LiganddockingandbindingsiteanalysiswithPyMOLandAutodock/Vina
DanielSeeliger
·BertL.deGroot
Received:22January2010/Accepted:26March2010/Publishedonline:17April2010©TheAuthor(s)2010.ThisarticleispublishedwithopenaccessatSpringerlink.com
AbstractDockingofsmallmoleculecompoundsintothebindingsiteofareceptorandestimatingthebindingaffinityofthecomplexisanimportantpartofthestructure-baseddrugdesignprocess.Forathoroughunderstandingofthestructuralprinciplesthatdeterminethestrengthofaprotein/ligandcomplexboth,anaccurateandfastdockingprotocolandtheabilitytovisualizebindinggeometriesandinteractionsaremandatory.HerewepresentaninterfacebetweenthepopularmoleculargraphicssystemPyMOLandthemoleculardockingsuitesAutodockandVinaanddemonstratehowthecombinationofdockingandvisuali-zationcanaidstructure-baseddrugdesignefforts.KeywordsDocking·Virtualscreening·Autodock·Vina·PyMOLIntroduction
Virtualscreeningofcompoundlibrarieshasbecomeastandardtechnologyinmoderndrugdiscoverypipelines[1].Ifasuitablestructureofthetargetisavailablemoleculardockingcanbeusedtodiscriminatebetweenputativebindersandnon-bindersinlargedatabasesofchemicalsandtoreducethenumberofcompoundstobe
ThisworkwasfundedbytheDeutscheForschungsgemeinschaft(DFG)grantNo.GR207914
D.Seeliger(&)·B.L.deGroot
ComputationalBiomolecularDynamicsGroup,Max-Planck-InstituteforBiophysicalChemistry,AmFassberg11,37077Gottingen,Germanye-mail:dseelig@gwdg.deB.L.deGroot
e-mail:bgroot@gwdg.de
subjectedtoexperimentaltestingsubstantially.Visualexaminationofpredictedbindinggeometries(dockingposes)therebycontributescruciallytothefurtherdevelop-mentofaleadcompoundeithertowardsenhancedbindingaffinity,towardsreducedsideeffectsortowardsreducedsusceptibilitytodrugresistancerelatedmutations.OverthelastyearsthePyMOLmoleculargraphicssystem[2]hasevolvedfrombeingapowerfulmolecularviewerwithexceptional3D-capabilitiesintoaplatformforseveralpro-gramsandapplicationswhichmakeuseofPyMOL’sversatilevisualizationproperties.
Throughitsmulti-layerarchitectureandtheuseofthepowerfulobject-orientedscriptinglanguagePythonatthetop-level,PyMOLisrelativelyeasytoextendandcus-tomizewithoutre-compilingthesourcecode.Extensionscaneithermakeuseofthewizard-interfaceortheplugin-interface,thelatterofwhichisthemorecommonlyuti-lized.Inthefieldofmolecularinteractionstherehavebeenseveral(plugin)-extensionsdevelopedthatgaingreatpop-ularity.TheAPBSplugin[3]isaninterfacetothepopularadaptivePoisson-Boltzmannsolver(APBS[4])programandprovideseasyaccesstoelectrostaticscalculationsandthevisualizationofpotentialenergysurfacesandchargedensitiesonproteinsurfaces.CAVER[5,6]performscalculationsofsubstratepathwaysandentrancetunnelsinproteinstructureswhicharevisualizedinPyMOL.CASTp[7,8,9]detectspocketsandvoidsinproteinstructurestodetermineandcharacterizebindingsites,andeMovie[10]providesanumberoffunctionalitiestocreateanimationsandmovies.
