randomly assigned to SNPs within each GWAS, so that the positive SNPs in each GWAS change at random, but the number of positive SNPs remained the same. This simulation generated distributions about the number of positive SNPs with 1~5 supporting GWAS, with the means of these distributions tallied as (E1, E2, E3, E4, E5); (vii) the significance of intersection for the real data was assessed by the minimum meta-false discovery rate (mFDR) calculated as mFDR = Minimum ([Ei]/[Ti]) for i = 1 to j, 1 <j < = 5. If mFDR < 0.05, a meta-signature was defined as those SNPs that are significantly identified (p-value < T) in at least j of 5 independent GWAS, where j is equal to i when mFDR was defined. The p-Value threshold (T) with 0.05 and 0.01 were calculated respectively and significant results were combined for further study. On the basis of the HapMap Linkage Disequilibrium data compiled from genotype data (HapMap data release rel#21 NCBI B35) [33], we further expanded this list using SNP pairs with strong linkage disequilibrium (r2> = 0.8) in all three HapMap populations. The protocol was implemented in Perl.
