lia55 user manual

Schrdinger Press

Liaison User Manual

Liaison 5.5User Manual

Liaison User Manual Copyright 2009 Schrdinger, LLC. All rights reserved.

While care has been taken in the preparation of this publication, Schrdinger

assumes no responsibility for errors or omissions, or for damages resulting from

the use of the information contained herein.

Canvas, CombiGlide, ConfGen, Epik, Glide, Impact, Jaguar, Liaison, LigPrep,

Maestro, Phase, Prime, PrimeX, QikProp, QikFit, QikSim, QSite, SiteMap, Strike, and

WaterMap are trademarks of Schrdinger, LLC. Schrdinger and MacroModel are

registered trademarks of Schrdinger, LLC. MCPRO is a trademark of William L.

Jorgensen. Desmond is a trademark of D. E. Shaw Research. Desmond is used with

the permission of D. E. Shaw Research. All rights reserved. This publication may

contain the trademarks of other companies.

Schrdinger software includes software and libraries provided by third parties. For

details of the copyrights, and terms and conditions associated with such included

third party software, see the Legal Notices for Third-Party Software in your product

installation at $SCHRODINGER/docs/html/third_party_legal.html (Linux OS) or

%SCHRODINGER%\docs\html\third_party_legal.html (Windows OS).

This publication may refer to other third party software not included in or with

Schrdinger software ("such other third party software"), and provide links to third

party Web sites ("linked sites"). References to such other third party software or

linked sites do not constitute an endorsement by Schrdinger, LLC. Use of such

other third party software and linked sites may be subject to third party license

agreements and fees. Schrdinger, LLC and its affiliates have no responsibility or

liability, directly or indirectly, for such other third party software and linked sites,

or for damage resulting from the use thereof. Any warranties that we make

regarding Schrdinger products and services do not apply to such other third party

software or linked sites, or to the interaction between, or interoperability of,

Schrdinger products and services and such other third party software.

June 2009

ContentsDocument Conventions ...................................................................................................... v

Chapter 1: Introduction ....................................................................................................... 11.1

1.2

1.3

1.4

Chapter 2.1

2.2

2.3

2.4

2.5

Chapter 3.1

3.2Liaison 5.5 User Manual iii

About Liaison............................................................................................................. 1

Liaison Binding Energy Models.............................................................................. 21.2.1 LIA Model Equation............................................................................................. 21.2.2 LiaisonScore Model Equation ............................................................................. 3

Running Schrdinger Software .............................................................................. 3

Citing Liaison in Publications ................................................................................. 3

2: Liaison Tutorial ............................................................................................... 5Preparation ................................................................................................................. 5

Starting Maestro ....................................................................................................... 6

Running the Liaison Simulations ........................................................................... 62.3.1 Setting Up the System ........................................................................................ 72.3.2 Starting and Monitoring the Liaison Job ............................................................. 8

Generating and Validating an LIA Model of Binding Affinity ............................ 92.4.1 Importing Liaison Results into Strike for Model Creation .................................... 92.4.2 Selecting Training Set Molecules ...................................................................... 112.4.3 Creating the LIA Model to Predict Binding Affinities ......................................... 112.4.4 Analyzing the LIA Binding Affinity Model .......................................................... 132.4.5 Predicting Binding Affinities with the LIA Model................................................ 132.4.6 Analyze LIA Binding Affinity Predictions for the Test Set .................................. 152.4.7 Making Predictions for Additional Molecules..................................................... 16

Generating and Applying LiaScore and ELR Binding Affinity Models.......... 16

3: Protein and Ligand Preparation........................................................ 19Protein Preparation ................................................................................................. 19

Checking the Protein Structures .......................................................................... 213.2.1 Checking the Orientation of Water Molecules................................................... 22

Contents

Liaison 5.5 iv

3.2.2 Checking for Steric Clashes.............................................................................. 223.2.3 Resolving H-Bonding Conflicts ......................................................................... 22

3.3 Ligand Preparation ................................................................................................. 23

Chapter 4.1

4.2

4.3

Chapter 5.1

5.2

5.3

5.4

Getting H

Index ......User Manual

3.3.1 Using LigPrep for Ligand Preparation ............................................................... 243.3.2 Using Other Programs for Ligand Preparation.................................................. 26

4: Running Liaison........................................................................................... 27Overview of Liaison Tasks..................................................................................... 27

The Liaison Panel .................................................................................................... 294.2.1 Systems Tab ..................................................................................................... 294.2.2 Parameters Tab................................................................................................. 304.2.3 Analysis Tab...................................................................................................... 32

Running Liaison From the Command Line ........................................................ 33

5: Technical Notes for Liaison.................................................................. 35Models of Ligand Binding...................................................................................... 355.1.1 Limitations of Free-Energy Perturbation ........................................................... 365.1.2 Advantages of Linear Response Methods ........................................................ 365.1.3 Advantages of Liaison....................................................................................... 37

LiaisonScore Binding Free Energy Model .......................................................... 38

Application to HIV Reverse Transcriptase .......................................................... 39

Application to -Secretase (BACE) Inhibitors.................................................... 43

elp ............................................................................................................................. 45

........................................................................................................................................ 49

Document Conventions

In additionthis docume

Links to oththis: Docum

In descriptienclose a csymbol | ssyntax that

File name,tions. To obpath or dire% at each e

In this docuenter text m

References

Font

Sans serif

Monospace

Italic

Sans serifuppercaseLiaison 5.5 User Manual vto the use of italics for names of documents, the font conventions that are used innt are summarized in the table below.

er locations in the current document or to other PDF documents are colored likeent Conventions.

ons of command syntax, the following UNIX conventions are used: braces { }hoice of required items, square brackets [ ] enclose optional items, and the bareparates items in a list from which one item must be chosen. Lines of commandwrap should be interpreted as a single command.

path, and environment variable syntax is generally given with the UNIX conven-tain the Windows conventions, replace the forward slash / with the backslash \ inctory names, and replace the $ at the beginning of an environment variable with and. For example, $SCHRODINGER/maestro becomes %SCHRODINGER%\maestro.

ment, to type text means to type the required text in the specified location, and toeans to type the required text, then press the ENTER key.

to literature sources are given in square brackets, like this: [10].

Example Use

Project Table Names of GUI features, such as panels, menus,menu items, buttons, and labels

$SCHRODINGER/maestro File names, directory names, commands, envi-ronment variables, and screen output

filename Text that the user must replace with a valueCTRL+H Keyboard keys

Liaison 5.5 vi User Manual

Chapter 1

Liaison User Manual

Chapter 1: Introduction

1.1 ALiaison pre(LIA) modetraining setcomplex attion data anand . Theswith the sasampling tim

Liaison is rfrom Maesdescribed innotes, backg

Maestro isis designedfacilitate thmain Maestfor automatsive molecuabout the M

Protein Prstructures tProtein Predescribed in

The Impaccan be carrminimizatiopendently oCommand R

The StrikeLiaison. FoLiaison 5.5 User Manual 1bout Liaisondicts ligand-receptor binding affinities using a linear interaction approximationl that has been fitted to a set of known binding free energies. For each ligand in the, Liaison runs molecular mechanics (MM) simulations of the ligand-receptorboth endpoints of the binding process, bound ligand and free ligand. The simula-d empirical binding affinities are analyzed to generate the Liaison parameters: , ,e parameters are subsequently used to predict binding energies for other ligandsme receptor. Liaison simulations use a continuum solvation model to shortenes and speed convergence.

un primarily from the Maestro graphical user interface. A tutorial in using Liaisontro appears in Chapter 2. Liaison can also be run from the command line, as

Chapter 5. Utilities and scripts are run from the command line. Liaison technicalround, and references are provided in Chapter 5.

Schrdingers powerful, unified, multi-platform graphical user interface (GUI). Itto simplify modeling tasks, such as molecule building and data analysis, and also toe set up and submission of jobs to Schrdingers computational programs. Thero features include a project-based data management facility, a scripting languageing large or repetitive tasks, a wide range of useful display options, a comprehen-lar builder, and surfacing and entry plotting facilities. For detailed informationaestro interface, see the Maestro online help or the Maestro User Manual.

eparation is strongly recommended for protein and protein-ligand complex PDBo be used in Liaison. In most cases, this can be performed in Maestro, using theparation Wizard panel on the Workflows menu. Protein and ligand preparation are

Chapter 3.

t computational program runs the MM calculations for Liaison simulations, whichied out using molecular dynamics (MD), hybrid Monte Carlo (HMC), or energyn. Impact uses an OPLS-AA force field. Impact calculations can also be run inde-f Liaison. For more information, see the Impact User Manual and the Impacteference Manual.

statistical analysis package is used for the fitting and prediction analysis tasks ofr more information, see the Strike User Manual.


