To find confounders on we systematically tested for correlations between the unadjusted gene score, calculated from permuted DGI GWA study (see Materials and Methods) and six potential gene score confounders (listed in Table 1; correlations reported for z-scores). We examined both physical properties of genes - physical gene size and number of SNPs per kilobase for each gene, and genetic properties that consider the dependency between subsets of SNPs due to genetic linkage between proximal markers. The genetic properties tested included estimated number of independent SNPs per gene (SNPs in linkage equilibrium), number of recombination hotspots spanning each gene, genetic distance of the gene, and linkage disequilibrium (LD) unit distance per gene, normalized to the size of the gene and its extended boundaries (see Materials and Methods). We found significant correlations for all six properties tested (average values across 1,000 permuted data sets: 0.17<r<0.38; p<2e-70) (Table 1), suggesting that all variables may have a confounding effect on . A similar trend was observed using from the actual DGI GWA study (Table 1, column 2; 0.14<r<0.39, p<1e-74), and the T2D GWA meta-analysis, used below to test the mitochondria-diabetes hypothesis (Table S2).
