In order that genes with a large number of associated SNPs in high LD were not disproportionately represented in the risk panel, we generated a list of semi-independent SNPs for the panel and removed SNPs with an r2 greater than 0.50. LD estimations used for pruning the SNPs were based on the HapMap Phase 3 CEU data using the PLINK v1.07 LD function. Selection of which SNP of a pair of correlated SNPs to remove depended on a ranked list of SNPs based on the level of significance from the family-based association results and how closely the SNP on the Illumina chip matched the original family-based SNP. Table 3 summarizes the list of SNPs after pruning. Pruning resulted in a set of 22 SNPs in 15 genes, with several genes pruned out primarily due to correlations among the ADH SNPs.
