We also created discovery and replication samples by splitting just the FSCD and COGEND portion of the SAGE GWAS sample in half, and then assessing for clinical validity in the COGA GWAS sample. Of the list of SNPs that met nominal significance criteria in both halves of the SAGE sample, the majority of SNPs did not share the same direction of effect, suggesting that many of these results could be false positives. This study also explored the effect of using a more stringent r2 threshold of 0.25 to prune the list of candidate gene SNPs before creating sum scores; results were similar.
