Transcriptional Regulatory Networks in S. cerevisi

TranscriptionalRegulatoryNetworksinSaccharomycescerevisiaeLaurensKraal∗3130967CancerGenomics&DevelopmentalBiology(UU)MasterThesisSupervisor:ProfessorDr.FrankHolstegeAugust10,2009AbstractOneofthemaingoalsofsystemsbiologyisreverseengineeringnetworkstructuresfromexperimentaldata.Usingquantitativenetworksmeasures,biologicalprocessescanbecharacterizedandunderstoodonasystemslevel.Thispaperreviewsthescien-tificliteratureonthetranscriptionalregulatorynetworksofSaccharomycescerevisiaetodiscusswhatkindofstructurescanbeidentifiedandwhatthismeansbiologically.Thisdiscussioncomprisesnetworkcharacteristics,networkdynamicsinbiologicalprocesses,androbustness;aninherentemergentpropertyofnetworks.1IntroductionSystemsBiologyTraditionally,biologyisareductionistsci-ence.Overtheyearsithassuccessfullyidentifiedmanycomponentsandinterac-tionsofvariousnaturebyreducingthemtoelementaryunitsthatcanbeisolatedandstudiedindependently[1].However,therearemanypropertiesoflivingsystemsthathavenohopeofbeingunderstoodusingthismethod.Asystemisaholisticentitywhose∗31LeroySt,Dorchester,MA02122,USA+1-617-606-1949,[e-mail]actionsbothwitheachother,aswiththeoutsideworld.Itsbehaviorandorganizationemergesoutofthiscomplexinterplay.Complexsystemslikethecell,comprisenumerousfunction-allyandstructurallydiversecomponentsthatinteractinnetworkstogeneratecoher-entbehaviors[2].Thebehaviorisafunctionofnetworkpropertiesandtheelementsin-volved.Althoughmolecularbiologyhasun-coveredmanyfactsaboutnetworkcompo-nents,biologicalsystemscannotbeunder-stoodonthislevel.Asystem-levelperspec-tiveisnecessarytounderstandthefunctionandbehaviorsofbiologicalsystemssuch1asorganelles,cells,tissues,organisms,andecologies.Systemsbiologyistheinterdisci-plinaryfieldthattriestogenerateandan-alyzesystem-leveldataandtriestounder-standbiologicalmechanismsinthecontextoftheirsystem.Itaimstodevelopquantita-tivemathematicalmodelstogainaformalunderstandingofbiologicalprocesses[3].Asbiologicalsystemscompriseintricatenetworks,graphtheoryisemployedtoan-alyzeandquantifythem.Fortranscrip-tionalregulatorynetworks,thisinvolvesabstractinginteractionsfromexperimentaldatageneratedbymicroarrayandChIP-chiptodirectedgraphswithtranscriptionfactorsandtargetgenesasnodesandtheirinteractionsaslinksbetweenthem(seefig-ure1).Interactionsbetweengenescanbeidentifiedfromexperimentaldatausingreverse-engineeringmethods[3].Thispa-perdealswiththetranscriptionalregula-tionnetworksofthebaker’syeastSaccha-romycescerevisiaeasmuchworkonillumi-natingthenetworkorganizationhasbeendoneonthiseukaryoticorganism.OutlineThefirstsectionofthispaperdiscussesthemeasuresthatcanobtainedfromnetworkarchitecturesandtopology.Thenextsec-tiondealswiththeintegrationofthesemea-sureswithadynamicalanalysisofbiolog-icalprocesses.Thesectionafterthishasrobustness,anemergentpropertyinherenttocomplexnetworks,asitssubject.Thepaperendswithadiscussion.GraphTranscriptionnetworkgenexgeneyXYXYFigure1:Atranscriptionalrelationshipin-volvesatranscriptionfactorbindingtoaregulatorysequenceofagenetoregulateitsexpression.Thisrelationshipisrepresentedinagraphbyanodetargetinganothernode(figureadjustedfrom[4]).2NetworkMeasuresThestudyofcomplexnetworkstookofmorethan50yearsago,withthetheworkdonebytheHungarianmathematiciansPaulErd¨osandAlfr´edR¨enyi[5].Theysuggestedthatcomplexnetworkscouldbemodeledasregularobjects,suchasasquareoradiamondlattice,orasarandomnet-workwithrandomlinksconnectingthenodesofthenetwork.Modelingrandomnetworksenabledthestudyofthestructureandpropertiesofcomplexnetworks.Oneofthemostelementarycharacteristicsofanodeisitsdegree(orconnectivity).Thedegreecomponentkisameasureforhowmanylinksthenodehastoothernodesanditdeterminesmanypropertiesofthesys-tem.Oneofthepredictionsofrandomnet-worktheoryisthatsuchasystemisdemo-craticoruniform.Asthelinksconnectingthenodeswillbealmostcompletelyevenly2distributed,thenodesfollowabellshapedPoissondistribution.Nodesthathavesig-nificantlymoreorfewerlinkstoothernodesthantheaveragedegree,k(de-notestheaverage),arehardtofindinran-domnetworks.Randomnetworksarealsocalledexponential,astheprobabilitythatanodehasconnectionstokothernodesde-creasesexponentiallyforalargek.[6]However,intheabsenceofdataonlargenetworks,thepredictionsofrandomnet-worktheorywererarelytestedintherealworld.Inthisageofcomputerizeddataacquisition,thetopologicalinformationofmanyreal-worldnetworkshavebecomeavailableandithasbeenfoundthatthetopologicalpropertiesofrealnetworks,suchasbiologicalnetworks,cannotbeexplainedbyrandomnetworkmodels[7].Instead,mostrealnetworksare’scale-free’,whichmeanstheyarecategorizedbyadegreedis-tributionthatapproximatesapower-law(P(k)kγwhereindicates’proportionalto’).ThedegreedistributionischaracterizedbythetheprobabilityP(k)thatanodeinthenetworkinteractswithkothernodes[8].Thedegreedistributionisessentialindif-ferentiatingbetweennetworks.Thepeakeddegreedistributionofarandomnetworkin-dicatesthatthesystemhasnohighlycon-nectednodesandmostnodeshaveatypicaldegree(Seefigure2)[8].Inapowerlawdis-tribution,mostnodeshavejustafewcon-nec