PCA was performed using the multi-dimensional scaling procedure implemented in PLINK v1.06 (http://pngu.mgh.harvard.edu/purcell/plink/) [61]. The vast majority of samples resided in a single cluster including all the individuals of self-reported European ancestry, with several moderately outlying samples corresponding to self-reported Hispanic and African ancestry. No samples were excluded from further analyses; the vectors determined for the first two principal components were used as ancestry control for all statistical analyses.