InthepresentworkwedescribeapluginforPyMOLwhichallowstocarryoutmoleculardocking,virtualscreeningandbindingsiteanalysiswithPyMOL.ThepluginrepresentsaninterfacebetweenPyMOLandtwopopulardockingprograms,Autodock[11,12]and
123
418JComputAidedMolDes(2010)24:417–422
AutodockVina[13]andmakesextensiveuseofaPythonscriptcollection(AutodockTools[14])forthesetupofdockingruns.Sincevisualizationiscrucialforstructure-baseddrugdesign,severaltoolshavebeendevelopedtoaddvisualsupportfortheautodocksuite.ThevisualizerAutoDockToolsoffersacompletemolecularviewerandagraphicalsupportforallstepsrequiredforsetupandanalysisofdockingruns.Raccoon(http://autodock.scripps.edu/resources/raccoon),BDT[15]andDOVIS[16]areothergraphicaluserinterfacesforAutodockwithaspecialfocusonlarge-scalevirtualscreening.RaccoonandBDTfocusonastraightforwarddataorganizationimportantforvirtualscreeningbutdonotprovidemolecularviewingfunctionalitywhereasDOVISusesanembeddedJavaviewer.ThePyMOLplugindescribedhereisdevelopedspecificallytomakeuseofPyMOL’sexceptionalmolec-ularviewingcapabilities.PyMOListhemostfrequentlyusedprogramforgeneratingpublicationqualitypicturesofmolecularstructuresandoffersmultipleadvancedrender-ingoptions.Additionallyitprovidesexceptional3D-viewingfunctionalitieswhichcanbeveryusefulinstruc-ture-baseddrugdesign.SincePyMOLsupportsseveralcommonlyusedfileformatsforelectrondensitymapsitisalsothepreferredtoolforcrystallographers.Hence,aneasytohandleAutodock/Vina-pluginforPyMOLisexpectedtolowerthebarrierforscientistwhoarenotdockingexpertstomakeuseofthesepopulardockingprotocolswithintheirpreferredenvironmentandtouseitinconjunctionwithotherapplicationsavailableforPyMOL.
ThepluginprovidesfunctionalitytocarryouttheentireworkflowofadockingstudywithvisualsupportofPy-MOLandagraphicaluserinterface.Inthecurrentversiontheplugincoversthefollowingoperations:Bindingsitedefinitionandadjustment,automaticfilepreparationsforreceptordefinition,straightforwardselectionofflexibleresidues,ligandfilepreparation,generationandviewingof
Fig.1Definitionofadockingboxaroundareferenceligand.Position,sizeandvisualizationpropertiescanbeadjustedwiththeplugin
Fig.2Selectionofsidechainswithinthebindingsiteforthesetupofdockingrunswithflexiblesidechains
affinitygridmaps,viewingofdockingposes,andanalysisandexportofvirtualscreeningresults.
Adockingstudyusuallystartswiththedefinitionofabindingsite,ingeneralarestrictedregionoftheprotein.ThesizeandlocationofthisbindingsiteisvisualizedinPyMOLandcanbeadjustedinteractively.Optionallyres-idueswithinthebindingsitecanbedefinedtobeflexibleduringdocking.Subsequently,thenecessaryfilesforthereceptordefinitionaregeneratedautomatically.Similarly,filepreparationsformultipleligandscanbecontrolledvia
123
JComputAidedMolDes(2010)24:417–422theplugin.TheactualdockingcalculationscanbebelaunchedfromwithinPyMOLandtheresultsbevisualized.Furthermore,theresultsofmultipledockingrunsareautomaticallyanalyzedandarankedlistofthedockedposesisgeneratedanditcanbeexportedindifferentdataformatsforfurtheranalysis.Methods
Bindingsitedefinition
BothAutodockandVinauserectangularboxesforthedefinitionofthebindingsite.Intheplugin,theboxcentercanbydefinedeitherbyprovidingexplicitcoordinatesor,moreuserfriendly,bydefiningaPyMOLselection(e.g.areferenceligand).TheboxcenteristhencalculatedfromthemeancoordinatesoftheatomsfromthePyMOLselectionandthedockingboxdisplayedinthePyMOLwindow.Thesizeandtheexactpositionoftheboxcanalsobeadjustedtotheuser’sdemands.Forvisualizationpurposesthepluginfurthermoreallowstochosebetweentwodisplayoptionsandthecoloroftheboxframe(seeFig.1).
BindingsitedefinitionsdefinedherecanalsobeexportedtoinputfilesforeitherAutodockorVina.Setupandexecutionofdockingruns
AutodockandVinaneedreceptorandligandrepresentationsinafileformatcalledpdbqtwhichisamodifiedproteindatabank[17]formatcontainingatomiccharges,atomtypedefinitionsand,forligands,topologicalinformation(rotat-ablebonds).ThesefilepreparationsarecarriedoutbythepluginusingscriptsfromtheAutodockToolspackage.LigandsforsubsequentdockingrunscaneitherbepreparedonebyonethroughPyMOLselectionsorbyspecifyingadirectorycontainingalibraryofligandstobedocked.
Afterbindingsitedefinitionandreceptorandligandpreparation,dockingrunscanbedirectlylaunchedfromPyMOL.Alternatively,runinputfilescanbewrittentostartthedockingrunsfromthecommandline.BothAutodockandVinaallowforflexibilityofpredefinedsidechainsduringdocking.Herethepluginfacilitatestheselectionofflexiblesidechains.SidechainswithinthedockingboxcanbevisualizedstraightforwardlyandPyMOLselectionscanbetranslatedintoaflexiblereceptordefinition(Fig.2).