Liaison 5.5 2

1.2 Liaison Binding Energy Models

1.2.1 LIA Model EquationA Liaison sbuild a modgies. The aenergy of ththis type is(LIA), or a A novel feaexplicit solSOlvatioN.(Hansson,charging inwhich the inand bound s

G =

Here < > resents the frethe van der(SGB) contsurface area

e surface area lost by contact with the receptor. The net electrostatic interaction-ntinuum solvent is given by:

Ucoul + 2 Urxnf

l is the Coulomb interaction energy and Urxnf is the SGB-solvent reaction-fielde factor of 2 compensates for the division by 2 made in the definition of the reac-ee energy.)lications, the coefficients , , and are determined empirically by fitting to thelly determined free energies of binding for a training set of ligands. In such appli-

isons simulation task is used to calculate the values of Uvdw, U elec, and U cav for theplexed) and unbound (free) states of the training-set ligands, and its analysis task isve values for the , , and fitting coefficients. The fitted equation can then be usedhe binding affinities of additional ligands. In the current version of Liaison, a


constant term is added to G in the fitting process, and is adjusted during the fit. This corre-sponds to an extension of the strict linear response model.

1.2.2 LiaisonScore Model EquationLiaison alsosimulation.Score in Liusing the al

G =

where a is t

1.3 RTo run anyhost from ainstallationcommand a

Once you hfollowing c

$SCHRODINIt is usuallyThis directoMaestro, inUser Manu

1.4 CThe use of t

Liaison, verLLC, New

csh/tcsh:bash/ksh:Liaison 5.5 User Manual 3

calculates a scoring function similar to GlideScore over the course of the LRMThis scoring function, previously called GlideScore in Liaison, is called Liaison-aison 5.5. The average LiaisonScore can then be used to predict binding energiesternate model:

a() + bhe LiaisonScore coefficient and b is a constant.

unning Schrdinger SoftwareSchrdinger program on a UNIX platform, or start a Schrdinger job on a remoteUNIX platform, you must first set the SCHRODINGER environment variable to thedirectory for your Schrdinger software. To set this variable, enter the followingt a shell prompt:

ave set the SCHRODINGER environment variable, you can start Maestro with theommand:

GER/maestro &a good idea to change to the desired working directory before starting Maestro.

ry then becomes Maestros working directory. For more information on startingcluding starting Maestro on a Windows platform, see Section 2.1 of the Maestroal.

iting Liaison in Publicationshis product and its components should be acknowledged in publications as:

sion 5.5, Schrdinger, LLC, New York, NY, 2009; Strike, version 1.8, Schrdinger,York, NY, 2009.

setenv SCHRODINGER installation-directoryexport SCHRODINGER=installation-directory

Liaison 5.5 4 User Manual

Chapter 2

Liaison User Manual

Chapter 2: Liaison Tutorial

This chapteality of Liaaffinities. Itapply the Latom locatiowhich mayand appliedtwo steps, tand applica

The exercis

How t

How tusing

How tfor no

How tdict bi

You will uspanels to cr

To do theseand Strike 1

2.1 PBefore youtree. Someallow yoututorial/be generateLiaison 5.5 User Manual 5r contains tutorial exercises to help you quickly become familiar with the function-ison using the Maestro interface. Liaison is used to simulate and predict bindingdoes so by generating for each protein-ligand complex the descriptors necessary toIA equation, the LiaScore, which is the GlideScore computed over discrete proteinns rather than over a gridded protein representation, and the LiaScore components

be used to generate an ELR model. Models for the binding affinity are then createdfrom Liaison-generated descriptors via Strike. Thus, the Liaison process involves

he simulation of binding in Liaison to generate a set of descriptors and the creationtion of binding affinity models from the Liaison descriptors with Strike.

es in this chapter demonstrate:

o perform Liaison simulations on multiple ligands

o create a validated model for the , , and coefficients for the LIA equationStrike

o generate and apply the results of Liaison simulations to predict binding affinitiesvel ligands

o create and apply LiaScore and Extended Linear Response (ELR) models to pre-nding affinities

e the Liaison panel to set up and run Liaison simulations and then use the Strikeeate, validate, and apply binding affinity models from Strike.

exercises you must have access to an installed version of Maestro 9.0, Liaison 5.5.8. For installation instructions, see the Installation Guide.

reparationstart Maestro and begin the exercises, you must first create a local tutorial directoryexercises in this tutorial produce files that are needed in subsequent exercises. Toto begin at any exercise you choose, the $SCHRODINGER/impact-vversion/liaison directory contains copies of the relevant input and output files that will

d as you perform the tutorial.


Liaison 5.5 6

To create a local directory tree:

1. At a shell prompt, change to a directory in which you have write permissions.

2. Create a local base directory:

m

In the

3. In thetutor

l

2.2 SYou do notbefore, this

To start Ma

1. Set thson ar

2. Chang

c

3. Enter

$You are now

2.3 RIn this exercjob that calfile, allowinELR model

The Liaisonsingle 2.8 G

csh/tcsh/basUser Manual

kdir basedir

text, this directory is referred to as the base directory.

base directory, create a soft link to the $SCHRODINGER/impact-vversion/ial/liaison directory by entering the following command:

n s $SCHRODINGER/impact-vversion/tutorial/liaison .

tarting Maestroneed to start Maestro until you begin an exercise. If you have not started Maestro

section contains instructions.

estro:

e SCHRODINGER environment variable to the directory in which Maestro and Liai-e installed:

e to the desired working directory.

d basedir

the command:

SCHRODINGER/maestro & ready to proceed with the exercises below.

unning the Liaison Simulationsise, starting with a receptor and a set of prepared ligands, you will set up and run a

culates Liaison descriptors. The descriptors are saved in a Maestro file and a CSVg you to use them in the creation, validation, and application of LIA, LiaScore ands of binding affinity.

calculations to be run in this exercise require about 1.5 hours of CPU time on aHz Xeon processor, though by taking advantage of multiple processors this time

sh: setenv SCHRODINGER installation_pathh/ksh: export SCHRODINGER=installation_path


can be reduced sharply. The output files from the Liaison simulation have been prepared forthis exercise so the tutorial may be completed whether you choose to run the Liaison simula-tions or not.

2.3.1 SBefore runnand ligands

1. From

The Li

2. In thefile is s

3. Click

4. In the1rt1_

The ligand-Liaison 5.5 User Manual 7

etting Up the Systeming Liaison on a series of ligand-receptor complexes, you must import the receptor and set the parameters for the Liaison simulations.

the Applications menu in the main window, choose Liaison.

aison panel opens with the Systems tab displayed.

Specify structures section, ensure that Take complexes from a Maestro Pose Viewerelected.

Browse.

file selector, navigate to your basedir/liaison directory and import the filehept_analogs_pv.mae.

protein complexes to be simulated have now been specified.

Figure 2.1. The Systems tab of the Liaison panel.


Liaison 5.5 8

To prepare the Liaison simulation parameters:

5. In the Job options section of the Parameters tab, choose OPLS_2001 from the Force fieldoption menu.

6. In the

2.3.2 SThe job takdo not needstarting wit

With the lisimulation c

1. Click

The S

2. Chang

3. Select

To runcessor

otherw

4. Click

The M

While the hdisplays thefinished. Tsubprocess

Before the jhept_anal

hept_anla

hept_anal

hept_analUser Manual

Specify restrained/frozen shells section, choose Huge from the option menu.

tarting and Monitoring the Liaison Jobes approximately 1.5 hours on a 2.8 GHz Xeon processor. The Liaison simulationsto be run to continue with the tutorial. If you prefer, you may continue the tutorial

h Section 2.4.

gands and receptors defined and Liaison simulation parameters set, the Liaisonan be started.

Start.

tart dialog box opens.

e the Job Name to hept_analogs.

the host where the Liaison simulations are to be run from the Host option menu.

on the local machine ensure that localhost is selected. If the host is a multipro-machine, specify the number of available CPUs in the Use N Processors text box,ise leave its value at one.

Start.

onitor panel opens with the File tab displayed.

ept_analogs job is in progress, the Status column in the Jobs tab for this jobtext running. When the job is complete, the status changes to incorporated :

he hept_analogs job spawns a subprocess named hept_analogs_sim. Thisin turn spawns a job for each ligand-receptor complex.ob is launched the following input files are written:ogs.inp LSBD command input to run Liaison simulationogs-ligs.mae Ligand structure input for Liaison simulationogs-rec.mae Receptor structure input for Liaison simulationogs-lia_cons.mae Ligand structure from which restrained/frozen receptor

atoms are determined


When the Liaison simulation finishes, the calculated Liaison results are incorporated into theProject Table along with the input ligand geometries, and the basedir directory will contain thefollowing job output files:

If you wantthe project i

2.4 GA

Before youtory tree, astory tree, do

A Strike lic

2.4.1 I

You will bethe LiaisondescriptorsFor this tuto

1. Choos

A proj2. Enter

3. Click

The P

hept_anal

hept_anal

hept_anal

hept_analLiaison 5.5 User Manual 9

to stop working on the tutorial now, choose Close Project from the Project menu. Ifs a scratch project, you will be prompted to save it or delete it.

enerating and Validating an LIA Model of Bindingffinitybegin the exercises shown below, you must first have created a local tutorial direc-described in Section 2.1 on page 5. If you have not yet set up your tutorial direc- so now, then proceed with the exercises.

ense is required for these exercises.

mporting Liaison Results into Strike for Model Creationimporting the Liaison descriptors that will be used to create the LIA model throughpanel. The data could also be imported directly into the Project Table. The Liaisonare stored in both the hept_analog-final.mae and hept_analog.csv files.rial you will import the data using the Maestro file, into a new project.e New from the Project menu in the main window.ect selector is displayed.liasim in the Name text box, and click Save.

