test04 7.44 10.47 þþ þþ 49.6 �4.74 Test32 6.9 36.75 þþ þþ 10.3 þ3.57
Test05 7.36 12.54 þþ þþ 26.5 �2.11 Test33 7.37 12.3 þþ þþ 13 �1.06
Test06 7.25 16.4 þþ þþ 9.1 þ1.8 Test34 6.69 59.21 þþ þþ 23 þ2.57
Test07 5.62 694.05 � � 1700 �2.45 Test35 6.1 232.81 þ � 1167 �5.01
Test08 8.75 0.52 þþ þþ 0.63 �1.21 Test36 7.19 18.64 þþ þþ 70 �3.75
Test09 5.55 810.35 þ þ 252 þ3.22 Test37 6.6 72.99 þþ þþ 64 þ1.14
Test10 5.59 742.71 þ þ 176 þ4.22 Test38 7.56 7.94 þþ þþ 12 �1.51
Test11 5.61 704.32 þ � 1029 �1.46 Test39 7.62 6.95 þþ þþ 34 �4.9
Test12 5.62 688.04 þ � 2737 �3.98 Test40 7.15 20.71 þþ þþ 50 �2.41
Test13 5.61 705.46 þ � 3776 �5.35 Test41 5.62 687.77 þ þ 109 þ6.31
Test14 5.62 693.24 þ þ 766 �1.1 Test42 7.7 5.83 þþ þþ 7 �1.2
Test15 5.62 688.73 þ þ 696 �1.01 Test43 6.64 65.5 þþ þþ 90 �1.37
Test16 5.62 688 þ þ 113 þ6.09 Test44 6.58 75.85 þþ þþ 30 þ2.53
Test17 6.29 147.92 þ þþ 33 þ4.48 Test45 7.69 5.9 þþ þþ 20 �3.39
Test18 5.61 706.06 þ þ 160 þ4.41 Test46 5.61 707.2 þ þ 400 þ1.77
Test19 5.61 698.38 þ þ 690 þ1.01 Test47 7.17 19.62 þþ þþ 3 þ6.54
Test20 5.62 687.73 þ þ 139 þ4.95 Test48 5.87 394.02 þ þ 315 þ1.25
Test21 7.17 19.73 þþ þþ 5.3 þ3.72 Test49 5.62 688.3 þ � 2100 �3.05
Test22 6.6 73.57 þþ þþ 13 þ5.66 Test50 5.62 687.68 þ þ 590 þ1.16
Test23 6.82 43.4 þþ þþ 25 þ1.74 Test51 5.46 991.4 þ � 1200 �1.21
Test24 5.61 710.87 þ þ 174 þ4.09 Test52 5.38 1193.39 � � 5700 �4.78
Test25 5.61 706.13 þ þ 324 þ2.17 Test53 7.66 6.27 þþ þþ 2 þ3.13
Test26 5.62 687.7 þ þ 808 �1.17 Test54 7.68 6.01 þþ þþ 30 �4.99
Test27 7.91 3.55 þþ þþ 10 �2.82 Test55 8.81 0.45 þþ þþ 0.4 þ1.12
Test28 7.88 3.82 þþ þþ 13 �3.4
aFit value calculated by geometry fitting between the hypothesis and compound; the higher the value, the better the fit.bEstimated IC50 value.cActual IC50 value.dError factor calculated as the ratio of measured activity to estimated activity or the inverse if the estimated activity is greater than the
measured activity. If measured activity is greater than estimated activity, the error factor is negative and vice versa.
Trp531
Gly485
Val491Ser217
Ser241Lys142
Met191Leu154
Gly268Leu278
Phe381
Ile238
Phe192
Phe381
Met436
Met495Phe432
Figure 4 Superposition of the pharmacophore Hypo-1 with the FAAH crystal structure. The molecule in red is OL135 in the crystal
structure and the molecule in blue is OL135 in the training set.
Pharmacophore modeling and virtual screening for new FAAH inhibitors 33
Figure 5 Mapping of the hit compounds (a) ZINC06623865 and (b) ZINC02260731 from the Zinc Natural Products database onto the
pharmacophore Hypo-1. Cyan sphere, HBA; green sphere, HY_ar; yellow/brown sphere, RA; black sphere, EV.
Dong-sheng Zhao et al.34
molecules matched by the model, A is the number of known
FAAH inhibitors in the decoy and N is the number of
compounds in the decoy. As before flexibility was taken into
account by applying the ‘‘Diverse Conformation Generation’’
protocol in Discovery Studio 2.1. This generates diverse
conformers for all the compounds in the database using the
Poling algorithm by first selecting the best conformer genera-
tion option with an energy threshold of 20 kcal/mol and a
maximum conformation number of 300 and then using the
‘‘Ligand Pharmacophore Mapping’’ protocol with the same
parameters as in the test set analysis. In this way 27 of 30
FAAH inhibitors (90%) and 976 compounds in the Zinc
Natural Products database ( hit rate of 1.04%) matched
Hypo-1 model with estimated IC50 valueso100 nmol/L.
