To compare LD differences of associated SNPs to the genome-wide background, we used the varLD targeted option that tests the null hypothesis that the correlations in LD between SNPs from a given window are equal in both populations. We implemented 1,000 permutations to calculate the empirical p-value for each 50-SNP window. Then, we built three cumulative distributions correspondent to each of the three sets of SNPs: replicated (n = 47), non-replicated (n = 64) and random genomic (n = 4,700) SNPs. The SNPs selected for the latter dataset were ascertained from HapMap Phase 2 in order to randomly match the minor allele frequencies of replicated and non-replicated SNPs in Europeans. Finally, we randomly created 100 groups of 47 genomic SNPs to calculate the median and 95% empirical CI of permutation p-values available at Figure 4.