Open/Close Project Table on the main toolbar.

roject Table panel is displayed.

ogs.log LSBD log summary fileogs-final.mae Structure file of input ligand geometries with calculated

Liaison resultsogs.csv Comma-separated value file with calculated Liaison

resultsogs.tar.gz Compressed tarball containing the full set of Liaison

results


Liaison 5.5 10

4. Choos

The Li

5. In the

A file

6. Navig

If youdirecto

7. In the

The mTable,

8. Close User Manual

Figure 2.2. The Analysis tab of the Liaison panel.

e Liaison from the Applications menu in the main window.

aison panel opens with the Systems tab displayed.

Analysis tab, click Browse.

selector opens.

ate to and import the file hept_analogs-final.mae.

ran the Liaison simulation described in Section 2.3, this file is in your basedirry. Otherwise, you can find a copy of it in your basedir/liaison directory.

Analysis tab, click Fit or Predict in Strike.

olecules and data are imported from hept_analogs-final.mae into the Project and the Strike Build QSAR Model panel opens.

the Liaison panel.


2.4.2 Selecting Training Set MoleculesIt is often important before model generation to separate available data into training and testsets. The training set is used to train a model to predict binding affinities. The resultant modelis then validset. The curproperty. TyProject TablStrike creatLIA modelmolecules a

1. In the

The Ethe Pr

2. Select

3. Select

4. Click

The E

5. Click

The Eing se

2.4.3 CThe experimresponse oroutlined in, an

In this exeselected inthe exerciseApplicationsLiaison 5.5 User Manual 11

ated by applying it to the prediction of binding affinities for molecules in the testrent data set has been separated into training and test sets as indicated in the Setpically this separation is done with the Random option in the Select menu of thee panel.

es models using selected entries in the Project Table. Since we are to generate anusing only molecules in the training set, its important to ensure only training setre selected in the Project Table:Project Table panel, choose Select > Only.ntry Selection panel opens. In the Properties tab there is a list of properties stored inoject Table.Set from the list in the Properties tab.

Matches and enter Train in the text field.

Add.

SL text box is updated with the chosen matching condition.

OK.

ntry Selection panel closes, and in the Project Table only the molecules in the train-t are selected.

reating the LIA Model to Predict Binding Affinitiesental binding affinities given in the property Activity (kcal/mol) will be used for the

dependent variable from which a linear model will be created. The LIA equation isdetail in the Liaison Manual and it estimates binding affinities using ,

d terms from Liaison.

rcise, you will create an LIA model for the twelve training set molecules youthe previous exercise. The Build QSAR Model panel should already be open fromin Section 2.4.1. If not, choose Build QSAR Model from the Strike submenu of the menu in the main window.


Liaison 5.5 12

1. In the, and Liaison .

ntrol-click to select the second and third of these descriptors. These are the termse used for an LIA model, and will be the independent variables in the model.

the Regression method option menu, choose Multiple Linear Regression.

that Use all selected is selected.

Choose, to the right of the Activity Property text box.

hoose Activity Property dialog box opens.

the Activity (kcal/mol) property and click OK.roperty is the response or dependent variable which a model will be created tot.


6. Click Start.

A Start dialog box opens.

7. Enter lrm_model as the job name.8. Click

The Strikeincorporaterow with na

2.4.4 A

Once the LIpossesses pwe will viewmodel, and

The basics oprovide anthese includrial we arelog unit, ancoefficientsOPLS_2005not need tothe fundam

Next, you w

1. Click

The Vmodelbasic rdepen

2.4.5 P

Once the Lbinding affiapplied to pcalculated tProject TabLiaison 5.5 User Manual 13

Start to begin the Strike job.job should finish in a matter of seconds and the predicted binding affinities ared in the Project Table as Predicted Activity1.1. In the Build QSAR Model panel, ame lrm_model.1.1 is added to the Results table that corresponds to the LIA model.

nalyzing the LIA Binding Affinity ModelA model has been created, it must be analyzed to see if it makes intuitive sense andredictive power that does not arise by chance. From the Build QSAR Model panel

a plot of predicted versus experimental activities, analyze the fundamentals of theassess its predictive power prior to making predictions on test set molecules.

f each model are displayed in the Results table in the Build QSAR Model panel andat-a-glance overview of a model. For a multiple linear regression (MLR) modele the standard deviation, R2, F-statistic, and P-value. For the purposes of this tuto-

interested in models with an R2 greater than 0.6, a standard deviation lower than 1d a P-value less than 0.05. For an LIA model to make intuitive sense the , , and

calculated when using the OPLS_2001 force-field should all be positive. Forthe cavity term is calculated in a different fashion, and the gamma coefficient does

be positive for the LIA model to make intuitive sense. For further information onental metrics of an MLR model see the Strike User Manual.

ill view all the available information on the LIA model.

View in the Build QSAR Model panel.

iew QSAR Model panel opens. This panel displays the Strike output from LIAcreation. From this display all the information on the model is available, from

egression statistics to validation tests such as the results of the leave-group-out anddent variable randomization testing results.

redicting Binding Affinities with the LIA ModelIA model has been analyzed and found to be suitable, it may be applied to predictnities for molecules in the test set. In a similar fashion the LIA model may beredict binding affinities for any molecule for which the LIA descriptors have beenhrough Liaison. What is essential is that the molecules must be imported into thele where they can be acted upon by Strike.


Liaison 5.5 14

First the traselection inTrain as th

1. In the

The si

2. In the

The P

3. In the

This s

4. Click

A Star

5. Enter

6. Choos

The Strikethe LIA moUser Manual

Figure 2.4. The Strike Predict based on QSAR model panel.

ining set molecules must be selected in the Project Table. If you have changed thethe Project Table, repeat the exercise in Section 2.4.2, using Test instead of

e text to match, and skip to Step 2.

Project Table panel, choose Select > Invert.x test set molecules are selected in the Project Table.Build QSAR Model panel, click Predict.

redict based on QSAR model panel opens, and the Build QSAR Model panel closes.

Select model to use for prediction table ensure that the LIA model is selected.

hould be the model named lrm_model.1.1.

Start.

t dialog box opens.

lrm_pred as the job name.e a host, then click Start to begin the Strike job.job should finish in a matter of seconds, and the predicted binding affinities fromdel are incorporated into the Project Table as Activity (kcal/mol) Strike Prediction.


2.4.6 A

We are intecules not inaffinities mbetween theexercise yobetween the

1. Ensure

This s

2. Choos

The S

3. Enter

4. Selectbetwe

5. ClickLiaison 5.5 User Manual 15

Figure 2.5. The Strike Univariate and Bivariate statistics panel.

nalyze LIA Binding Affinity Predictions for the Test Setrested in viewing how well the LIA model predicted binding affinities for mole-cluded in the training set. Qualitatively the predicted and experimental binding

ay be plotted using the plotting facilities of Maestro. Quantitatively, the correlationpredicted and experimental binding affinities is given in the R2 statistic. In this

u will generate an R2 statistic for the test set molecules showing the correlation predicted and experimental binding affinities.

that only the six test set molecules are selected in the Project Table.hould be the case from the previous exercise unless you have changed the selection.

e Applications > Strike > Statistics in the main window.

trike Univariate and Bivariate statistics panel opens.

lrm_pred as the job name.Activity (kcal/mol) and Activity(kcal/mol) StrikePrediction to analyze the correlation

en these two properties.

Start to begin the calculation.


Liaison 5.5 16

6. The calculation should take only a few seconds. Once complete the bivariate and univari-ate statistics are displayed in the Results section, including the R2 statistic.

