Category: Fst/Gst

Failure of Fst/Gst

Population genetics.

Both “movement” fetishists as well as anti-racist liars like to misuse Fst/Gst in genetic distance discussions (*) to promote their respective agendas.  Unfortunately for them, Fst/Gst is not really a (direct) measure of genetic distance, and particularly fails even as an indirect proxy when comparing populations that exhibit different levels of heterozygosity (e.g., human ethnies) and/or when considering loci with more than two allele variants.  To have the ill-informed trying to parse differences of, say, Fst/Gst = 0.0060 vs. 0.0065 and trying to make relevant conclusions from that is laughable.  The following are a small sampling of links to cite the next time some idiot tries to play such games (emphasis added):

See here.

See also here:

Likewise, when diversity is equated with heterozygosity, standard similarity measures formed by taking the ratio of mean within-subpopulation diversity to total diversity necessarily approach unity when diversity is high, even if the subpopulations are completely dissimilar (no shared alleles). None of these measures can be interpreted as measures of differentiation or similarity. 

At Wikipedia:

Also, strictly speaking FST is not a genetic distance, as it does not satisfy the triangle inequality. As a consequence new tools for measuring genetic differentiation continue being developed.

And this article here:

One underutilized approach is the coupling of indirect metrics of gene flow (e.g. F-statistics, Dest_Chao) with more direct measures such as kinship or parentage analyses (e.g. Loiselle et al. 1995; Selkoe et al. 2006; Buston et al. 2009; Christie et al. 2010; Palsbøll et al. 2010). Broadly speaking, kinship analyses provide an index of the relative relatedness of all genotyped individuals in a data set, and parentage is a distinct case of kinship whereby the most likely parents of individual juveniles are identified (Vekemans & Hardy 2004; Jones & Arden 2003; reviewed in Blouin 2003; Jones et al. 2010). Kinship coefficients (also known as coefficients of coancestry) are widely interpreted as the probability of identity by descent of the genes, but they are more properly defined as ‘ratios of differences of probabilities of identity in state’ (Hardy & Vekemans 2002, p. 23) from homologous genes sampled randomly from each pair of individuals (Hardy & Vekemans 2002; Rousset 2002; Blouin 2003; Vekemans & Hardy 2004).

By comparison, F-statistics and Dest_Chao are often blind to the relatedness of individuals; different population samples with the same kinship structure can have very different levels of genetic differentiation among them and vice versa.

*True, Salter used Fst in On Genetic Interests, but only because there was no other data available for that purpose at that time.  And Salter makes clear in the book that the proper approach would be to use data from global assays of genetic kinship which did not (and still do not) exist for human ethnies.  It is interesting that population geneticists and ecologists will calculate genetic kinship for plant and non-human animal species, but are either too lazy or politically-motivated to do so for human population groups. However, anecdotal evidence from the genetic kinship data that companies such as 23andme and DeCode used to present to their customers suggest that human genetic kinship findings would not be to the liking of either the fetishists or the anti-racists. 

Advertisements