Prediction of Protein-Protein Interactions Using Protein Signature Profiling.pdfVIP

维普资讯 Article Prediction ofProtein—Protein InteractionsUsingProtein Signature Profiling M ahmoodA．MahdaviandYen—HanLin DepartmentofChemicalEngineeringUniversityofSaskatchewan，Saskatoon，SK s’N5A9，Canada Protein dom ainsare conserved and functionally independentstructuresthatplay an im portant role in interactionsam ong related proteins．Domain-dom ain inter- actionshavebeenrecentlyusedtopredictprotein-proteininteractions(PPI1．In general，theinteraction probability ofa pairofdomainsisscored using a trained scoring function．Satisfying athreshold，theprotein pairscarryingthosedomains are regarded as “interacting”． In this study．the signature contents ofproteins wereutilizedtopredictPPIpairsinSaccharomycescerevisiae，Caenovhabditisele- gans．and Homosapiens．Sim ilaritybetweenproteinsignaturepatternswasscored and PPIpredictionswere drawn based on the binary sim ilarity scoring function． Resultsshow thatthetrue positive rate ofprediction by theproposed approach isapproximately32％ higherthan thatusingthemaximum likelihoodestimation methodwhencompared withatestset．resultingin22％ increaseintheareaan- derthereceiveroperatingcharacteristic(ROC)curve．W henproteinscontaining oneortwosignatureswereremoved，thesensitivity ofthepredicted PPIpairsin- creased significantly．Thepredicted PPIpairsareon average 11tim esm orelikely to interactthan the random selection ata confidence levelof0．95．and on aver- age 4timesbetterthan those predicted by eitherphylogeneticprofiling or gene expression profiling． Key words：protein-protein interaction，protein signature，ROC curve Introduction Protein-proteininteraction(PPI)isthekeyelement utilized asinput to predictmore accurate interac— ofanybiologicalprocessina livingcel1．Proteinsin— tionsinanotherorganism