2.4.7 Making Predictions for Additional MoleculesOnce an LIities may bThe steps a

1. Run a

2. Impor

You ca

3. Open

4. Ensure

5. Click

The geityX.1

2.5 GB

In additioncalculated pdure is theLiaison , Liaison , and Liaison terms from the Select descriptors to bethe model table of the Build QSAR Model panel, simply select the LiaScore

Model generation, validation, and application then follow exactly as with the LIAple Strike input and output for the generation and application of the LiaScore onlye current dataset may be found in:

ODINGER/impact-vversion/tutorial/liaison/analysis/ls_*terested in expanding the range of descriptors to correlate with binding affinities,linear response (ELR) model may be most appropriate. An example of an ELR

ld be to use the LiaScore components, eight descriptors named with a lia prefix, toinding affinity model. With an ELR model you can take advantage of Strikes auto-

ble selection or sophisticated partial least squares and principal component linear


fitting. Sample input and output for the generation and application of an ELR model usingautomatic variable selection starting with the eight LiaScore components may be found in:

$SCHRODINGER/impact-vversion/tutorial/liaison/analysis/liacomp_*Liaison 5.5 User Manual 17

Chapter 3

Liaison User Manual

Chapter 3: Protein and Ligand Preparation

The qualityand the ligaPreparationstructures fsive ligandand ligand s

3.1 PA typical Pmetal ions,or charges.ysis cannotatom forcestates must

This sectionbe performmenu on thePreparation

After proceand the ligamost cases,to determin

You may ondescribed in

1. Impor

The pligand

2. Simpl

For costructuLiaison 5.5 User Manual 19of Liaison results depends on reasonable starting structures for both the proteinnd. Schrdinger offers a comprehensive protein preparation facility in the ProteinWizard, which is designed to ensure chemical correctness and to optimize protein

or use with Glide and other products. Likewise, Schrdinger offers a comprehen-preparation facility in LigPrep. It is strongly recommended that you process proteintructures with these facilities in order to achieve the best results.

rotein PreparationDB structure file consists only of heavy atoms, can contain waters, cofactors, andand can also be multimeric. The structure generally has no information on bondingTerminal amide groups can also be misaligned, because the X-ray structure anal-usually distinguish between O and NH2. For Liaison calculations, which use an all-field, atom types and bond orders must be assigned, the charge and protonationbe corrected, side chains reoriented if necessary, and steric clashes relieved.

provides an overview of the protein preparation process. The entire procedure caned in the Protein Preparation Wizard panel, which you open from the Workflows

main toolbar. This tool and its use is described in detail in Chapter 2 of the Protein Guide.

ssing, you will have files containing refined, hydrogenated structures of the ligandnd-receptor complex. The prepared structures are suitable for use with Liaison. Innot all of the steps outlined need to be performed. See the descriptions of each stepe whether it is required.

occasion want to perform some of these steps manually. Detailed procedures areChapter 3 of the Protein Preparation Guide.

t a ligand/protein cocrystallized structure, typically from PDB, into Maestro.

reparation component of the protein preparation facility requires an identified.

ify multimeric complexes.

mputational efficiency it is desirable to keep the number of atoms in the complexre to a minimum. If the binding interaction of interest takes place within a single


Liaison 5.5 20

subunit, you should retain only one ligand-receptor subunit to prepare for Liaison. If twoidentical chains are both required to form the active site, neither should be deleted.

Determine whether the protein-ligand complex is a dimer or other multimer con-taining duplicate binding sites and duplicate chains that are redundant.

Is

3. Locate

Waterjudgedthey amore

makin

Theseattachwatersare keprepar

4. AdjusProbleIncomare dishould

MetalroModatom tFurtheGB, G

It mayreceptassignbenefimust bGlu 14the cotallizeligandUser Manual

f the structure is a multimer with duplicate binding sites, remove redundant bindingites and the associated chains by picking and deleting molecules or chains.

any waters you want to keep, then delete all others.

molecules in the crystallographic complex are generally not used unless they arecritical to the functioning of the proteinligand interaction. When waters are used,

re later included in the protein as structural waters. Keeping structural waters islikely to be important for Liaison than for other programs such as Glide, whereg a site more accessible by removing all waters may be necessary for docking.

waters are identified by the oxygen atom, and usually do not have hydrogensed. Generally, all waters (except those coordinated to metals) are deleted, but

that bridge between the ligand and the protein are sometimes retained. If waterspt, hydrogens will be added to them by the preparation component of the proteination job. Afterwards, check that these water molecules are correctly oriented.t the protein, metal ions, and cofactors.ms in the PDB protein structure may need to be repaired before it can be used.plete residues are the most common errors, but may be relatively harmless if theystant from the active site. Structures that are missing residues near the active site be repaired.

ions in the protein complex cannot have covalent bonds to protein atoms. The Mac-el atom types for metal ions are sometimes incorrectly translated into dummyypes (Du, Z0, or 00) when metal-protein bonds are specified in the input structure.rmore, isolated metal ions may erroneously be assigned general atom types (GA,C, etc.).be necessary to adjust the protonation of the protein, which is crucial when the

or site is a metalloprotein such as thermolysin or an MMP. In such a case, Glides a special stability to ligands in which anions coordinate to the metal center. Tot from this assignment, groups such as carboxylates, hydroxamates, and thiolatese anionic. The protein residues that line the approach to the metal center (such as3 and His 231 in thermolysin) need to be protonated in a manner compatible with

ordination of an anionic ligand such as a carboxylate or hydroxamate. The co-crys-d complex therefore needs to be examined to determine how the protein and thes should be protonated. In some cases, two or more protonation states of the protein


may need to be used in independent docking experiments to cover the range of physicallyreasonable ligand dockings.

Cofactors are included as part of the protein, but because they are not standard residues itis sometimes necessary to use Maestros structure-editing capabilities to ensure that mul-tiple b

F C I

t

S S

5. AdjusIf thethey minteracfound

If yousites, t

6. Run a

This ipotent

7. Review

Ip

Eo

3.2 CAfter you hprotein struLiaison 5.5 User Manual 21

onds and formal charges are assigned correctly.

ix any serious errors in the protein.heck the protein structure for metal ions and cofactors.

f there are bonds to metal ions, delete the bonds, then adjust the formal charges ofhe atoms that were attached to the metal as well as the metal itself.et charges and correct atom types for any metal atoms, as needed.et bond orders and formal charges for any cofactors, as needed.

t the ligand bond orders and formal charges.complex structure contains bonds from the ligand or a cofactor to a protein metal,ust be deleted. Glide models such interactions as van der Waals plus electrostatictions. Glide cannot handle normal covalent bonds to the ligand, such as might bein an acyl enzyme.

are working with a dimeric or large protein and two ligands exist in two activehe bond orders have to be corrected in both ligand structures.

restrained minimization of the protein structure.

s done with impref, and should reorient side-chain hydroxyl groups and alleviateial steric clashes.

the prepared structures.

f problems arise during the restrained minimization, review the log file, correct theroblems, and rerun.xamine the refined ligand/protein/water structure for correct formal charges, bondrders, and protonation states and make final adjustments as needed.

hecking the Protein Structuresave completed the protein preparation, you should check the completed ligand andctures.


Liaison 5.5 22

3.2.1 Checking the Orientation of Water MoleculesYou only need to perform this step if you kept some structural waters. Reorienting the hydro-gens is not strictly necessary, as their orientation should have been changed during refinement,but it is use

If the orienSection 3.9

When yourefinement

3.2.2 CYou shouldrestraint-op

Steric clashContacts fobad contactder Waals rcontacts. Bdisplays on

If steric claprocedure,default of 0the prepared

3.2.3 R

You shouldthe ligand ointerveningacceptor-ac

Some of theThe preparacally occursone or more

reasonablerepresentedUser Manual

ful to check that the orientation is correct.

tation is incorrect, reorient the molecules by using the procedure outlined in of the Protein Preparation Guide.

have corrected the orientation of the retained water molecules, you should run aon the adjusted protein-ligand complex.

hecking for Steric Clashesmake sure that the prepared site accommodates the co-crystallized ligand in the

timized geometry obtained from the structure preparation.

es can be detected by displaying the ligand and protein in Maestro and using thelder in the Measurements panel to visualize bad or ugly contacts. Maestro definess purely on the basis of the ratio of the interatomic distance to the sum of the vanadii it assigns. As a result, normal hydrogen bonds are classified as bad or uglyy default, Maestro filters out contacts that are identified as hydrogen bonds, andly the genuine bad or ugly contacts.

shes are found, repeat the restrained optimization portion of the protein preparationbut allow a greater rms deviation from the starting heavy-atom coordinates than the.3 . Alternatively, you can apply an additional series of restrained optimizations to ligand-protein complex to allow the site to relax from its current geometry.

esolving H-Bonding Conflictslook for inconsistencies in hydrogen bonding to see whether a misprotonation ofr the protein might have left two acceptor atoms close to one another without anhydrogen bond. One or more residues may need to be modified to resolve such anceptor or donor-donor clash.

se clashes are recognized by the preparation process but cannot be resolved by it.tion process may have no control over other clashes. An example of the latter typi-in an aspartyl protease such as HIV, where both active-site aspartates are close toatoms of a properly docked ligand. Because these contact distances fall within any

cavity radius, the carboxylates are not subject to being neutralized and will both beas negatively charged by the preparation process. However, when the ligand inter-


acts with the aspartates via a hydroxyl group or similar neutral functionality, one of the aspar-tates is typically modeled as neutral.

