The iControlDB subjects were then evaluated for population structure to identify ancestral outliers. We implemented the pairwise population concordance test in PLINK, which is based on the observed proportion of IBS loci pairs, and we identified eight subjects who were significantly different (P<0.0005) from 95% of the rest of the population. The ancestral outliers were confirmed using the STRUCTURE program [21] and were excluded from further analyses. Ancestral proportions of all iControlDB subjects were inferred via comparison to HapMap CEU, YRI, CHB, and ASW populations in STRUCTURE. As shown in Figure S2, the resulting ancestral proportions indicate that the identified ancestral outliers have relatively high Asian ancestry (up to 78%) or European ancestry (up to 98%). Given that the minimum African ancestral proportion among ASW subjects was 52.7%, we excluded 179 subjects with an African ancestry less than 60%. A list of subjects excluded due to ancestral misclassification is provided in Table S1. The final analysis dataset included 595 (71.7%) subjects.
