G ST is still a useful measure of genetic differentiation — a comment on Jost's D

In a recent issue of Molecular Ecology, Jost (2008) suggests that the widely used measure of genetic differentiation GST and its relatives do not measure differentiation. His conjecture is based on the now well-known dependence of GST on average within population heterozygosity (HS) which prevents GST from taking values larger than average homozygosity (1 − HS; Jin & Chakraborty 1995; Charlesworth 1998; Nagylaki 1998; Hedrick 1999, 2005). To eliminate the influence of heterozygosity on GST, Jost (2008) suggests that genetic diversity is quantified in terms of effective number of alleles rather than heterozygosity, and he derives an alternative measure, D, claimed to be independent of heterozygosity. He proposes that D should replace GST when differentiation is the quantity of interest. We are not convinced that GST has served its time as a measure of genetic differentiation. Whereas GST is mathematically constrained by HS, there is no such constraint on D, implying that D can take any value in the range 0–1 regardless of HS. The lack of such a restriction does not imply that D is expected to change independently of heterozygosity, however. We argue that D shares the same problems of dependence on HS and mutation rate as GST and its relatives, and that these problems are sometimes even more pronounced for D than for GST. There is current and renewed interest in the assessment of diversity in ecology and population genetics (e.g. Pavoine et al. 2005; Ricotta & Szeidl 2006; Jost 2007). The D measure suggested by Jost quantifies diversity at a gene locus as the inverse of gene identity (1/J), equivalent to the effective number of alleles (Crow & Kimura 1970), rather than as 1 − J which is a basic quantity for GST. An attractive characteristic of an approach using 1/J is that it is based on purely mathematical requirements for measures of diversity, and that it lends itself to partitioning diversity in a hierarchical way that is intuitively appealing. However, despite the utility of measures based on 1/J for apportioning the allelic diversity at a locus, there are other circumstances where GST is more appropriate. This measure has to a large extent been used to describe population structure with special focus on the effects of genetic drift and migration, which are the evolutionary forces reflecting demographic population characteristics that under selective neutrality are expected to affect all loci in the same way. As pointed out by Slatkin (1991), for assessment of population structure it would be desirable to have a measure of differentiation that does not confound the purely demographic processes of genetic drift and migration with purely genetic processes such as mutation. Unfortunately, as has become increasingly obvious during the past decade (e.g. Hedrick 1999), GST is affected by mutation and heterozygosity in a way that sometimes makes it difficult to compare results obtained from loci with markedly different mutation rates. The point we make here is that Jost’s D is also affected by mutation and heterozygosity in a way that can be even more pronounced than for GST. Thus, although based on an approach for quantifying variation that is appealing for apportioning allelic diversity, the D measure shares the same limitations as GST when it comes to comparing divergence at loci with different mutation rates. In addition, in some cases D approaches mutation– migration–drift equilibrium at a markedly slower rate than GST, which has implications for its usefulness when estimating migration rates assuming equilibrium. GST is defined as