If residues need to be modified, follow these steps:

1. Place

2. Exami

3. Use yotomeri

4. Use ththe Pr

5. Re-mi

It is usuallyzation to colarge-scalewithout daninal complechanged theit is best to

3.3 LiTo give theligand structures suppli

1. They m

2. They m

3. Theywith n

4. They m

5. They m

For exprotonacceptLiaison 5.5 User Manual 23

the refined protein-ligand complex in the Workspace.

ne the interaction between the ligand and the protein (and/or the cofactor).ur judgment and chemical intuition to determine which protonation state and tau-c form the residues in question should have.

e structure-editing capabilities in Maestro to resolve the conflict (see Section 3.8 ofotein Preparation Guide for procedures).nimize the structure.

sufficient to add the proton and perform about 50 steps of steepest-descent minimi-rrect the nearby bond lengths and angles. Because this optimizer does not make

changes, the partial minimization can be done even on the isolated ligand or proteinger of altering the conformation significantly. However, if comparison to the orig-x shows that the electrostatic mismatch due to the misprotonation has appreciablypositions of the ligand or protein atoms during the protein-preparation procedure,

reprotonate the original structure and redo the restrained minimization.

gand Preparationbest results, the structures that are used must be good representations of the actualtures as they would appear in a protein-ligand complex. This means that the struc-ed to Liaison must meet the following conditions:

ust be three-dimensional (3D).ust have realistic bond lengths and bond angles.

must each consist of a single molecule that has no covalent bonds to the receptor,o accompanying fragments, such as counter ions and solvent molecules.

ust have all their hydrogens (filled valences).ust have an appropriate protonation state for physiological pH values (around 7).

ample, carboxylic acids should be deprotonated and aliphatic amines should beated. Otherwise a neutral aliphatic amine could improperly act as a hydrogen-bondor in the docking calculations, or could occupy a hydrophobic region without


Liaison 5.5 24

incurring the large desolvation penalty that XP Glide docking would have assessed if theamine had been properly protonated.

Protonation states are particularly crucial when the receptor site is a metalloprotein suchas thermolysin or a MMP. If the metal center and its directly coordinated protein residueshave ato theamateto anio

6. They m

Maeststro foformamat be

All of the aLigPrep is d

3.3.1 UThe Schrdatom 3D strin SD, Maeline. For de

To run LigPbmin requirof commancommand li

The LigPrethe structurmize the strthe LigPrep

1. Conve

2. Select

3. Add h

4. Remo

5. NeutraUser Manual

net charge, Glide assigns a special stability to ligands in which anions coordinatemetal center. To benefit from this assignment, groups such as carboxylates, hydrox-s, and thiolates must be anionic. If there is no net charge, Glide gives no preferencens over neutral functional groups.

ust be supplied in Maestro, SD, Mol2, or PDB format.

ro transparently converts SD, MacroModel, Mol2, PDB, and other formats to Mae-rmat during structure import. However, Glide has no direct support for other

ts, so you should ensure that your structures are in Maestro, SD, Mol2, or PDB for-fore starting Liaison jobs.bove conditions can be met by using LigPrep to prepare the structures. Use ofescribed in the next section.

sing LigPrep for Ligand Preparationinger ligand preparation product LigPrep is designed to prepare high quality, all-uctures for large numbers of drug-like molecules, starting with 2D or 3D structuresstro, or SMILES format. LigPrep can be run from Maestro or from the commandtailed information on LigPrep, see the LigPrep User Manual.

rep, you must have a LigPrep license. The MacroModel commands premin ande LigPrep licenses when run in a LigPrep context, and are limited to a restricted setds when run using a LigPrep license. LigPrep can be run from Maestro or from thene.

p process consists of a series of steps that perform conversions, apply corrections toes, generate variations on the structures, eliminate unwanted structures, and opti-uctures. Many of the steps are optional, and are controlled by selecting options in panel or specifying command-line options. The steps are listed below.

rt structure format.

structures.

ydrogen atoms.

ve unwanted molecules.

lize charged groups.


6. Generate ionization states.

7. Generate tautomers.

8. Filter structures.

9. Gener

10. Gener

11. Remo

12. Optim

13. Conve

The LigPreApplicationsof the LigP

The simplesfor each sutures from ering conformor specified

The defaultminimize thon panel op

The stereofied maxim

The fifrom 3fied orthat Rwhileity inf

Genertures s

The Ionizatfound in the

RetainLiaison 5.5 User Manual 25

ate alternative chiralities.

ate low-energy ring conformations.

ve problematic structures.

ize the geometries.

rt output file.

p panel allows you to set up LigPrep jobs in Maestro. Choose LigPrep from themenu to open the panel. For details of panel options and operation, see Chapter 2

rep User Manual.

t use of LigPrep produces a single low-energy 3D structure with correct chiralitiesccessfully processed input structure. LigPrep can also produce a number of struc-ach input structure with various ionization states, tautomers, stereochemistries, andations, and eliminate molecules using various criteria including molecular weight

numbers and types of functional groups present.

options in the LigPrep panel remove unwanted molecules, add hydrogens, ande ligand structure (performing a 2D-3D conversion, if necessary). Below are notestions that produce more than one output structure per input structure.

izer can generate two stereoisomers per chiral center in the ligand, up to a speci-um. There are three Stereoisomers options.

rst two options, Retain specified chiralities (the default) and Determine chiralitiesD structure, generate both isomers only at chiral centers where chirality is unspeci-indeterminate; centers with known chirality retain that chirality. The difference is

etain specified chiralities takes its chirality data from the input file (SD or Maestro),Determine chiralities from 3D structure ignores input file chiralities and takes chiral-ormation from the 3D geometry.

ate all combinations varies the stereochemistry up to a maximum number of struc-pecified by Generate at most max per ligand. The default maximum is 32.

ion options allow you to generate all the ligand protonation states that would be specified pH range. The Ionization options are:

original state


Liaison 5.5 26

Neutralize

Generate possible states at target pH target +/- range. This is the default, and can gener-ate several different output structures for each input structure. The default pH target is 7.0with a +/- range of 2.0, so the default pH range is 5.0 9.0. Both the target and range set-tings cstates.

Generate lolowest ener

Desalt is se

Generate taligand, seleinput structthen the ta

3.3.2 UIf you prefeinstallationsThese utilitligands in th

Hydro

or the

Hydrodescri

StructpdbcoUser Manual

an be changed. You can use either the ionizer or Epik to generate ionization Epik is a separate product, so you must purchase this product to use it.

w energy ring conformations: number per ligand. The default is to generate only thegy conformation.

lected by default.

utomers is selected by default. The tautomerizer generates up to 8 tautomers percting the most likely tautomers if more than 8 are possible. If you are sure that theures are already in the correct tautomeric form for docking to a particular target,utomerizer should be turned off by deselecting Generate tautomers.

sing Other Programs for Ligand Preparationr to prepare the ligands with other programs, you can do so. Schrdinger softwareinclude a number of utilities that can be used to perform some of the above tasks.

ies are also used by LigPrep. One of these, the Ionizer, can be used to preparee required protonation states. Some of the other tasks can be performed as follows:

gen atoms can be added in Maestro with either the Add hydrogens toolbar button:

Hydrogen Treatment panel (select Hydrogen Treatment from the Edit menu).gen atoms can also be added (or removed) using the utility applyhtreat, which isbed in Appendix D of the Maestro User Manual.

ure file format conversion can be done from the command line with utilities such asnvert, sdconvert, and maemmodsee Appendix D of the Maestro User Manual.

Chapter 4

Liaison User Manual

Chapter 4: Running Liaison

Liaison is abeen fitted tpredicting s

The LiaisonBased Descperformed uthe Strike U

4.1 OBefore youproperly prfile includes

To run a Li

1. SpecifYou crecept

2. Speciftab.

3. Set coters ta

4. Click

5. In the

If youthe cuten tofrom a

When the jvarious comnot incorpoLiaison 5.5 User Manual 27method of predicting ligand-protein binding free energies using a model that haso known binding energy values. The process involves two steps, a fitting step and atep. Each step is carried out as two tasks, a simulation task and an analysis task.

panel is used to set up and run the simulation tasks. It runs the Ligand & Structure-riptors script (lsbd) to set up and run the Liaison job. The analysis tasks aresing the Build QSAR Model panel of Strike. For more information on Strike, seeser Manual. A tutorial introduction to the Liaison tasks is given in Chapter 2.

verview of Liaison Tasksrun a Liaison simulation, you should ensure that the receptor and the ligands are

epared, as described in Chapter 3. You should also ensure that the ligand structure the known binding energies.

aison simulation:

y the systems to be simulated in the Specify structures section of the Systems tab.an take the receptor and ligands from a Glide pose viewer file, or you can take theor from the Workspace or a file and take the ligands from the Project Table or a file.y the kind of system to be simulated in the Job options section of the Parameters

nstraints if required in the Specify restrained/frozen shells section of the Parame-b.