Bindingsiteanalysiswithinteractionmaps
Autodockusesinteractionmapsfordocking.Priortotheactualdockingrunthesemapsarecalculatedbythe
419
Fig.3Autodockgridmapsdisplayedwithdifferentcontourlevels.aMapforinteractionsofaliphaticcarbonatomsatcontourlevel5kcal/mol.bSamemapatcontourlevel−0.3kcal/mol.cHydrogenbonddonormapatcontourlevel−0.5kcal/mol
123
420JComputAidedMolDes(2010)24:417–422
programautogrid.Foreachligandatomtype,theinterac-tionenergybetweentheligandatomandthereceptoriscalculatedfortheentirebindingsitewhichisdiscretizedthroughagrid.Thishastheadvantagethatinteractionenergiesdonothavetobecalculatedateachstepofthedockingprocessbutonlylookedupintherespectivegridmap.Inadditiontospeedingupadockingrunsthegridmapsontheirowncanalsoprovidevaluehintsforligandoptimization.Sinceagridmaprepresentstheinteractionenergyasafunctionofthecoordinatestheirvisualinspectionmayrevealpotentialunsaturatedhydrogenacceptorsordonorsorunfavourableoverlapsbetweentheligandandthereceptor.ThepluginthereforeprovidesthefunctionalitytovisualizethesegridmapsinPyMOL.ThemapsgeneratedbyautogridareconvertedtoafileformatreadablebyPyMOL(DXformat)whichallowstodrawisosurfacesandisomeshesanalogoustoelectrondensitymaps.Sinceseveralmapscanbeloadedandcontrolledsimultaneously,arapidinspectionofseveralinteractiontypesismadeveryeasily.Figure3showshowthesegridmapscanbecontrolledviatheplugin.
InFig.3Aanisosurfaceatacontourlevelof5kcal/molfortheinteractionoftheproteinwithaliphaticcarbonatomsisshown.Suchasettingmaybeusedtogetavisualimpressionoftheoverallshapeofthebindingsite.Ligandmodificationswhichcauseapenetrationofsuchawallwillmostlikelynotenhancetheaffinity.InFig.3Bthesamemapisvisualizedatacontourlevelof−0.3kcal/mol.Ascanbeseen,theshapeofthesurface,hereshownasiso-mesh,roughlydescribesanenvelopeoftheligandandrevealsputativespotsofattractiveinteractionsthatmayguidefurtherligandoptimization.Likewise,hydrogenbonddonororacceptorinteractionmapscanguideligand
optimizationsincetheymightrevealunsaturatedacceptorordonorpositions(Fig.3C).
Thepluginprovidesfunctionalitytohandledifferentinteractionmapsandrepresentationsatdifferentcontourlevelsatthesametimeandhence,offersthepossibilitytovisualizedifferentbindingsitepropertieswhichmaypro-videvaluableinsightsforstructure-baseddrugdesign.
Analysisofdockingresults
DockingposesgeneratedbythedockingprogramscanbedirectlyloadedintoPyMOLthroughtheplugin.Posesformultipleligandsmaybehandledsimultaneouslyusinganintuitivenotebooklayout(seeFig.4).Foreachdockingpose,metainformationcontainingthedockingscoreisdisplayedinasmalltextviewer,allowingdirectanalysisofconfiguration/scorerelationships.Moreover,resultsfrommultipledockingrunsaresummarizedinatable(seeFig.5).Thedockingposesarerankedaccordingtotheirdockingscoresandboththerankedlistofdockedligandsandtheircorrespondingbindingposesmaybeexported.Forinstance,therankedlistofdockingresultscanbeexportedinaCSVfileformatwhichcanbedirectlyimportedintoprogramslikeExcel.
Conclusion
WepresentanovelpluginforthepopularmoleculargraphicssystemPyMOLwhichallowstoperformdockingstudiesusingAutodockorAutodock/Vina.Theplugincoversallfunctionalitiesfortheentireworkflowofa
Fig.4Analysisofdockingposes.Left:PyMOLviewerwithdisplayeddockingposes.Right:Poseviewerpageoftheplugin.Posesfrommultipledockingrunsmaybeanalyzedsimultaneouslyusinganintuitivenotebooklayout
123
JComputAidedMolDes(2010)24:417–422Fig.5Virtualscreening
analysis.Arankedlistofdockedligandsisgenerated
automaticallyanditcanbeexportedindifferentdata
formats.Additionally,dockingposesfromdifferentligandscanbeexportedinasinglefile
421
dockingrunplusadditionalfunctionalitytoprepare,exe-cuteandanalyzevirtualscreeningtasks.Sincevisualsupportisanimportantaspectofstructure-baseddrugdesign,thepluginisexpectedtoenhancetheseeffortsbyallowingthecombineduseoftwowidelyuseddockingprogramsandPyMOL.Thepluginisavailablefreeofchargewithsourcecodeandmaybeobtainedfromhttp://wwwuser.gwdg.de/~dseelig/adplugin..
installationandoperationondifferentplatforms(MacOSandWindows)nosupportisprovidedfortheseoperatingsystems.