Taking this into account, an EF of 83.89 was calculated
for Hypo-1, showing it is highly restrictive in picking hits
from a chemistry database. The high performance of this
virtual screening further proves that Hypo-1 is very useful as
a 3D query tool in the screening of other databases.
Two compounds from the Zinc Natural Products database
predicted to be highly active (ZINC06623865 and
ZINC02260731) based on mapping onto Hypo-1 are shown
in Fig. 5.
4. Conclusions
This paper reports the use of ligand-based pharmacophore
generation and virtual screening methods to identify the ligand
requirements of FAAH inhibitors. The study presents the 3D
QSAR pharmacophore models for FAAH inhibitors and
establishes that the best pharmacophore Hypo-1 contains four
features (two hydrogen-bond acceptors, one aromatic hydro-
phobic group and one aromatic ring group) and two excluded
volumes. The pharmacophore model provides high predict-
ability for estimating the activities of a variety of compounds.
The virtual screening results provide a total of 976 new hits
from 93,527 molecules, proving the model is highly restrictive
in picking up the novel hits from the chemical databases. To
progress the discovery of novel FAAH inhibitors, some of the
hit compounds are currently synthesized and tested for activity
determined. In addition, the pharmacophore Hypo-1 mapping
approach is shown to be useful in understanding the impor-
tance of various chemical features in FAAH inhibitors, in
further virtual screening and in the design and optimization of
a new generation of FAAH inhibitors.
References
1. Pillarisetti S, Alexander CW, Khanna I. Pain and beyond: fatty
acid amides and fatty acid amide hydrolase inhibitors in cardio-
vascular and metabolic diseases. Drug Discov Today 2009;14:
1098–111.
2. Ahn K, Johnson DS, Mileni M, Beidler D, Long JZ, McKinney
MK, et al. Discovery and characterization of a highly selective
FAAH inhibitor that reduces inflammatory pain. Chem Biol 2009;
16:411–20.
3. Mileni M, Garfunkle J, DeMartino JK, Cravatt BF, Boger DL,
Stevens RC. Binding and inactivation mechanism of a humanized
fatty acid amide hydrolase by a-ketoheterocycle inhibitors
revealed from cocrystal structures. J Am Chem Soc 2009;131:
10497–506.
4. Mileni M, Johnson DS, Wang Z, Everdeen DS, Liimatta M, Pabst
B, et al. Structure-guided inhibitor design for human FAAH by
interspecies active site conversion. Proc Natl Acad Sci USA
2008;105:12820–4.
5. Myllymaki MJ, Kasnanen H, Kataja AO, Lahtela-Kakkonen M,
Saario SM, Poso A, et al. Chiral 3-(4,5-dihydro-oxazol-2-
yl)phenyl alkylcarbamates as novel FAAH inhibitors: insight into
FAAH enantioselectivity by molecular docking and interaction
fields. Eur J Med Chem 2009;44:4179–91.
6. Kimball FS, Romero FA, Ezzili C, Garfunkle J, Rayl TJ,
Hochstatter DG, et al. Optimization of a-ketooxazole inhibitors
of fatty acid amide hydrolase. J Med Chem 2008;51:937–47.
7. Mor M, Lodola A, Rivara S, Vacondio F, Duranti A, Tontini A,
et al. Synthesis and quantitative structure–activity relationship of
fatty acid amide hydrolase inhibitors: modulation at the N-portion
of biphenyl-3-yl alkylcarbamates. J Med Chem 2008;51:3487–98.
8. Ahn K, Johnson DS, Fitzgerald LR, Liimatta M, Arendse A,
Stevenson T, et al. Novel mechanistic class of fatty acid amide
hydrolase inhibitors with remarkable selectivity. Biochemistry
2007;46:13019–30.
9. Wang X, Sarris K, Kage K, Zhang D, Brown SP, Kolasa T, et al.
Synthesis and evaluation of benzothiazole-based analogues as
novel, potent, and selective fatty acid amide hydrolase inhibitors.
J Med Chem 2009;52:170–80.
10. Muccioli GG, Fazio N, Scriba GK, Poppitz W, Cannata F,
Poupaert JH, et al. Substituted 2-thioxoimidazolidin-4-ones and
imidazolidine-2,4-diones as fatty acid amide hydrolase inhibitors
templates. J Med Chem 2006;49:417–25.
11. Minkkila A, Saario SM, Kasnanen H, Leppanen J, Poso A,
Nevalainen T. Discovery of boronic acids as novel and potent
inhibitors of fatty acid amide hydrolase. J Med Chem 2008;51:
7057–60.
12. Feledziak M, Michaux C, Urbach A, Labar G, Muccioli GG,
Lambert DM, et al. b-Lactams derived from a carbapenem
chiron are selective inhibitors of human fatty acid amide
Pharmacophore modeling and virtual screening for new FAAH inhibitors 35
hydrolase versus human monoacylglycerol lipase. J Med Chem
2009;52:7054–68.
13. Guo YS, Guo ZR. Comparative pharmacophore analysis of dual
dopamine D2/5-HT2A receptor antagonists. Acta Pharm Sin
2009;44:314–20.
