Welcometothe2018D3Rworkshop!
WorkshoponChallengesinDockingandScreeningUSNationalInstitutesofHealth,2005
Participantsfrompharma,academia,USgovernmentAcademic:MikeGilson,ArtOlson,BrianShoichetGovernment:ChrisAustin,AnneChaka,JayneKapur,JannaWehrlePharma:JeffBlaney,WendyCornell,DebbieLoughney,CathyPeishoff,EmanuelePerola
ConclusionsComputationalpredictionsofposesandaffinitiesneedtoimproveDatasetsfrompharmacouldhelp
Workshopreport:http://bit.ly/2pLpy8C
EvaluationofProtein-LigandModelingMethods
Testshavebeenrunwithknowledgeoftheexperimentalresults
Differentmethodshavebeentestedfordifferentsystems
Valueofblinded,community-wide,predictionchallenges
NIH-fundedinitiativeCSAR2010-2014(Carlson,U.Michigan)D3R2014-present(Amaro&Gilson,UCSanDiego)
Pharmaaspotentialsourceofdatahighlyrelevanthithertounpublished
NIH-U01Resource,UniquePurposeblindedpredictionchallengestodriveadvancesinCADD
CentralGoal:Utilizepreviouslyunpublisheddatasetsasbenchmarksfordevelopersofprotein-ligandmodelingtechnologies
SynergywithPublicDatabases:Publicreleaseofmoreindustrialcrystalstructuresandaffinitydata
BroaderGoals: UtilizeblindeddatasetstodriveimprovementofallCADDtechnologiesandtofostereducationanddisseminationofmethods
MorepredictiveCADDmethodsbenefiteveryone!
DrugDesignDataResource(D3R)
D3RProjectTeam
MikeGilsonRommieAmaro Huanwang Yang
JasmineYoung
StephenBurley
MikeChiu
Conor Parks
JeffWagner JeffGretheChrisChuras
Zied Gaieb
CrystallographyGroup
Chenghua Shao
D3RScientificAdvisoryBoard
Aled EdwardsSGC
DavidMobleyUCIrvine
JohnMoultUMaryland
AdrianRoitbergUFlorida
Torsten SchwedeBiozentrum
CharlesGrimshawTakeda
MartinStahlRoche
UniqueHubforCADDCommunity
CoherentCADDdatasets
Blindedchallenges:Protein-ligand,modelsystems
Evaluationmetrics
Capturinganddisseminatingworkflows
Workshopsandnetworking
ChallengeTypes
GrandChallenges:ligand-proteinposesandaffinities
SAMPL:affinities,physicalpropertiesofsimplersystemswithDavidMobley,JohnChodera,&MichaelShirts
CELPP:automated,weeklyposepredictionchallenge
Stage1:Predictposesandaffinitiesofmultipleligandsforaprotein
Stage1b:Releaseco-crystalstructureswithoutligandstoenableself-docking(Isolatesevaluationofdockingalgorithm)
*Co-crystalstructureswithligandsreleased*
Stage2:Predictaffinitiesagain
Alldatareleased,depositedtoPDB,BindingDB
GrandChallenges
GrandChallenge201535participants,355submissions
HSP90:focusonpotencypredictionsDatafromAbbvieandCarlson’sCSARproject8cocrystalstructures(.6-2.0Åresolution)180IC50s(5nM-20µM)Threeseries:benzimidazolones,aminopyrimidines,benzophenone-likeVariedwater-mediatedinteractions;open/closedconformations
MAP4K4:focusonposepredictionsDatafromGenentech30cocrystalstructures(1.6– 2.5Åresolution)18IC50data(3.1nM - 10µM)DiversechemotypesbindinginATPsiteOpen/closedP-loopstructures
GrandChallenge249participants,262submissions
Farnesoid XReceptor(FXR):posesandpotenciesDatafromRoche36cocrystalstructures(resolutions<2.6Å)102IC50s(0.3nM-260µM)Threeseries+misc:sulfonamides,benzimidazoles,spirosHelixshiftsandvariedwater-bridges
GrandChallenge3
CathepsinSposes&IC50sJanssenPharmaceuticals
24cocrystalstructures,3.0Å136IC50s,3– 8500nM
27participants303submissions
KinaseKdsSGC-UNC/DiscoverX
Selectivity
VEGFR285(0.62to>104 nM)
JAK289(0.66to>104 nM)
p38-⍺72(0.28to>104nM)
ActivityCliffs
JAK217(53to>104 nM)
ActivityCliffs
TIE218(3.4to>104 nM)
Mutations
ABL112(49to>104nM)
11participants94submissions
6participants25submissions
6participants32submissions
6participants11submissions
SAMPLBlindedPredictionChallengesStarted2008
SmallmoleculehydrationfreeenergiesNicholls,Mobley,Guthrie,Chodera,Bayly,Cooper,Pande
Simplemodelsystems,e.g.,Host-guestbindingaffinitiesWater-organicphasepartitioncoefficientsSmallmoleculepKavaluesSmallmoleculehydrationfreeenergies
Advantagesvs.protein-ligandchallengesCalculationsfareasiertoconvergeTroubleshootingbyisolationofspecificissuesReducedambiguity(protonationstates,missingresidues…)
https://en.wikipedia.org/wiki/SAMPL_Challenge
SAMPL6Host-GuestandpKaChallengesHost-Guest:124submissionsfrom6groups
pKas:95submissionsfrom10groups
DeepCavityCavitandHostsBruceGibb,TulaneU.
