One of the major reasons that so much human genetic work is fixated on ascertaining the nature of population substructure is that different populations may respond differently to particular drugs. Of course population identification is only a rough proxy, but in many cases it is a good one. Years ago, Neil Risch reported the utility of genetic markers in differentiating individuals into distinct groups, and the high fidelity of these identifications with self-report. But this sort of generalization is contingent on particular conditions. In Brazil centuries of admixture have resulted in a considerable amount of gene flow between various phenotypic races. That is, Brazilians may identify as white, mixed or black, but the two ostensibly non-mixed categories often are characterized by non-trivial admixture. This means that the marginal utility of genetic tests to ascertain the individual's genetic background above & beyond their self-identification may be greater in Brazil than in the United States.
The GNB3 825C>T allele and genotype distribution differed significantly across the three self-reported groups...with a trend for increasing frequency of both the GNB3 825T allele and the TT genotype from White to Intermediate to Black individuals...Non-linear logistic regression showed that the odds of having the GNB3 825C>T allele increase monotonically...with increasing ACA throughout the ACA range...observed in the overall population sample, irrespective of "racial/color" self-identification.
The T allele is extant at a frequency of 80% among Africans and 30% among Europeans. Among Brazilians the distribution is:
These data are as one would expect, roughly. But these are averages...there's a lot of intra-population variation:
...The average proportion of individual cluster 2 ancestry among the self-identified White, Intermediate and Black subjects enrolled in this study was 0.326 (range 0.136-0.567), 0.497 (0.208-0.836) and 0.651 (0.229-0.897), respectively. From this trend we inferred that cluster 2 is an estimate of the "African component of ancestry" (ACA).
Note that the "mixed" category nearly spans the African ancestry proportions of individuals who might also identify as black or white. Because of this variation within these populations, the authors contend that a model of continuous genetic variation (where the proportion of African ancestry is the variable) adds more power:
...Taking advantage of the fact that the individuals enrolled in this study were previously typed with a validated set of AIMs, we applied a non-linear logistic regression model...to investigate the association of the GNB3 825C>T polymorphisms with the estimated individual ancestry in the ACA. The results of this exercise are presented graphically in Fig. 1. The ACA associates significantly with the 825C>T polymorphisms in our Brazilian population sample. Specifically, the odds of having the 825T allele increase...with increasing ACA throughout the range of the ACAs...observed in this population sample, irrespective of self-reported "color/racial" identity.