The density plots were drawn for the histograms of log(PGR) values with different colors and line styles representing each of 11 different HapMap3 populations. PGR represents Predicted Genetic Risk (PGR) with higher risk on the right. Seven diseases showed significantly differential PGR across populations, including type 2 diabetes (A), colorectal cancer (B), cleft palate (C), type 1 diabetes (D), prostate cancer (E), Parkinson's disease (F), and lung cancer (G). p_Afr shows the likelihood of observing lager log(PGR, African)-log(PGR, Other) values after randomly replacing disease genotypes with global frequency-matched genomic genotypes. For example, an average value of log(PGR) is 0.648 in African populations and −0.224 in other populations for T2D. After randomly replacing T2D genotypes with control genomic genotypes, there is only 4.7×10−3 chance finding an average value of log(PGR, African)-log(PGR, Other) larger than 0.872. Similarly, p_Asi and p_Eur represent the likelihoods of observing more extreme values of log(PGR, Asian)-log(PGR, Other) and log(PGR, European)-log(PGR, others) using randomly selected genomic genotypes, respectively. All p values were calculated as two-sided p values. SNPs used in each figure are summarized in Table S4.