Statistical properties of the variation at linked microsatellite loci: implications for the history of human Y chromosomes [published erratum appears in Mol Biol Evol 1997 Mar;14(3):354]

It has recently been suggested that observed levels of variation atmicrosatellite loci can be used to infer patterns of selection in genomesand to assess demographic history. In order to evaluate the feasibility ofthese suggestions it is necessary to know something about how levels ofvariation at microsatellite loci are expected to fluctuate due simply tostochasticity in the processes of mutation and inheritance (geneticsampling). Here we use recently derived properties of the stepwise mutationmodel to place confidence intervals around the variance in repeat scorethat is expected at mutation-drift equilibrium and outline a statisticaltest for whether an observed value differs significantly from expectation.We also develop confidence intervals for the time course of the buildup ofvariation following a complete elimination of variation, such as might becaused by a selective sweep or an extreme population bottleneck. We applythese methods to the variation observed at human Y-specificmicrosatellites. Although a number of authors have suggested thepossibility of a very recent sweep, our analyses suggest that a sweep orextreme bottleneck is unlikely to have occurred anytime during the lastapproximately 74,000 years. To generate this result we use a recentlyestimated mutation rate for microsatellite loci of 5.6 x 10(-4) along withthe variation observed at autosomal microsatellite loci to estimate thehuman effective population size. This estimate is 18,000, implying aneffective number of 4,500 Y chromosomes. One important general conclusionto emerge from this study is that in order to reject mutation-driftequilibrium at a set of linked microsatellite loci it is necessary to havean unreasonably large number of loci unless the observed variance is farbelow that expected at mutation-drift equilibrium.