Start.

Start dialog box, set the job name and select the host and number of processors.want to choose a remote machine or batch queue as the host for the job, ensure thatrrent working directory is mounted on the remote host. Liaison input files are writ-the current working directory. Liaison does not have the ability to copy input files local directory to a remote scratch directory.

ob finishes, the results are incorporated into the Project Table. The scores and theponents that are used in the analysis task are added as properties. If the job does

rate, select it in the Monitor panel and click Monitor.


Liaison 5.5 28

To run a Liaison fitting job:1. Select the results of a Liaison simulation:

If the simulation results are already in the Project Table, select the relevant entries,a

Is

ILP

The B

2. Select

F F

3. Choos

4. For De

5. Click

If theyou m

6. Click

7. Set job

To run a Li

1. Selectenergy

2. Choos

3. Select

4. Click

5. Set jobWhen the joadded to theUser Manual

nd choose Build QSAR Model from the Strike submenu of the Applications menu.

f the simulation results include both the training set and the test set, you shouldelect the training set for the fitting.

f the simulation results are not incorporated, click Browse in the Analysis tab of theiaison panel, navigate to and select the file jobname-final.mae, and click Fit orredict in Strike.

uild QSAR Model panel of Strike opens, with the Liaison results loaded.

the descriptors required for the fit from the list (click, shift-click, control-click):or an LRM model, select Uvdw, Uele, and Ucav.or a LiaScore model, select LiaScore.

e Multiple Linear Regression from the Regression method option menu.

scriptors, ensure that Use all selected is selected.

Choose to choose the activity property.

activity property is not in the Project Table, you can import it from a CSV file, butust do this before clicking Choose.

Start.

options in the Start dialog box, and click Start.

aison prediction job:the results of a Liaison simulation for the test set (the ligands whose binding you want to predict) in the Project Table.e Predict from the Strike submenu of the Applications menu.

the model you want to apply from the list.

Start.

options in the Start dialog box, and click Start.b finishes, you can view the results by clicking View. The predicted values are also Project Table.


4.2 The Liaison PanelThe main part of the Liaison panel consists of three tabs:

Systems tab Param Analys

Below thesWrite button

To open the

4.2.1 SThis tab has

Take compl

Use a MaeBrowse to nLiaison 5.5 User Manual 29

eters tabis tab

e tabs on the left are the Start button, for starting the Liaison simulation, and the, for writing the Liaison input files for later use from the command line.

Liaison panel, choose Liaison from the Applications menu in the main window.

ystems Tab a single section, Specify structures, for defining the system to be studied.

exes from a Maestro pose viewer file option and controls

stro (Glide) pose viewer file as the source of the receptor and the ligands. Clickavigate to the file, or enter the path in the File text box.

Figure 4.1. The Systems tab of the Liaison panel.


Liaison 5.5 30

Take complexes from separated ligand & protein structures option and controls

Specify the receptor and the ligand structures separately.

Take ligands from: Select either Selected entries in Project Table or File. If you select a file,click Brows

Take receptnavigate to

4.2.2 P

This tab hadefining res

The optionsoptions sectComplex Siand are des

Sampling m

Choose theMolecular D

Minimizatio

Choose anmethods. ADescent.

Simulation t

Specify thezation samp

Temperatur

Specify the10 ps Not a

Residue-ba

Set the valutions of an apair list if aUser Manual

e to navigate to the file, or enter the path in the File text box.

or from: Select either Workspace entry or File. If you select a file, click Browse tothe file, or enter the path in the File text box.

arameters Tabs two sections, one for setting the simulation parameters (Job options) and one fortrained and frozen shells (Specify restrained/frozen shells).that apply to both the complex and the free ligand are in the upper part of the Job

ion. Options that apply to one or the other are in tabs labeled Ligand Simulation andmulation in the lower part of this section. The controls in these tabs are identical,cribed below.

ethod option menu

method for performing the simulation, from Minimization, Hybrid Monte Carlo, orynamics.

n algorithm option menu

algorithm for performing the minimization steps in any of the three samplingvailable algorithms are Truncated Newton, Conjugate Gradient, and Steepest

emperature text box

temperature of the simulation in K. Default: 300 K Not available with the Minimi-ling method.

e relaxation time text box

time scale, in picoseconds, on which heat is exchanged with the heat bath. Default:vailable with the Minimization sampling method.

sed cutoff distance text box

e (in ) for the cutoff distance of non-bonded interactions. All pairwise interac-tom in one residue with an atom in another residue are included on the non-bonded

ny such pair of atoms is separated by this distance or less. Default: 15 .


Force field o

Select the fLiaison simSelect the Ononpolar m

Use ligand

Select thisfree and the

Maximum m

Specify theligands and

Heating tim

Set the timean HMC orLiaison 5.5 User Manual 31

Figure 4.2. The Parameters tab of the Liaison panel.

ption menu

orce field. The force field options are OPLS_2005 (the default) and OPLS_2001.ulations are run using Surface Generalized Born (SGB) continuum solvation.PLS_2005 force field if you want the calculation to use the improved parametrized

odel instead of the default SGB terms that are used with OPLS_2001.

input partial charges (if they exist) optionoption to use the input charges for the ligand in Liaison calculations (for both the bound state).inimization steps text box

maximum steps to take during any minimization. Can be set independently for complexes. Default: 1000.

e text box

in picoseconds over which the system is heated before the LIA task is started, in MD simulation. Default: 10 ps.


Liaison 5.5 32

Simulation t

Set the simages for the

In the Sperestrained othe frozen rflexible. Reregion bounfrozen. You

Note: Youwhe

4.2.3 A

This tab proanalysis. Yothat containthis file areUser Manual

Figure 4.3. The Analysis tab of the Liaison panel.

ime text box

ulation time for the LIA task, in an HMC or MD simulation. In this task the aver- van der Waals, Coulombic, reaction field and cavity terms are determined.

cify restrained/frozen shells section, you can specify which residues are to ber frozen, by choosing cutoffs for the start of the restrained region and the start ofegion. Residues with any atoms inside the restrained region boundary are treated assidues with any atoms inside the frozen region boundary but outside the restraineddary are restrained. Residues with all atoms outside the frozen region boundary are can choose from a range of options for the location of the two boundaries.

should make sure that you make the same choice of restrained and frozen shellsn you run Liaison for a particular system, otherwise the results will be invalid.

nalysis Tabvides an interface to Strike for performing the fit and predict steps of the Liaisonu can use Strike independently. If you use these controls, you must specify the files the results of the Liaison simulation, named jobname-final.mae. The structures inimported into the Project Table; when you click Fit or Predict in Strike, these entries


are selected in the Project Table and used as input to Strike. Clicking this button opens theStrike Build QSAR Model panel. For more information, see Chapter 3 of the Strike UserManual.

4.3 ROnce you hthen run the

$SCHRODINwhere job-oLiaison 5.5 User Manual 33

unning Liaison From the Command Lineave set up a Liaison job in Maestro, you can click Write to write out the input file, Liaison simulation job from the command line with the following command:GER/lsbd [job-options] input-fileptions are the standard Job Control options, described in the Job Control Guide.

Chapter 5

Liaison User Manual

Chapter 5: Technical Notes for Liaison

5.1 MThe typicalligand is prin this cavit

Figure 5.1water. Th

The princip

The deor patboth eand elonger

The exregionmenta

The eait costto chaFigure

.

.

.

.

.

.

.

.Liaison 5.5 User Manual 35odels of Ligand Bindingbinding site for a ligand is the active-site cavity of a protein receptor. When no

esent, this cavity is filled with water molecules. When a ligand binds to the proteiny, it displaces water molecules in the active site, which return to bulk solvent.

. Schematic view of ligand binding in a receptor cavity with displacement ofe ligand and the receptor both change, to a greater or lesser extent, from a

free (f) to a bound (b) conformation.

al factors determining the strength and specificity of binding are as follows:

gree to which hydrophobic groups on the ligand interact with hydrophobic pocketsches on the protein surface to release water into the bulk. This release is favorablenergetically (more hydrogen bonds are formed by the released water molecules)ntropically (the released waters are less constrained orientationally and are no confined to a restricted cavity).tent to which the ligand forms hydrogen bonds or metal ligations in hydrophilic

s with appropriately placed polar or charged groups on the receptor. Such comple-rity is essential for achieving adequate binding affinity and specificity.

se with which the ligand fits into the protein cavity. An important question is whats the ligand (and the protein) in energy and/or entropy to accomplish this fiti.e.,nge from the free to the bound conformation, as indicated schematically in5.1. Energy will be required if the ligand has to be distorted away from its natu-

Lb RbRf+

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Lf


Liaison 5.5 36

rally preferred, low-energy conformation into a higher-energy conformation when itbinds to the receptor. At the same time, entropy will be lost if the ligand is very flexible insolution and then is confined to a small number of conformations in the receptor cavity.Both effects, as well as similar restrictions on the conformation of protein side chains, actagains

5.1.1 L

Molecular sligand calcapproachesDuring theand in pharachieved, F

FEP capplic

FEP ctime s

Inaccu

5.1.2 A

The limitatiseveral yearstudies by Jdifficulties.

