Admixture Studies and the Detection of Selection

See allHide authors and affiliations

Science  15 Jun 1973:
Vol. 180, Issue 4091, pp. 1137-1143
DOI: 10.1126/science.180.4091.1137


Three methods can be used to search for evidence of natural selection from admixture studies. These include classification of the admixture estimates into two groups; calculation of the rank correlation between estimates from more than one population; and the testing of admixture estimates for homogeneity. The use of these methods is discussed with special reference to black-white admixture in the United States. Using revised estimates of African gene frequencies, derived from a consideration of the geographical origin of the slaves, we calculated admixture estimates and their variances for five U.S. black populations; Claxton (Georgia); Sapelo Island (Georgia); James Island (South Carolina); Charleston (South Carolina); Oakland (California). Two out of the five populations yielded heterogeneous admixture estimates but all other tests were non-significant. The data provide little evidence for the action of selection. The few, inconsistent significant results are more indicative of the action of random drift or biased gene frequency estimates than natural selection, and in general these effects cannot be differentiated. It seems doubtful that admixture studies can ever provide unequivocal evidence for the action of natural selection in human populations. In the search for natural selection, perhaps admixture studies should only be used as a preliminary screening device.