We develop a new, self-contained proof that the expected number of generations required for gene allele fixation or extinction in a population of size n is O(n) under general assumptions. The proof relies on a discrete Markov chain analysis. We further dev
Fisher-Wrightandcoalescenceproblems11choosearandomsetA,theprobabilityitseparatesA1fromA2isexactly50%.SincemorethanonehalfofthesetsAarestopped,theremustthereforebesomestoppedAthatseparatesA1fromA2.Thus,atleastoneofA1orA2hasdisappearedfromthepopulation.SincethisargumentappliestoanypairofallelesA1,A2,itfollowsthatthereisonlyonealleleleftinthepopulation.
Notethattheaboveproofdependsonlyonthefactthat,foranysubpopulation,theexpectedtimeforittoeitherbecometheentirepopulationortobeeliminatedisO(n)generations.ThusthepropertyofTheorem6isarobustphenomenon.
Theorem6issimilartoacoalescencetheorem.Theviewpointofcoalescenceisthatthecurrentgenerationistheendofanevolutionaryprocess,andoneconsiderswhichevolutionarysequencescouldhaveledtothecurrentgeneration.Thatis,unlikeFisher-Wright,whereoneconsiderstheevolutionoffuturegenerations,thecoalescenceviewpointconsidersthepossiblepastevolutionaryprocesses.Theexpectedcoalescencetimeisde nedtobetheexpectednumberofgenerationselapsedsinceallindividualsinthepresentgenerationhadacommonancestor.
Theusualassumptionforcoalescenceisthattheindividualsinagenerationchoosetheirparentsatrandomfromthepreviousgeneration,andasKingman[17]notes,thisismathematicallyequivalenttothebinomialprobabilities(1),Togeneralizethistootherevolutionaryprocesses,suchasthehypergeometricprocess,itisnecessarytode nethetime-reversalsoftheprocesses.Wedothisonlyforprocessesofthefollowingtype.De nition4.AMarkovprocessMonnindividualsiscontrolledbyfunctionproba-bilitiesprovidedthereisaprobabilitydistributionP(f)onthefunctionsf:[n]→[n],where[n]={1,2,...,n},suchthattheprocessMevolvesasfollows.Givengenera-tiontcontainingindividualsnumbered1,2,...,n,choosearandomfunctionfaccord-ingtothedistributionPand,foreachi,letthei-thindividualofgenerationt+1inheritthealleleofthef(i)-thindividualofgenerationt.
Asde ned,theseMarkovprocessesdi erfromourpreviousnotionofMarkovprocesssincetheindividualsarenumberedorindexed.ToreverttothepreviouskindofMarkovprocess,theindividualscouldberandomlypermutedineveryevolutionarystep.Equivalently,theprobabilitydistributiononfunctionscouldberequiredtobeinvariantunderpermutationsofthedomainandrangeofthefunction,thatis,itcould
热门文档