Correlation between type 2 diabetes gene association scores and potential gene score confounders.

Pearson's correlation coefficients were calculated between , or and six different physical and genetic properties of genes. is a vector of the unadjusted best SNP per gene z-scores for all genes in the genome, is a vector of corrected gene z-scores using regression analysis for all genes, and is a vector of corrected gene z-scores using phenotype permutation analysis for all genes. This was computed for 1,000 phenotype permutation data sets of the Diabetes Genetics Initiative (DGI) GWA study and the actual DGI GWA study. Aside for gene size, all gene properties were converted to per kilobase (kb) units for each gene by dividing by gene region size using the extended physical boundaries. All correlations between and the six variables were statistically significant (mean p<2e-70 across 1,000 DGI permutations and p<1e-74 for the actual DGI study). Similar correlations were obtained for the five latter variables in Table 1 before normalizing to gene region size (data not shown).
†These gene properties were significant in almost all 1,000 DGI GWA permutations tested under a step-wise multivariate linear regression model of regressed against the six gene properties (see Table S3).
*The linkage disequilibrium units per kb variable was significant under the regression model for about half of the permutations tested (Table S3).