OpenAccessThisarticleisdistributedunderthetermsoftheCreativeCommonsAttributionNoncommercialLicensewhichper-mitsanynoncommercialuse,distribution,andreproductioninanymedium,providedtheoriginalauthor(s)andsourcearecredited.
References
Systemrequirements
ThepluginhasbeendevelopedonLinuxwithPyMOLversion1.2andrequiresMGLToolsversion1.5.4andNumPyversion1.3.Althoughusersreportedsuccessful
1.KitchenD,DecornezH,FurrJ,BajorathJ(2004)Dockingandscoringinvirtualscreeningfordrugdiscovery:methodsandapplications.NatRevDrugDiscov3(11):935–949
2.DeLanoWL(2002)ThePyMOLmoleculargraphicssystem.http://www.pymol.org
123
422
3.LernerMG,CarlsonHA(2008)Apbspluginforpymol.Uni-versityofMichigan,AnnArbor
4.BakerN,SeptD,JosephS,HolstM,McCammonJ(2001)Electrostaticsofnanosystems:applicationtomicrotubulesandtheribosome.ProcNatlAcadSciUSA98(18):10,0375.Petr
ˇekM,OtyepkaM,Bana´sˇP,Kosˇinova´P,KocˇaJ,Damborsky´J(2006)CAVER:anewtooltoexploreroutesfromproteinclefts,pocketsandcavities.BMCBioinformatics7(1):316
6.Damborsky
`J,PetrekM,Bana´sˇP,OtyepkaM(2007)Identifica-tionoftunnelsinproteins,nucleicacids,inorganicmaterialsandmolecularensembles.BiotechnolJ2:62–67
7.LiangJ,EdelsbrunnerH,WoodwardC(1998)Anatomyofpro-teinpocketsandcavities:measurementofbindingsitegeometryandimplicationsforliganddesign.ProteinSci7:1884–18978.LiangJ,EdelsbrunnerH,FuP,SudhakarP,SubramaniamS(1998)AnalyticalshapecomputingofmacromoleculesII:iden-tificationandcomputationofinaccessiblecavitiesinsideproteins.Proteins:StructFunctGenet33:18–29
9.LiangJ,EdelsbrunnerH,FuP,SudhakarP,SubramaniamS(1998)Analyticalshapecomputationofmacromolecules:I.Molecularareaandvolumethroughalphashape.Proteins:StructFunctGenet33(1):1–17
10.HodisE,SchreiberG,RotherK,SussmanJ(2007)eMovie:a
storyboard-basedtoolformakingmolecularmovies.TrendsBiochemSci32(5):199–204
123
JComputAidedMolDes(2010)24:417–422
11.MorrisGM,GoodsellDS,HallidayDS,HueyR,HartWE,Belew
R,OlsonAJ(1998)Automateddockingusingalamarckiangeneticalgorithmandandempiricalbindingfreeenergyfunction.JCompChem19:1639–1662
12.HueyR,MorrisGM,OlsonAJ,GoodsellDS(2007)Asemi-empiricalfreeenergyforcefieldwithcharge-baseddesolvation.JCompChem28:1145–1152
13.TrottO,OlsonA(2010)AutoDockVina:improvingthespeed
andaccuracyofdockingwithanewscoringfunction,efficientoptimization,andmultithreading.JCompChem31:455–46114.SannerM(1999)Python:aprogramminglanguageforsoftware
integrationanddevelopment.JMolGraphicsMod17:57–6115.VaqueM,ArolaA,AliagasC,PujadasG(2006)BDT:aneasy-to-usefront-endapplicationforautomationofmassivedockingtasksandcomplexdockingstrategieswithAutoDock.Bioinfor-matics22(14):1803
16.ZhangS,KumarK,JiangX,WallqvistA,ReifmanJ(2008)
DOVIS:animplementationforhigh-throughputvirtualscreeningusingAutoDock.BmcBioinformatics9(1):126
17.BermanH,WestbrookJ,FengZ,GillilandG,BhatT,WeissigH,
ShindyalovI,BourneP(2000)Theproteindatabank.URLciteseer.ist.psu.edu/berman02protein.html
因篇幅问题不能全部显示，请点此查看更多更全内容