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-Wrightandcoalescenceproblems3manygenerations,whenstartingfrominitialallelefrequencyofp=i/N,whereH(p)isthebaseeentropyfunction.W.EwenshasshownthatthisisanupperboundforthemeanstoppingtimeforthediscreteFisher-Wrightproblemand,in[6],estimatedtheerrorinthedi usionequationapproximationwithalogarithmicadditiveterm.Thisarticlepresentsanewself-containedproofthattheexpectednumberofgenera-tionsrequiredforgeneallele xation/extinctionisO(nH(p)).Ofcourse,thisresultisaweakeningofthejust-discussedapproximationbythedi usionequation.However,thedi usionapproximationproofsarelongandarduous.Incontrast,ourproofsavoidthedi usionequationapproachaltogetherandworksolelywithdiscreteMarkovchains.OurproofmethodsapplytoawiderangeofMarkovmodels:ouronlyassump-tionsareameanconditionandavariationconditionasde nedinSection2.Forinstance,CookandRacko suggestedtousthefollowinghypergeometricmodelforreproduction.(Schensted[21]earlierstudiedthehypergeometricmodel.AgeneralizedhypergeometricmodelhasbeensuggestedbyM¨ohle[20].)Assumethatapopulationconsistsofdiscrete,non-overlappinggenerationsof xedsizen.The(i+1)stgenerationisobtainedfromthei-thgenerationasfollows:Eachparentindividualinthei-thgenerationhastwoo spring,yielding2npotentialo spring;arandomlychosensetofnofthesepotentialo springssurvivestobethe(i+1)stgeneration.Thus,eachparentindividualwillhavezero,oneortwoo springsurviveinthesucceedinggener-ation.Thehypergeometricmodelisbasedonsamplingfromasetofsize2nwithoutreplacement,whereastheFisher-Wrightmodelusessamplingwithreplacement.Forthehypergeometricmodel,ifagenerationcontainsiindividualswithalleleA,thentheprobabilitythatthenextgenerationcontainsjindividualswithalleleAisequalto
2i 2n 2i
qi,j=jnn j.(2)
Thehypergeometricprocesssatis esthemeanandvariationconditions,andthusourtheoremsimplythatithasexpectedabsorptiontimeO(nH(p)).
Sinceourapproachisdi erentfromthetraditionalapproach,itislikelyourresultscoversomecaseswhichcannotbecoveredbythedi usionequationapproach.Infact, weprovesimilarstoppingtimeresultsoftheformO(nforMarkovchainssatisfyingaweakerassumptiononthevarianceofthenumberofindividualswithagivenallele.Thisweakerassumptionisde nedinSection2astheweakvariation
热门文档