Assessing Protein Co-evolution in the Context of the Tree of Life Assists in the Prediction of the I

蛋白质互作用问题研究的论文

doi:10.1016/j.jmb.2005.07.005J.Mol.Biol.(2005)352,1002–1015

AssessingProteinCo-evolutionintheContextoftheTreeofLifeAssistsinthePredictionoftheInteractome

FlorencioPazos1* ,JuanA.G.Ranea2,DavidJuan3andMichaelJ.E.Sternberg1

StructuralBioinformaticsGroup,DivisionofMolecularBiosciences,ImperialCollegeLondon,LondonSW72AZUK

BiomolecularStructureandModellingUnit,DepartmentofBiochemistryandMolecularBiology,UniversityCollegeLondon,LondonWC1E6BTUK

ProteinDesignGroup

NationalCentreforBiotechnol-ogy(CNB-CSIC),MadridSpain

321

Theidenti cationofthewholesetofproteininteractionstakingplaceinanorganismisoneofthemaintasksingenomics,proteomicsandsystemsbiology.Oneofthecomputationaltechniquesusedbymanyinvestigatorsforstudyingandpredictingproteininteractionsisthecomparisonofevolutionaryhistories(phylogenetictrees),underthehypothesisthatinteractingproteinswouldbesubjecttoasimilarevolutionarypressureresultinginasimilartopologyofthecorrespondingtrees.Here,wepresentanewapproachtopredictproteininteractionsfromphylogenetictrees,whichincorporatesinformationontheoverallevolutionaryhistoriesofthespecies(i.e.thecanonical“treeoflife”)inordertocorrectbytheexpectedbackgroundsimilarityduetotheunderlyingspeciationevents.WetestthenewapproachinthelargestsetofannotatedinteractingproteinsforEscherichiacoli.Thisassessmentofco-evolutioninthecontextofthetreeoflifeleadstoahighlysigni cantimprovement(P(N)bysigntestw10E–6)inpredictinginteractionpartnerswithrespecttotheprevioustechnique,whichdoesnotincorporateinformationontheoverallspeciationtree.Forhalfoftheproteinswefoundarealinteractoramongthe6.4%topscores,comparedwiththe16.5%bythepreviousmethod.WeappliedthenewmethodtothewholeE.coliproteomeandproposefunctionsforsomehypotheticalproteinsbasedontheirpredictedinteractors.Thenewapproachallowsusalsotodetectnon-canonicalevolutionaryevents,inparticularhorizontalgenetransfers.Wealsoshowthattakingintoaccountthesenon-canonicalevolutionaryeventswhenassessingthesimilaritybetweenevolutionarytreesimprovestheperformanceofthemethodpredictinginteractions.

q2005ElsevierLtd.Allrightsreserved.

*Correspondingauthor

Keywords:proteininteractions;phylogenetictree;co-evolution;interac-tome;horizontalgenetransfer

Introduction

Thenetworkofprotein–proteininteractionsofagivenorganism(“interactome”)containsextensiveinformationaboutitsbiology,sinceproteininter-actionsareinvolvedinmostcellularprocesses(structuralmacromolecularcomplexes,signalling

Presentaddress:F.Pazos,ProteinDesignGroup,

NationalCenterforBiotechnology(CNB-CSIC),CampusUAM,Cantoblanco,28049Madrid,Spain.

Abbreviationsused:HGT,horizontalgenetransfer;16SrRNA,ribonucleicacidmoleculeinthesmallsubunitoftheribosome;ORF,openreadingframe;GO,geneontology.

E-mailaddressofthecorrespondingauthor:pazos@cnb.uam.escascades,metabolism,transcriptioncontrol,etc.).Experimentaltechniquesforthedeterminationofproteininteraction,mainlyvariationsoftheyeasttwo-hybridsystem1andaf nitypuri cation,2,3havebeenappliedinahigh-throughputapproachaimingtocoverasmuchaspossibleoftheinteractomeofanumberofmodelorganisms.4–7Knowledgeofthecompleteinteractomeallowedsomeofthe rststudiesofbiologicalnetworksfromasystemsbiologypointofview,extractingimport-antdataonthetopology,connectivityandevolutionoftheproteininteractionnetwork.8–14

Severalcomputationaltechniqueshaveemergedtocomplementtheseexperimentalapproaches.Thesecomputationaltechniquescanbeusedtoguideexperimentsrestrictingthenumberofpairstotestexperimentally,15to lterasetofexperimental

0022-2836/$-seefrontmatterq2005ElsevierLtd.Allrights

reserved.

第1页下一页