In conLRMtein. Tment o

Only ienter iand prto genant mothe cabecaussimulaUser Manual

t binding.

imitations of Free-Energy Perturbationimulation methods have been used to calculate binding free energies of protein-ulations since the pioneering applications of free energy perturbation (FEP)by McCammon, Kollman, Jorgensen, and others approximately 20 years ago.

past two decades, many FEP calculations have been carried out in academic groupsmaceutical and biotechnology companies. But while notable successes have beenEP methods are in limited use in drug-discovery projects, for several reasons:alculations are typically limited to small changes in ligand structure, restricting theability to the very last phase of lead optimization.

alculations are very expensive computationally, and often cannot be completed on acale compatible with the schedule of a given drug-discovery project.racies in force fields and sampling methods can lead to errors in FEP predictions.

dvantages of Linear Response Methodsons of FEP motivated the development of linear-response (LR) methods by Aqvists ago (Hansson, T.; Aqvist, J. Protein Eng. 1995, 8, 11371145). Since that time,orgensen and others have shown that LR methods can effectively address the above In comparison with the FEP approach, the advantages of LRM are as follows:

trast to FEP, where a large number of intermediate windows must be evaluated,requires simulations only of the ligand in solution and the ligand bound to the pro-he idea is that one views the binding event as replacement of the aqueous environ-f the ligand with a mixed aqueous/protein environment.

nteractions between the ligand and either the protein or the aqueous environmentnto the quantities that are accumulated during the simulation. The protein-proteinotein-water interaction are part of the reference Hamiltonian, and hence are usederate conformations in the simulation, but are not used as descriptors in the result-del for the binding free energy. This eliminates a considerable amount of noise in

lculationsfor example, that arising from variations in the total energy that resulte slightly different geometries of the protein are obtained for each ligand moleculeted.


As long as the binding modes of the ligand are fundamentally similar, LRM calculationscan be applied to ligands that differ significantly in chemical structure.

LRM calculations are less computationally expensive than FEP calculations.

The Lset ofenergyallowsthe corepresabsorbevent,of theouslyvalueparamThis ehave sreturntime,projec

5.1.3 A

LiaisonSattractive fe

The usculatiosolvenmore.

An evenergysimulathe mito a si

WhenNewtotemposcreenLiaison 5.5 User Manual 37

RM approach allows the binding-energy model to be calibrated by using a trainingcompounds for which experimental binding affinities are known. The use of theterms as descriptors in the fitting equation introduces an empirical element thatsome of the limitations in the theoretical framework (for example, the neglect of

st in energy and entropy of fitting the ligand into the protein site) and the physicalentation (as reflected by errors in the force field or solvation model) to be partiallyed into the parameterization. Moreover, some of the steps involved in the bindingsuch as the removal of water from the protein cavity and subsequent introductionligand, are not inherently linear. If the linear-response approximation was rigor-

valid, the coefficients of the terms would each be 0.5, corresponding to the mean-approximation to the charging integral. In practice, optimization of the fittingeters yields coefficients that are significantly different from the ideal value of 0.5.mpirical element sacrifices generality: the method probably requires the ligands toimilar binding modes, and new parameters must be developed for each receptor. In, one can obtain a reasonable level of accuracy with a modest expenditure of CPUunder assumptions that are quite reasonable for many structure-based drug-designts.

dvantages of Liaisonchrdingers continuum-solvent implementation of LRMhas a number of highlyatures in addition to those listed in Section 5.1.2:

e of the Surface Generalized Born (SGB) continuum model greatly speeds the cal-n because the various interaction terms converge much faster than in an explicit-t simulation. As a result, the required CPU time is reduced by a factor of 10 or

en greater reduction in computational effort can be achieved by using a simpleminimization protocol, rather than a molecular dynamics or Hybrid Monte Carlo

tion, and obtaining the LRM fitting data from the lowest-energy point reached bynimization. While there is sometimes a small degradation of accuracy as comparedmulation, the speed of the calculation is qualitatively enhanced.

energy-minimization sampling is coupled with a highly efficient Truncatedn minimizer, Liaison calculations are fast enough to be applied routinely in a con-rary drug-discovery context. This approach makes Liaison very attractive foring a large number of ligands.


Liaison 5.5 38

Schrdingers automatic atom-typing scheme for the OPLS-AA force field (Jorgensen,W. L.; Maxwell, D. S.; Tirado-Rives, J. J. Am. Chem. Soc. 1996, 118, 1122311235)assigns charges, van der Waals, and valence parameters with no human intervention. Akey feature of OPLS-AA is that, via fitting to liquid-state simulations, excellent reproduc-tion of

The Maestra training seto generateOnce the paffinities caperform scobinding ene

It is importGuidefor a

5.2 LiLiaison calcsimulation.Score. Thegies using t

G =

instead of thselected infitting coeffi

This capabii.e., beforeavailable. Trelax and en

AdvantagRecall thatprotein atomsite and impa ligand thprotein. Buothers mayUser Manual

condensed-phase properties is obtained.

o graphical user interface makes it easy to set up and run Liaison calculations. First,t of compounds for which experimental binding affinities are available can be usedsimulation data that are then employed to determine the optimal LRM parameters.arameters have been determined, libraries of compounds with unknown bindingn be run and their binding affinities can be predicted. Alternatively, Liaison canring on relaxed protein-ligand complexes, allowing direct prediction of ligand

rgies.

ant that the protein structures be correctly prepared. See the Protein Preparationdescription of Schrdingers protein-preparation facility.

aisonScore Binding Free Energy Modelulates a scoring function, similar to GlideScore 3.5 SP, over the course of the LRMThis function, previously known as GlideScore in Liaison is now called Liaison-average LiaisonScore found in the simulation can be used to predict binding ener-he formula

a() + be LRM equation described above. The LiaisonScore binding energy model can be

the Analysis tab. The Analyze task can then be used to derive values for the a and bcients.

lity allows Liaison to be employed in the early stages of a drug-discovery project,a set of ligands with known binding affinities suitable for training a LRM model ishis approach transcends scoring in Glide itself because it allows the protein site toables the ligand to undergo full (not just torsional and rigid-body) optimization.

es of LiaisonScoredocking with Glide is normally done with the vdW radii of non-polar ligand and/or

s scaled back by 10 to 20%. This rescaling creates more room in the rigid proteinlicitly allows for breathing motions a protein site may carry out to accommodate

at is slightly larger in some dimension than the ligand co-crystallized with thet while some protein sites may readily expand (or contract) upon ligand binding,be less adaptable, and the device of rescaling radii used by Glide will miss this


distinction. Moreover, Glide further compensates for the limitations of docking into a rigidprotein site by allowing intramolecular contacts within the docked ligand pose to be shorterthan are physically realistic. These two aspects mean that Glide at times will give good scoresto ligands that are too large to bind in a given receptor site. There can also be cases in whichthe rigid sitnon-polar ra

In principlescoring afteto the trueGlideScorere-scored Gsmall will spose into a

CalibratioIf a trainingvalues can b

Gl

Calibrationcorrelationscale and m

5.3 AThe Jorgenapproachesapplicationanalogs (J.encompass29702987class, andFigure 5.2,methyl benzvariations oLiaison 5.5 User Manual 39

e is somewhat too large for an active binder, particularly with the scaling down ofdii. As a result an active ligand may score more poorly than it should.

, such false positives and false negatives can be eliminated by repeating the Glider relaxing the ligand-protein complex. In this way, if the protein can readily adapt(not Glide compacted) size and shape of the docked ligand, the recomputedvalue should be good. But if the protein site cannot respond appropriately, a poorlideScore value should be obtained. Note, however, that sites that are much tooimply generate incorrect docked poses, and minimization will not transform such acorrectly docked structure.

n of GlideScoresset of ligands with known binding affinities is available, calibrated GlideScore

e obtained as

ideScore(calibrated)i = a GlideScore(raw)i + b (1)

has no effect on the rank order of calculated binding affinities or on the computedcoefficient, but can place the computed GlideScores on the correct experimentalake it easier to interpret the calculated values.

pplication to HIV Reverse Transcriptasesen group has been a leader in the development and testing of linear-responsefor systems of pharmaceutical interest using explicit-solvent methods. An initialto HIV reverse transcriptase examined the binding of 20 HEPT and 20 nevirapineMed. Chem. 2001, 44, 145154). More recently, this work has been extended to200 different inhibitors covering 8 chemical classes (J. Med. Chem. 2002, 45,

). Nevirapine is a FDA-approved anti-HIV drug of the non-nucleoside inhibitorone of the HEPT analogs (MKC-442) has been in clinical trials. As shown inthe variation in the R2 sidechain is particularly sizable, ranging from hydrogen toyl ether. It would be very difficult, and very time consuming, to examine structuralf this magnitude using free-energy perturbation methods.