14. Tarzia G, Duranti A, Tontini A, Piersanti G, Mor M, Rivara S,
et al. Design, synthesis, and structure–activity relationships of
alkylcarbamic acid aryl esters, a new class of fatty acid amide
hydrolase inhibitors. J Med Chem 2003;46:2352–60.
15. Romero FA, Du W, Hwang I, Rayl TJ, Kimball FS, Leung D,
et al. Potent and selective a-keto heterocycle-based inhibitors of
the anandamide and oleamide catabolizing enzyme, fatty acid
amide hydrolase. J Med Chem 2007;50:1058–68.
16. Matuszak N, Muccioli GG, Labar G, Lambert DM. Synthesis
and in vitro evaluation of N-substituted maleimide derivatives as
selective monoglyceride lipase inhibitors. J Med Chem 2009;52:
7410–20.
17. Alexander JP, Cravatt BF. Mechanism of carbamate inactivation
of FAAH: implications for the design of covalent inhibitors and
in vivo functional probes for enzymes. Chem Biol 2005;12:1179–87.
18. Seierstad M, Breitenbucher JG. Discovery and development of
fatty acid amide hydrolase (FAAH) inhibitors. J Med Chem
2008;51:7327–43.
19. Minkkila A, Myllymaki MJ, Saario SM, Castillo-Melendez JA,
Koskinen AM, Fowler CJ, et al. The synthesis and biological
evaluation of para-substituted phenolic N-alkyl carbamates as
endocannabinoid hydrolyzing enzyme inhibitors. Eur J Med Chem
2009;44:2994–3008.
20. Cravatt BF, Lichtman AH. Fatty acid amide hydrolase: an
emerging therapeutic target in the endocannabinoid system. Curr
Opin Chem Biol 2003;7:469–75.
21. Wang YR, Xu J, Uveges A, Ramarao MK, Rogers KE, Jones PG.
A novel scintillation proximity assay for fatty acid amide hydro-
lase compatible with inhibitor screening. Anal Biochem 2006;354:
35–42.
22. Timmons A, Seierstad M, Apodaca R, Epperson M, Pippel D,
Brown S, et al. Novel ketooxazole based inhibitors of fatty acid
amide hydrolase (FAAH). Bioorg Med Chem Lett 2008;18:
2109–13.
23. Saario SM, Poso A, Juvonen RO, Jarvinen T, Salo-Ahen OM.
Fatty acid amide hydrolase inhibitors from virtual screening of the
endocannabinoid system. J Med Chem 2006;49:4650–6.
24. Wang XQ, Sarris K, Kage K, Zhang D, Brown SP, Kolasa T,
et al. Synthesis and evaluation of benzothiazole-based analogues
as novel, potent, and selective fatty acid amide hydrolase inhibi-
tors. J Med Chem 2009;52:170–80.
25. Keith JM, Apodaca R, Xiao W, Seierstad M, Pattabiraman K,
Wu J, et al. Thiadiazolopiperazinyl ureas as inhibitors of fatty
acid amide hydrolase. Bioorg Med Chem Lett 2008;18:4838–43.
26. Ortar G, Cascio MG, De Petrocellis L, Morera E, Rossi F,
Schiano-Moriello A, et al. New N-arachidonoyl serotonin analo-
gues with potential ‘‘dual’’ mechanism of action against pain.
J Med Chem 2007;50:6554–69.
27. Morera E, De Petrocellis L, Morera L, Moriello AS, Ligresti A,
Nalli M, et al. Synthesis and biological evaluation of piperazinyl
carbamates and ureas as fatty acid amide hydrolase (FAAH) and
transient receptor potential (TRP) channel dual ligands. Bioorg
Med Chem Lett 2009;19:6806–9.
28. Garfunkle J, Ezzili C, Rayl TJ, Hochstatter DG, Hwang I, Boger DL.
Optimization of the central heterocycle of a-ketoheterocycle inhibitorsof fatty acid amide hydrolase. J Med Chem 2008;51:4392–403.
29. Myllymaki MJ, Saario SM, Kataja AO, Castillo-Melendez JA,
Nevalainen T, Juvonen RO, et al. Design, synthesis, and in vitro
evaluation of carbamate derivatives of 2-benzoxazolyl- and 2-
benzothiazolyl-(3-hydroxyphenyl)-methanones as novel fatty acid
amide hydrolase inhibitors. J Med Chem 2007;50:4236–42.
30. DeMartino JK, Garfunkle J, Hochstatter DG, Cravatt BF, Boger
DL. Exploration of a fundamental substituent effect of alpha-
ketoheterocycle enzyme inhibitors: potent and selective inhibitors
of fatty acid amide hydrolase. Bioorg Med Chem Lett 2008;18:
5842–6.
31. Sit SY, Conway C, Bertekap R, Xie K, Bourin C, Burris K, et al.
Novel inhibitors of fatty acid amide hydrolase. Bioorg Med Chem
Lett 2007;17:3287–91.
32. Zinc Natural Products [DB]. /http://zinc.docking.org/vendor0/
npd/index.htmlS.
33. Diller DJ, Li R. Kinases homology models, and high throughput
docking. J Med Chem 2003;46:4638–47.