VirtualScreeningofMultitargetKinaseInhibitors
Table2.MolecularDescriptorsUsedinThisWork
descriptorclass
simplemolecularproperties82
no.ofdescriptorsinclass
18
descriptors
articles
chemicalproperties83
molecularconnectivityandshape82,84335
electrotopologicalstate82,8542
no.ofC,N,O,P,S,no.oftotalatoms,no.ofrings,no.ofbonds,no.ofnon-Hbonds,MW,no.ofrotatablebonds,no.ofH-bonddonors,no.ofH-bondacceptors,no.of5-memberaromaticrings,no.of6-memberaromaticrings,no.ofNheterocyclicrings,no.ofOheterocyclicrings,no.ofSheterocyclicringsSandersonelectronegativity,molecularpolarizability,aLogp
Schultzmoleculartopologicalindex,Gutmanmoleculartopologicalindex,Wienerindex,Hararyindex,gravitationaltopologicalindex,molecularpathcountoflength1-6,totalpathcount,BalabanindexJ,0-2ndvalenceconnectivityindex,0-2ndorderdeltachiindex,Poglianiindex,0-2ndsolvationconnectivityindex,1-3rdorderKiershapeindex,1-3rdorderkappaalphashapeindex,Kiermolecular exibilityindex,topologicalradius,
graph-theoreticalshapecoef cient,eccentricity,centralization,Logpfromconnectivity
sumofestateofatomtypesCH3,dCH2,ssCH2,dsCH,aaCH,sssCH,dssC,aasC,aaaC,sssC,sNH3,sNH2,ssNH2,dNH,ssNH,aaNH,dsN,aaN,sssN,ddsN,aOH,sOH,ssO,sSH;sumofestateofallheavyatoms,allCatoms,allheteroatoms;sumofestateofH-bondacceptors;sumofHestateofatomtypeHsOH,HdNH,HsSH,HsNH2,HssNH,HaaNH,HtCH,HdCH2,HdsCH,HaaCH,HCsats,HCsatu,Havin;sumofHestateofH-bonddonors
ships,53quantitativestructure-activityrelationships54andVStools.13,17,21Weused981Dand2Ddescriptors(Table2)derivedfromoursoftware,55whichinclude18descriptorsintheclassofsimplemolecularproperties,3descriptorsintheclassofchemicalproperties,35descriptorsintheclassofmolecularconnectivityandshape,and42descriptorsintheclassofelectrotopologicalstate.
SupportVectorMachinesMethodology.SVMtraininganditsapplicationisschematicallyillustratedinFigure2.SVMisbasedonthestructuralriskminimizationprincipleofstatisticallearningtheory.56Itconsistentlyshowsout-standingclassi cationperformance,islesspenalizedbysampleredundancy,haslowerriskforover tting,iscapableofaccommodatinglargeandstructurallydiversetrainingandtestingdatasets,andisfastinperformingclassi cationtasks.57,58However,likeallmachinelearningmethods,theperformanceofSVMiscriticallydependentonthediversity
(53)Tong,W.;Xie,Q.;Hong,H.;Shi,L.;Fang,H.;Perkins,R.
Assessmentofpredictioncon denceanddomainextrapolationoftwostructure-activityrelationshipmodelsforpredictingestrogenreceptorbindingactivity.EnViron.HealthPerspect.2004,112(12),1249–1254.
(54)Ijjaali,I.;Petitet,F.;Dubus,E.;Barberan,O.;Michel,A.Assessing
potencyofc-JunN-terminalkinase3(JNK3)inhibitorsusing2DmoleculardescriptorsandbinaryQSARmethodology.Bioorg.Med.Chem.2007,15(12),4256–64.
(55)Xue,Y.;Yap,C.W.;Sun,L.Z.;Cao,Z.W.;Wang,J.F.;Chen,
Y.Z.PredictionofP-glycoproteinsubstratesbyasupportvectormachineapproach.J.Chem.Inf.Comput.Sci.2004,44(4),1497–505.
(56)Vapnik,V.N.Thenatureofstatisticallearningtheory;Springer:
NewYork,1995.
(57)Pochet,N.;DeSmet,F.;Suykens,J.A.;DeMoor,B.L.
Systematicbenchmarkingofmicroarraydataclassi cation:as-sessingtheroleofnon-linearityanddimensionalityreduction.Bioinformatics2004,20,3185–95.
oftrainingdatasets.OurearlierstudyshowedthattheyieldsofSVMtrainedbysparsedatasetscanbereducedby10%-20%ifthetrainingdatasetisreducedby50%,andthereductionisinacompounddiversitydependentmannerwithhigherreductionpercentageforhigherdiversitycom-pounds.13Becauseofthelimitedknowledgeofknowninhibitorsformanykinasetargets,suf cientlygoodSVMVStoolsmaynotbereadilydevelopedforthesetargets.Nonetheless,SVMVStoolswithcomparableperformancesorpartiallyimprovedperformancesincertainaspects(e.g.,reducedfalse-hitratesatcomparableinhibitoryield)areusefultocomplementotherVStools.Thedetailedmath-ematicalalgorithmsofSVMaredescribedinourearlierpublications13,21,55,59aswellasintheestablishedliter-ature.56-58Readersarereferredtothisliterature.OurSVMVSmodelsweredevelopedbyusingahardmarginc)100,000,andtheirσvaluesareintherangeof0.1-2.IntermsofthenumbersoftruepositivesTP(trueinhibitors),truenegativesTN(truenoninhibitors),falsepositivesFP(falseinhibitors),andfalsenegativesFN(falsenoninhibitors),theyieldandfalse-hitratearegivenbyTP/(TP+FN)andFP/(TP+FP)respectively.
ResultsandDiscussion
Dual-InhibitorsandNon-Dual-InhibitorsoftheStudiedKinasePairs.AsshowninTable1,thenumbersofdual-inhibitorsandnon-dual-inhibitorsofthekinasepairsare58,1,316and622forEGFR-PDGFR,71,1,303and392for
(58)Li,F.;Yang,Y.Analysisofrecursivegeneselectionapproaches
frommicroarraydata.Bioinformatics2005,21,3741–3747.(59)Ung,C.Y.;Li,H.;Yap,C.W.;Chen,Y.Z.Insilicoprediction
ofpregnaneXreceptoractivatorsbymachinelearningapproaches.Mol.Pharmacol.2007,71(1),158–68.
VOL.7,NO.5MOLECULARPHARMACEUTICS1551
Synthesis and Characterization of Pyrazoleanthrone Derivatives as Aurora A Kinase Inhibitors_生产/经营管理_经管营销_专业资料。Chem.Res.Chin.Univ.2013.29(6)...
Synthesis and Characterization of Pyrazoleanthrone Derivatives as Aurora A Kinase Inhibitors_生产/经营管理_经管营销_专业资料。Chem.Res.Chin.Univ.2013.29(6)...
热门文档