In order to verify self-reported ancestry and assess admixture within racial groups, we used the software Structure version 2.1 [32], which implements a model-based clustering method for inferring population structure using genotype data for an unlinked subset of the markers selected because the minor allele frequencies for each population are characteristic of ancestry. There were 11 participants who self-reported mixed ancestry, reported their primary ancestry inconsistently or did not report ancestry. Of the 11 participants in question, one was excluded for a low QC call rate, five were determined to be of mixed heritage (<80% ethnic identity) and were excluded, and five were found to exceed the 80% threshold and were included. Because the population definitions became more stringent with the exclusion of participants with measured admixture, two individuals previously measured >80% fell below this threshold and were excluded. Ultimately, 11 Asians, 13 Hispanics and four people with ambiguous ethnicity were not analyzed further because of small sample size for these groups. Three pairs of individuals were found to be related through the program Relpair version 2.0.1 [33] and were determined subsequently to be siblings; in each case, the one with the most complete data was retained.
