Saturday, January 10, 2009

Selection or demography in differences between human populations?   posted by p-ter @ 1/10/2009 08:13:00 AM

Dan MacArthur points to a paper claiming that large allele frequency differences between populations are due to demographic effects. The data the authors are working with is a set of a few thousand markers (SNPs and others) genotyped on 53 populations from across the world. Their main points boil down to two things:

1. "Large" allele frequency differences are "surprisingly" common between human populations.

2. Such allele frequency differences are not enriched near genes (as would be expected if genes are more likely targets of positive selection than non-genic regions).

This work can essentially be seen as a push back against the trend towards finding "evidence" of positive selection in the human genome in any gene one finds interesting, and the authors cite a number of papers that fully or partially base their claims for selection on allele frequency differences between populations. As a warning about the caveats in such types of analysis, this is a useful paper, but it's important not to overstate what the data actually say:

1. When the authors say that large allele frequency differences are common, it's important to define "large". In this case, they're talking about things with an allele frequency difference of 0.3 or above. That is, if an allele is at 30% frequency in Africa and 60% frequency in Asia, that counts. How you define large is obviously subjective, and personally I wouldn't have chosen that threshold. But in any case, the authors are right to say that if you see an allele frequency difference of 0.3 and 0.4 between continental populations in your favorite gene, that alone is not strong evidence for selection.

2. The enrichment (or lack thereof) of large allele differences near genes was more comprehensively studied in a paper from about a year ago. The authors there found that there is indeed such an enrichment, but that it occurs at a more stringent definition of "large" than the one considered here. So the fact that allele frequency differences of 0.3 are not enriched near genes is not all that surprising.

To summarize, this paper shows that many claims about selection on individual loci based entirely on modest (what the authors call large) allele frequency differences between populations are massively overstating their evidence. But then again, you already knew that.