8guestmoleculeswithbothOAandTEMOAhostvariants
Cucurbit[8]urilHostLyleIsaacs,U.Maryland
10guestmolecules
SmallMoleculepKasJohnChodera,SKMCC
25compounds
NewSAMPLing Challenge
Testsefficiencyofconformationalsamplingmethods
BindingfreeenergyconvergencewithNumberofenergyevaluationsWallclocktimeTotalCPUtime
Host-guestsystemswithspecificsetups
PosepredictionvariationsSoftwarepackages;.e.g.AutoDockVina,Glide,rDOCK,Gold,RosettaLigand,SurflexLigandoverlayoftenused;e.g.ROCS,PoPSSRelaxationandrescoring;e.g.Moleculardynamics,MMPB/SACombinations;e.g.Gold-PlantsPLP-rDock,RosettaLigand-Omega-ROCS,Surflex-Grim
AffinitypredictionandrankingLigand-basedStructure-based
“Lowresolution”dockingandscoring“Highresolution”freeenergymethods
Machine-learning
WideRangeofProtein-LigandMethods
Whatwehavelearned…
Accuracyofdockingandscoringcorrelatespoorlywithsoftware choice,andsuccessfulposepredictiondependsonothermethodologicalfactors; e.g.,
ligandandproteinpreparationchoiceofproteinconformationtreatmentofxtal waters.
Posepredictionbenefitsfromuseof known ligand-proteincocrystalstructures;e.g.,byligandoverlay
Humaninspectionandintervention donotconsistentlyimproveresults
Accuracyofposesusedcorrelatespoorlywithscoringaccuracy
Applicationoffreeenergymethodstohost-guestsystemspointstoneedforbetterforcefields
Explicitsolventfreeenergymethodshavenotyetoutperformedfasterscoringmethods
Rigorousevaluationofpredictionsisnon-trivialandcanbecontroversial
https://drugdesigndata.org/about/what-we-have-learned
FromOurGCParticipants…
Ithasmadememoreawareofthechallengesofsampling.I'vebeenworkingonbetterwaystoincludethisintoourprotocolsandmethods.
Iwouldpaymoreattentiontothereceptorconformationsandflexibility.
TheD3Rchallengesallowedustovalidateourdockingprotocol
DockingseemstobeimprovedbymachinelearningandIplantoincorporatesuchapproaches.
...itwilldefinitelychangetheIdodockingtoavoidorminimizefalsepositives.
Ithasmademepleasantlysurprisedwhenascoringfunctionactuallydeliversausefulresultandmakesmeveryskepticalofpeoplewhoblindlytrustthescorethattheyget.
SpecialIssuesinJCAMDthankstoTerryStouch,SeniorEditor-in-Chief
GC201514articles,2016
GC223articles,2017
SAMPL51/212articles,2016
SAMPL52/217articles,2017
Saturday
Sunday
Tuesday
Wednesday
PDB pre-releaseInChIsProtein sequencesForthcoming IDs
D3R releases InChIs and protein structures for docking
D3R opens for submissions
D3R submission window closesPDB releases structures
D3R evaluates predictions against released structures
D3R scriptsEliminate trivial ligandsPick protein structures
TowardGreaterStatisticalPowerContinuousEvaluationofLigandPosePredictions(CELPP)
CapturingComplexWorkflows
Full description of methods
Reproducibility
Evaluation on new datasets
Application to drug design projects
Method1OMEGA,SHAFTS,Amber11
Method2GLIDE-CCDC-GOLD,Amber14,MMGBSa
Method3WaterMap,SHAPEScreening,StructuralInteractionFingerprint,DFT/B3LYP/6-31G*,GLIDE-SP-XP,Induced-fit-docking,Emodel/GlideScore-SP,BindingPoseMetadynamics
https://drugdesigndata.org
WebPortalforData,Challenges,CommunityActivities
MealsLightbreakfasts:todayandtomorrowLunchtodayandtomorrowDinnertoday,here;onyourowntomorrow
ShuttlesBothmornings7:30amand7:45amThursdayevening:8:05pmand8:15pmFriday:3:30pm,3:45pmand5:30pm
Posters:Onwalls,pleaseusebluetapeprovided
ContactPeopleMeganMurphyIrisVillanuevaAnyonefromtheD3Rteam
Practicalities
D3RTeamatUCSDandRutgersD3RScientificAdvisoryBoardSAMPLco-organizers:Profs.D.Mobley,J.Chodera,M.ShirtsDr.TerryStouch andtheJCAMDteamDrs.PeterLyster,PeterPreusch,andJannaWehrle,NationalInstitutesofHealthDataContributors:JanssenPharma,SGC-UNC,ChoderaLab,GibbLab,IsaacsLab,othersExternalevaluators:Drs.Neysa Nevins,MillLambert,PatWaltersAllchallengeparticipants
Acknowledgements