MolecularDockingStudies
Jinxia(Nancy)Deng,PhD
Sept25,2007
Outline
ReviewofStructureBasedDrug
Design
Moleculardocking
Dockingcasestudy
Drugdiscoveryprocess
Identifydisease
DA
Formulation
Humanclinicaltrials
(210years)
Fi
le
N
Preclinicaltesting
(13years)
Fil
eI
ND
Isolateprotein
involvedin
disease(25years)
Findadrugeffective
againstdiseaseprotein
(25years)
Scaleup
FDAapproval
(23years)
http://wwwpersonal.engin.umich.edu/~wildd/mfg501/lecture.ppt
Databasesearching
3Dstructuresearching
pharmacophoremodels
exclusionzonestorepresentreceptor
Docking
predictionofthestructureandbindingfreeenergy
ofaligandreceptorcomplexgivenonlythe
structuresofthefreeligandandreceptor
examinebindingmodelofknownligandsto
suggestmodification
screendatabasesof3Dstructuretofindnovel
ligands
Itincludes
ligandproteindocking,onebranchof
rationaldrugdesigntopredictwhich
moleculescandisplaceothersfrom
thesurfaceofaprotein,orchange
proteinfunction.
Proteinproteininteraction
ProteinDNAinteraction
Liganddocking
(lock&keymodel)
Target
structure
Small
molecule
database
Structure
determination
Designnew
inhibitors
Ligandproteindocking:
Dockingofasmallmolecule(theligand)on
alargemolecule(theprotein,socalledreceptor)
Proteinproteindocking:
Usuallythedockingsiteisamore"planar"surfacethanin
theligandproteindocking
B
ProteinDNAsimulation*
NucleicAcidsResearch2006,Vol
34,No.1133173325
Evaluatingastructure
Xraycrystallography
Mostcommon
Musthavebetterthan2.5
resolution
NMR
Maybepossibletodetermine
dynamics
Homologymodeling
Whennoexperimentalstructure
isavailable
Usesaknownstructureasa
templatefortheunknown
GrowthofthePDB
30000
NumberofEntries
25000
20000
15000
10000
5000
0
1977
1982
1987
1992
Year
http://www.rcsb.org
1997
2002
Identificationoftargetsite
Ideallyapocketorprotuberancewith
Hydrogenbonddonorsandacceptors
Hydrophobiccharacteristics
Varietyofsizesofmolecularsurfaces
Canbeanactivesite,anassemblysite,
oracommunicationsite
Cocrystallizedproteinligandcomplexes
canbeinvaluablefordeterminingtarget
sites
Dockingalgorithms
Molecularflexibility
bothligandandproteinrigid
flexibleligandandrigidprotein
bothligandandproteinflexible
searchalgorithm
usetoexploreoptimalpositionsoftheligandwithin
theactivesite
scoringfunction
valueshouldcorrespondtopreferredbindingmode
efficiencyveryimportantfordatabasesearching
TODAYSFOCUS
eHITS:ElectronicHigh
ThroughputScreening
(SimBioSysInc.)
GOLD:GeneticOptimization
forLigandDocking (CCDC)
eHITS:ElectronicHighThroughput
Screening(SimBioSysInc.)
Fast,felxibible
dockingofwholeand
partialstructuresto
targetreceptors
Designedforlibrary
screening
eHiTSSearch
Ligandisdividedinto
rigidfragmentsand
connectingflexiblechains
Allrigidfragmentsare
dockedindependently
Graphmatching
Flexiblechainfitting
Localenergyminimisation
eHiTSScoring
Empiricalbasedscoring
Manycomponents;Hydrogen
bonding,HydrophobicityElectrostatic
potential,VanderWaalscontact
energy,Metalioninteractions,etc.
