Lastly, the average proportion of predicted functional SNVs per individual varied between 2.3 and 4.2% (Fig. 4D). When the more liberal definition of functional SNVs was used, EA individuals have a significantly higher proportion of predicted functional SNVs compared with AA individuals (Fig. 4D; Wilcoxon-rank sum test; P < 10−15), consistent with empirical estimates and theoretical expectations because of the lower EA effective population size (28, 29). However, when the more conservative definition was used, this pattern was reversed, and AA individuals have a significantly higher proportion of predicted functional SNVs compared with EA individuals (Fig. 4D; Wilcoxon-rank sum test; P < 10−15). These results highlight how the definition of functional variants can influence inferences and underscore the importance of continued methodological development to robustly identify functionally important variation. Nonetheless, there was considerable rare genetic variation among individuals that is predicted to be functional, which could explain variability in disease risk and adverse drug response.