Liaison 5.5 40

F

The Liaisonfrom Glidein each casLiaison LR

Figure 5.3 sresults obtacorrelationexperimentaaffinities wiof the limitaand scoring

The resultssampling anFigure 5.5.relaxationGlideScoreexperimentaThus, the pThe slope oFigure 5.3 a

The convenGlide 3.5 Shigher r2 cobserved bi

O

NHR1 R1 = Me, Et, i-Pr, c-PrUser Manual

igure 5.2. Substituents for the HEPT series of inhibitors for HIV-RT.

calculations presented in this section use input geometries for the ligands obtained3.5 SP and XP dockings using Glide 3.5. Energy-minimization sampling was usede. Calibrated GlideScore values using Glide SP and Glide XP, LiaisonScores, andM-model predictions appear in the figures that follow.

hows the result obtained in the Glide 3.5 SP dockings, while Figure 5.4 shows theined using Glide 3.5 XP scoring. As can be seen, SP Glide gives only a very roughof the docked GlideScore values (shown as the calculated Gbind values) with thel binding affinities. This is not surprising, as correlating experimental bindingll be beyond the capabilities of Glide docking and scoring in some cases, becausetions of rigid-receptor docking discussed above. Figure 5.4 shows that XP docking gives a considerably better correlation.

of applying the LiaisonScore binding energy model, using energy-minimizationd starting from the Glide 3.5 SP docked geometries for the ligands, are shown inThe correlation is much tighter and the largest outliers are gone, showing that

of the protein and docked ligand geometries has improved the results. Thesevalues, unlike those in the previous two figures, have been put on the scale of thel binding affinities by applying the linear transformation shown in Equation (1).

redicted values now cover the approximately same range as the measured values.f the fitted least-squares line is still less than 1, but is larger than those seen innd Figure 5.4.

tional LRM fit, again using energy-minimization sampling and starting from theP ligand poses, is shown in Figure 5.6. As can be seen, LRM scoring produces aorrelation coefficient and a lower RMS deviation between the calculated andnding affinities.

O N

R2

R3 R3 = SPh, CH2Ph, OPh,

SPH-3,5-di-Me

R2 = H, Me, Et, Pr, CH2OCH3, CH2OCH2CH3,CH2OCH2CH2OH, CH2OCH2Ph


Figure 5.Gl

Figure 5.Gl

-7

al/m

ol)Liaison 5.5 User Manual 41

3. Correlation of calculated (GlideScore) and observed binding affinities foride 3.5 SP docking of HEPT inhibitors into HIV reverse transcriptase.

4. Correlation of calculated (GlideScore) and observed binding affinities foride 3.5 XP docking of HEPT inhibitors into HIV reverse transcriptase.

-13 -12 -11 -10 -9 -8 -7 -6 -5-10

-9

-8

r2 = 0.32

ExperimentalGbind (kcal/mol)

Cal

c. G

bin

d (

kc

-13 -12 -11 -10 -9 -8 -7 -6 -5-16

-15

-14

-13

-12

r2 = 0.60


Cal

c. G

bin

d (

kcal

/mo

l)


Liaison 5.5 42

HIV rever

Figure 5.6reverse tr

-7

-6

-5

al/m

ol)User Manual

Figure 5.5. Correlation of LiaisonScores for the HEPT inhibitors forse transcriptase, calculated by energy-minimization sampling starting from

Glide 3.5 SP docked geometries.

. Correlation of Liaison LRM scoring for the HEPT series of inhibitors for HIVanscriptase, calculated by energy-minimization sampling starting from Glide

3.5 SP docked geometries.

-13 -12 -11 -10 -9 -8 -7 -6 -5-14

-13

-12

-11

-10

-9

-8

r2 = 0.77RMSD = 1.08 kcal/mol


Cal

c. G

bin

d (

kc

-13 -12 -11 -10 -9 -8 -7 -6 -5-13

-12

-11

-10

-9

-8

-7

-6

-5

r2 = 0.82RMSD = 1.01 kcal/mol


Cal

c. G

bin

d (

kcal

/mo

l)


These results suggest that some of the limitations expected for rigid-receptor docking can beaddressed by relaxing the docked complexes before scoring them with GlideScore or via astandard LRM fitting model. LRM scoring gives the best results in this instance, but it remainsto be seen whether this will always be the case.

Normally wby rescorinmended progives correlFigure 5.4 aonly one ligmerely conR2groups (presumably

5.4 ATounge andof Liaison tis typicallyneutral inhistructure. Thave two orlogical pH.

Figure 5Liaison 5.5 User Manual 43

e would expect even better results from energy-minimization sampling followedg in Liaison when starting from the XP-docked poses. This indeed is our recom-cedure. In this instance, however, the results obtained using the same two modelsations that are better than those shown in Figure 5.3 but are not as good as those innd Figure 5.5. This is counterintuitive, especially in view of the fact that XP docksand in a manner that is inconsistent with the other cases, and here the variation

sists of a reversal of the positions adopted by the pseudo-symmetric C-R1 and N-see Figure 5.2). In contrast, SP Glide docks three ligands in an inconsistent, and incorrect, manner.

pplication to -Secretase (BACE) InhibitorsReynolds (J. Med. Chem. 2003, 46, 20742082) recently reported the application

o ten -secretase inhibitors of the general structure shown in Figure 5.7, where R1a BOC-capped dipeptide, and R2 and R3 are usually methyl or isopropyl. These

bitors have relatively high molecular weights and resemble HIV protease ligands inounge and Reynolds also studied two nanomolar co-crystallized octapeptides thatthree aspartate or glutamate residues and hence are negatively charged at physio-

.7. Schematic structure of neutral BACE inhibitors studied by Tounge andReynolds.

N NN

O

O O

H

H

R1

R2

R3

H

H

Ph


Liaison 5.5 44

Figure 5.8

A plot of thpresented inreasonablynegative calis notoriousstates to fal

Tounge andfits to the eemploy thei

-9

-8

/mo

l)User Manual

. Correlation of predicted and observed binding affinities for BACE inhibitorsusing LRM scoring and energy-minimization sampling.

eir data for energy-minimization sampling, taken from Figure 5 in their paper, isFigure 5.8; hybrid Monte Carlo sampling gave a similar result. The correlation is

good, but the negatively charged octapeptides (the two data points with the mostculated and observed binding affinities) fall off the line. This is not surprising, as itly difficult in binding energy correlations to get ligands having different charge

l on a common line.

Reynolds also presented results for other simulation conditions that gave poorerxperimental data. We nevertheless regard their results as promising, and intend tor series of BACE inhibitors in future efforts to improve Liaison.

-14 -13 -12 -11 -10 -9 -8 -7 -6-14

-13

-12

-11

-10


Cal

c. G

bin

d (

kcal

r2 = 0.61RMSD = 1.10

Liaison User Manual

Getting Help

Schrdingenot installedit or ask you

For help instion, see the

Maestro haan online he

Checkavailacontai

If youBalloonot apwindo

For inin the

For otHelp fplayed

The Help mpages, the N

If you do nsources:

Maest Maest Maest Maest Impac Impac Strike Liaison 5.5 User Manual 45r software is distributed with documentation in PDF format. If the documentation isin $SCHRODINGER/docs on a computer that you have access to, you should installr system administrator to install it.

talling and setting up licenses for Schrdinger software and installing documenta-Installation Guide. For information on running jobs, see the Job Control Guide.

s automatic, context-sensitive help (Auto-Help and Balloon Help, or tooltips), andlp system. To get help, follow the steps below.

the Auto-Help text box, which is located at the foot of the main window. If help isble for the task you are performing, it is automatically displayed there. Auto-Helpns a single line of information. For more detailed information, use the online help.

want information about a GUI element, such as a button or option, there may ben Help for the item. Pause the cursor over the element. If the Balloon Help doespear, check that Show Balloon Help is selected in the Maestro menu of the mainw. If there is Balloon Help for the element, it appears within a few seconds.

formation about a panel or the tab that is displayed in a panel, click the Help buttonpanel, or press F1. The help topic is displayed in your browser.

her information in the online help, open the default help topic by choosing Onlinerom the Help menu on the main menu bar or by pressing CTRL+H. This topic is dis- in your browser. You can navigate to topics in the navigation bar.

enu also provides access to the manuals (including a full text search), the FAQew Features pages, and several other topics.

ot find the information you need in the Maestro help system, check the following

ro User Manual, for detailed information on using Maestroro Command Reference Manual, for information on Maestro commandsro Overview, for an overview of the main features of Maestroro Tutorial, for a tutorial introduction to basic Maestro featurest User Manual, fo

lia55 user manual

Documents