Allparametersareconfigurable
Chemicalpropertiesmappedto
Connollysurface
Flagcompatibilitymatrixscorefor
receptorligandcontacts
Therearemanydockingprograms
eHITSisafullysystematicandexhaustive
dockingprogram
19.FragmentBasedeHiTS
Algorithm
Ligandsaredividedintorigid
fragmentsandflexibleconnecting
chains
RigidDock:Eachfragmentis
dockedINDEPENDENTLY
everywhereinthereceptor
PoseMatch:Afastgraph
matchingalgorithmfindsall
matchingsolutionstoreconstruct
theoriginalmolecule
LocalEnergyOptimization:
structureisoptimizedwithinthe
receptor
Ranking:structuresareranked
basedonscoringfunction
HN
H2
HN
H2
H2
HN
H2
NH
HN
O
O
H2
HN
N
N
H2
H2
HN
HN
H2
Reconnecte
dLigand
Pose:
O
N
HN
N
H2
20.DockingFragmentsRigiDock
Rigidfragmentdockingbasedon
Chemicalfeaturemappedpolyhedra
Polyhedronshrinkwrappedonto
molecularsurface(Connolly)
Chemicalfeatureflagsonvertices
Analoguecavityrepresentation
Rapidmappingofligandandcavity
polyhedra
21.FragmentBasedDocking
22.DockingFragment1
TheFragmentisdocked
everywhereinthereceptor
site,andclusteredtogive
asetnumberofposes
NOTE: In the illustrations shown, only 250 poses are displayed,
default parameters of eHiTS uses 640 for each fragment
23.DockingFragment2
24.DockingFragment3
25.DockingFragment4
26.Closebutdifferent
Fragment2:EtherOxygen
Fragment4:AlcoholOxygen
PolarCharacter
27.Posematching
Eachrigidfragmentisnowindividually
dockedtoseveralalternativesites
Taskistofindsetsofposessuchthat:
fragmentscanbelinkedbyflexiblechains
theydonotbumpintoeachother
Algorithm
Graphnodes:ligandfragmentposes
Graphedges:posepairsatdistance
compatiblewithchainlength
Cliquesencodedockingsolutions
Frag1
Frag2
Frag3
Frag4
TopRankedPose
RMSD:0.699
Score:8.507
29.SummaryofeHiTSFeatures
Fragmentbaseddockingengine
Speedupduetodatabasestorage
Statisticallyderivedempiricalscoring
function
Usertrainablescoringfunction(Family
based)
Automatichandlingofprotonationstate
DHFRCaseStudy
ExperimentObjectives
Showabilitytoreproducecrystal
structures
ShowthateHiTScanselectactiveligands
ofhumanDHFRfromadrugdatabase
IllustratetheeaseofuseofeHiTS
Nopdbpreparation,noligand
preparation
ShoweHiTScanbeusedinHTS
DihydrofolateReductase
(DHFR)
Playsanessentialroleinthe
buildingofDNA
jugglestwomoleculesin
thisreaction
Folate(purple)andNADPH
(green)
Thefirstenzymetargetedfor
cancerchemotherapy
Oct.2002PDBMoleculeoftheMonth:
http://www.rcsb.org/pdb/molecules/pdb34_3.html
DHFRBindingsite
Thedrugmethotrexateisdesignedto
mimicfolate,blockingtheenzyme'saction
Notetheinteractionbetweenfolateand
NADPH,thisisessentialfortheenzyme's
function
ActivesSelection
SearchedforDHFRcomplexesinthePDB
Obtained88complexes,allsources
Uponquickvisualinspection,eliminated17
complexes
Containednoligandinthebindingsite
Containedmultipleligandsinthebindingsite
Selected71DHFRcomplexesforstudy
Including19humancomplexes
19HumanDHFRcomplex
Com plex
1dhf
1dlr
1dls
1drf
1hfp
1hfq
1hfr
1km s
1km v
1m vs
1m vt
1ohj
1ohk
1pd8
1pd9
1s3 u
1s3 v
1s3 w
2dhf
Ligand
Form ula
# at om s
FOL 2(C19H17 N7 O6)
49
MXA C17 H19N5O2
43
MTX C20H22N8O5
55
FOL C19H17 N7 O6
49
MOT C20H22N6O6
54
MOT C20H22N6O6
54
MOT C20H22N6O6
54
LIH
C18H17 N7
42
LII
C18H19N5O2
44
DTM C18H22N6O3
49
DTM C18H22N6O3
49
COP C27 H27 N9O6
69
COP C27 H27 N9O6
69
CO4 C19H24N6O3
52
CO4 C19H24N6O3
52
TQD C19H3 9N5O3
66
TQD C19H3 9N5O3
66
TQT C17 H3 3 N5
55
DZF 2(C20H18N6O6)
50
CoF
Form ula
C21H28N7 O17 P3
C21H28N7 O17 P3
C21H28N7 O17 P3
C21H3 0N7 O17 P3
C21H3 0N7 O17 P3
Validation
EachDHFRligandwasremovedfromthe
proteinanddockedbackintoitsbindingsite
EHiTSwasallowedtodothissplitautomatically
Resultswerethenjudgedbyevaluatingthe
