Friday, November 17, 2006

Race & Brazil - the sequel   posted by Razib @ 11/17/2006 08:38:00 PM

3 years ago a study came out which suggested that the color classifications in Brazil don't match that well to the ancestry as ascertained by loci which exhibit disjoint frequencies between Africans and Europeans. Yanni points me follow up study which agrees with the previous findings. You can read the pre-print here. Since 2003 a lot of material has come out on traits like skin color. We now know that around ~1/3 of the quantitative difference in complexion between Africans and Europeans is due to one locus which manifests in disjoint allele frequencies. Overall the work seems to be moving toward confirming the inferred genetic architecture derived from pedigree analysis that skin color as a trait emerges from the combine action of 4-5 loci of large effect. How does this relate to this study? Below I referred to "Ecotype Persistence" to define the perpetuation of adaptive alleles and phenotypes even when most of the local genetic pool is replaced by immigrant alleles.

Consider a contrived scenario. You have an island population which is coadapted to a local highland parasite. Sea levels drop and recontact is made with the parent mainland population. During the period of isolation genetic drift and local evolution has resulted in a non-clinal variance in allele frequencies between the "island" and "mainland" population. Even though reunited by a narrow isthmus assume that cultural barriers prevent a great deal of intermarriage. Also, let us assume that the mainland population is very large relative to the island population. If you had a small amount of intermarriage per generation the island population could be genetically assimilated to the mainland population rather quickly. If at a locus, A, there was a 1% replacement of "native" alleles with "mainland" alleles, within 2,000 years 2/3 of the alleles in the island population would be derived from the mainland. Now, consider another locus, B, which is implicated in adaptation to the highland parasite. If you assume that an "island" allele confers greater than a 1% fitness benefit vis-a-vis the immigrant "mainland" alleles, then it can persist indefinitely. 2,000 years later locus B will have preserved the ancestral character of the population, but locus A will be similar to the state of the mainlanders. In fact, on the vast majority of the genome the island population will be assimilated to that of the mainland, but on specific adaptively beneficial alleles the ancestral character will be maintained in the totality.

The point is to illustrate the power of selection in confounding our assumptions about genetic admixture and long term evolutionary dynamics when we use a population level lens. One reason that the ornate fantasies of early 20th century historical physical anthropology were so popular is that they appealed to our bias in imagining the movements of peoples across time & space, with the admixture of originally pure Platonic races. The imaginings of Madison Grant and company were only moderately more rooted in fact than the migrations of Tolkein's elves.

Which gets me back to the paper above. First, caveats. The original paper was met with a lot of skepticism. Some of it is probably warranted as the flavor of Political Correctness is pretty transparent in the author's text and statements from what I can see. Nevertheless, the beauty of science is the judgement of reality. Review the results yourself. One Brazilian commenter noted that they used one population in one village, but the individual didn't bother to read the original paper where they used some control groups in other cities to double-check their finding. Though I believe that the results need to be viewed as only part of the picture (e.g., I suspect there are plenty of unadmixed Europeans in the far south of Brazil, and nearly pure Africans in places like Bahia), I think they do capture some of the reality of racial admixture in that nation. The basic results were that though self-identified and researcher identified blacks, browns (in the sense of being mixed race) and whites did differ genetically, the difference was far less than one might have assumed and did not match what one would have predicted if phenotype and ancestry tracked perfectly. What happened here? First, remember that a few genes of large effect induce the skin color difference. Though racial ascertainment is a gestalt perception based on many variables, those variables are finite, a very small subset of visible characters controlled by a small number of genes which do not necessarily have to be reflected in the rest of the genome. The recent prominence of black and white twins born from mixed-race parents illustrates the issue: the relatively small number of characters used to judge race can quickly be reshuffled and segregated in a mixed population. Consider a population of white Brazilians, assume that a number, x, of white looking mixed-race Brazilians "passed" every generation. Though the loci which control physical appearance were predominantly derived from their European ancestors, these individuals would still carry and introduce a larger number of African alleles which do not exhibit any visible outward effects. Over a number of generations a non-trivial proportion of non-European ancestry could easily introgress into a population which is overwhelmingly European on loci which shape the outward phenotype. Conversely, more African looking mixed-race individuals would bring into the black population European alleles. What you see here is the power of social selection in maintaining modal phenotypes via assortative mating!