Estimation, variance and optimal sampling of gene diversity

It is shown that, at a given locus, there is a unique sample size per population which yields the smallest variance of GST, regardless of the number of populations studied, and the optimal sampling strategy for measuring GST is considered.

Abstract

An extension of Nei's analysis of diversity in a subdivided population is proposed for a haploid locus. The differentiation G STbecomes a natural extension of Wright's F STand generalizes Weir and Cockerham's parameter of co-ancestry by relaxing the assumption of identical correlation for all the alleles. Inter- and intrapopulation variances of the estimated diversities and differentiation are derived. Finally, the optimal sampling strategy for measuring G STwhen a fixed number of individuals can be analysed is considered. It is shown that, at a given locus, there is a unique sample size per population which yields the smallest variance of G ST,regardless of the number of populations studied. These theoretical developments are illustrated with an analysis of chloroplast DNA diversity in a forest tree. The results emphasize the necessity of sampling many populations, rather than many individuals per population, for an accurate measurement of the subdivision of gene diversity at a single locus.