Multiple-Locus Allocation of Individuals to Groups as a Function of the Genetic Variation Within and Differences Among Human Populations

Multivariate (multiple-locus) analytic procedures are used to show that the amount of genetic overlap of two or more populations, as measured by the difficulty of correctly allocating individuals, may be formally related to the average genetic distance between pairs of populations. On theoretical grounds, we argue that the probability of correct allocation for individuals should: (1) increase with the number of segregating loci; (2) increase with increasing taxonomic disparity of the populations considered; and (3) decrease with an increase in the number of candidate populations. Genetic data on 5,214 individuals from seven South American Indian tribes are used to evaluate these theoretical predictions. We have analyzed genetic distance and classification with the seven tribes, as well as with seven village clusters within one tribe and seven villages within one cluster. At all three levels of taxonomic organization (tribal, cluster, village), the probability of correct allocation did increase with the number of loci, as predicted. This pattern was consistent with the computed average distances between populations, also as predicted. Some loci were more useful for allocation than others. In general, high-frequency polymorphic markers were the most useful, while tribally restricted (but low-frequency) markers were not helpful. There was consistent evidence that the genetic overlap for tribes was less than that for clusters and that the overlap for clusters was less than that for villages. As a consequence, tribal allocation was more accurate than cluster allocation and cluster allocation was more accurate than village allocation. To place these differences in wider perspective, we computed the genetic distance measure for a set of seven "racial" groups. The predicted overlap for races was substantially less than that for tribes. Finally, we compared our results with those of others, and showed that the differences result from our use of a multiple-locus rather than a single-locus analysis. The two types of results are strictly compatible, however, once this technical gap is bridged. The conclusion is inescapable that some human populations show substantial genetic differences, measured relative to the variation within populations. For populations as similar as nearby villages, genetic overlap is substantial; for populations as different as racial groups, genetic overlap is small.