All 266 samples included from the published Merck study were self-reported Caucasians. The SNP set was trimmed using linkage disequilibrium (LD)-based SNP pruning, removing all SNPs for with high pairwise LD (R2>0.8), as in [59]. PCA was performed using the multi-dimensoinal scaling procedure implemented in PLINK v1.07 (http://pngu.mgh.harvard.edu/purcell/plink/) [61]. No outliers were detected; the vectors determined for the first four principal components were used as ancestry control for all statistical analyses.
