A comparison of several measures of genetic distance and population structure with microsatellite data: Bias and sampling variance

Because of their rapid mutation rate and resulting large number of alleles, microsatellite DNA are well suited to examine the genetic or demographic structure of fish populations. However, the large number of alleles imply that large sample sizes are required for accurate reflection of genotypic frequencies. Estimates of genetic distance are often biased at small sample sizes, and biases and sampling variances can be affected by the number of, and distances between, alleles. Using data from a large collection of larval cod (Gadus morhua) from a single area, I examined the effect of sample size on seven genetic distance and two structure metrics. Pairs of samples (equal or unequal) of various sizes were drawn at random from a pool of 856 individuals scored for six microsatellite loci. (δμ)², D(SW), R(ST), and F(ST) were the best performers in terms of bias and variance. Sample sizes of 50 ≤ N ≤ 100 individuals were generally necessary for precise estimation of genetic distances and this value depended on number of loci, number of alleles, and range in allele size. (δμ)² and D(SW) were biased at small sample sizes.