Genome-wide scan of 29,141 African Americans finds no evidence of directional selection since admixture

Abstract

The extent of recent selection in admixed populations is currently an unresolved question. We scanned the genomes of 29,141 African Americans and failed to find any genome-wide-significant deviations in local ancestry, indicating no evidence of selection influencing ancestry after admixture. A recent analysis of data from 1,890 African Americans reported that there was evidence of selection in African Americans after their ancestors left Africa, both before and after admixture. Selection after admixture was reported on the basis of deviations in local ancestry, and selection before admixture was reported on the basis of allele-frequency differences between African Americans and African populations. The local-ancestry deviations reported by the previous study did not replicate in our very large sample, and we show that such deviations were expected purely by chance, given the number of hypotheses tested. We further show that the previous study's conclusion of selection in African Americans before admixture is also subject to doubt. This is because the FST statistics they used were inflated and because true signals of unusual allele-frequency differences between African Americans and African populations would be best explained by selection that occurred in Africa prior to migration to the Americas.

Figure 1

Ancestry at Each Location in the Genome in 29,141 African Americans This figure gives the proportion of European ancestry at each of the 118,006 SNPs common to all cohorts. The black line indicates the genome-wide average proportion of European ancestry. The red and blue lines indicate the threshold for genome-wide significance (p < 10⁻⁵) in our study and in the Jin et al. study, respectively. The dashed blue line indicates the significance threshold (p < 2.7 × 10⁻³) that was actually used in the Jin et al. study. The SD was computed empirically over all SNPs. It is clear that no region attained genome-wide significance in our scan. For the six loci reported under selection in Jin et al., dashed vertical lines indicate their location, and blue points indicate their deviation in local ancestry. These deviations are reported in relation to the genome-wide average ancestry proportion in our study. None of the six reported loci exceeded the threshold for genome-wide significance (p < 10⁻⁵) for the Jin et al. study (blue lines).