RMSDbetweenthecrystalstructurebinding
positionandthecomputeddockingpose
Standard(default)parametersforeHiTSwere
usedinalltheruns
ValidationAllSources
71PDBs
29Unique
Ligands
Topranked Closest
< 0.5
3.23%
17.74%
< 1.0
22.58%
51.61%
< 1.5
59.68%
69.35%
< 2.0
67.74%
85.48%
< 2.5
83.87%
91.94%
< 3.0
88.71%
91.94%
Av eRMSD
1.94
1.41
TopRankedandClosestRMSD
Comparison
100.00%
90.00%
PercentofStructures
80.00%
70.00%
60.00%
Topranked
Closest
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
<0.5
<1.0
<1.5
<2.0
RMSD
<2.5
<3.0
ValidationHuman
19PDBs
12Unique
Ligands
Topranked Closest
< 0 .5
10 .53% 21.0 5%
< 1.0
21.0 5% 63.16%
< 1.5
68 .42% 73.68 %
< 2.0
73.68 % 8 4.21%
< 2.5
8 4.21% 94.74%
< 3.0
8 9.47% 94.74%
AveRMSD
1.98
1.18
TopRankedandClosestRMSD
ComparisonHumanDHFR
100.00%
90.00%
PercentofStructures
80.00%
70.00%
60.00%
Topranked
Closest
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
<0.5
<1.0
<1.5
<2.0
RMSD
<2.5
<3.0
ResultsTheGood
1dyiComplexxrayligandinwhite
Closest,105.77
0.76RMS
TopRank,139.6
0.85RMS
ResultsTheBad
1ly4Complexxrayligandinwhite
Closest,50.32
0.89RMS
TopRank,55.11
2.23RMS
ResultsTheUgly
1rc4Complexxrayligandinwhite
Closest,43.22
4.38RMS
TopRank,7.87
4.90RMS
ValidationSummary
EHiTSwasabletoreproduceaccurately
(RMS<2.0)thecrystalstructurepositionof
DHFRligands85%ofthetime
67%ofthetime,eHiTS'highestranking
(bestscoring)posehadaRMS<2.0
ThisnumberimprovesforHumanDHFR
ligands,74%
ThisshowsthateHiTSisabletopredict
dockingposesforDHFRligands
GOLD:GeneticOptimizationfor
LigandDocking( CCDC)
calculatingdockingmodesofsmall
moleculesintoproteinbindingsites
geneticalgorithmforproteinligand
docking
fullligandandpartialproteinflexibility
energyfunctionspartlybasedon
conformationalandnonbondedcontact
informationfromtheCSD
choiceofscoringfunctions:GoldScore,
ChemScoreandUserdefinedscore
virtuallibraryscreening
GOLDcontains:
Ascoringfunction:
GOLDFitness=Shbext+Svdwex+Shbint+Svdwint
Searchingmechanismtoexplorebinding
mode:Geneticalgorithm.
Dockingoftheligand
in1BL7,aMAP
kinaseP38complex.
TheGOLDdocked
solution(red)is
comparedwiththat
observedinthePDB
(colouredby
element).RMSD=
0.5Angstroms.
Dockingoftheligandin
1JAP,amatrix
metalloprotease
complex.TheGOLD
dockedsolution(red)is
comparedwiththat
observedinthePDB
(colouredbyelement)
RMSD=0.8Angstroms
GeneticAlgorithm:selection,crossover,andmutation
operatorstofindtheoptimalsolution
Solutionsareencodedasgenome,i.e.
chromsome
ChromosomeA:
101100101100101011100101
ChromosomeB:
111111100000110000011111
Reproduction:viacrossover,whichpicks
randomlyfromparentgenometoproduce
newgeneration.
Mutation:chromosomecanundergo
spontaneouschanges(withsmall
probability)toproducenewgenerationsto
samplingdifferentpartofthesolution
space.
Afitnessfunction(orobjectivefunction)is
usedtoevaluateeachsolution.
Dockingapplications
Bindingmodeprediction
Novelleaddesignoroptimization
Databasescreeningtoidentifyhits
Designnovelleadsfrom
fragment
Docking
fragment
target
Linkage
Perfectligand
Challenging:receptorflexibility
Ligandbindingtothereceptorinducesarangeof
conformationalchanges:
Twoproblems
Localrearrangementofsidechains
Hingemotionsofdomain
Predictingconformationalchange
samplingthepossiblechangedconformationduring
dockingbecausetoomanydegreeoffreedomavailableto
abindingsiteandnumberofbindingsiteconformations
growsexponentiallywiththem.
VirtualScreening/Docking
eHiTS
DOCK
FlexX
FlexE
SLIDE
Flo98
ADAM
Hammerhead
GLIDE
GRAMM
GOLD
PPD
HINT
BIGGER
LIGPLOT
SITUS
ZDOCK/
RDOCK
VEGA
RosettaDock
FRED
Affinity
FlexiDock
CombiBUILD
MCSAPCR
AUTODOCK
MCDOCK
ProDOCK
ICM
DockVision
3DDock
HEX
eHiTSDemo
(A)