In a disease association study where cases and controls are drawn from distinct populations, there is a mathematical relationship between the distribution of allele frequency differences and the distribution of disease association statistics (see Materials and Methods). We obtained empirical agreement with this theoretical result by simulating a case-control study in which 100 unrelated samples with majority ancestry from region 4 were labeled as disease cases and 100 unrelated samples with majority ancestry from region 5 were labeled as controls. We computed Cochran-Armitage trend statistics and obtained a genomic control λ of 1.285, consistent with the predicted value of (1+NF ST) = 1.28 given the F ST of 0.0014 between the two regions (see Materials and Methods). After dividing by Cochran-Armitage trend statistics by the genomic control λ, the most significant association had a P-value of 3×10−6, which is not statistically significant after correcting for the number of SNPs tested. We repeated this analysis for all pairs of regions (4,5,6,8,10) with 100 unrelated samples available (see Table 1), and obtained similar results (minimum P-value of 4×10−7, which is not statistically significant after correcting for the number of SNPs and number of pairs of regions tested.)
