Finally, we combined heritability predictors and gene disease designations into several multiple regressions (Supplementary Table 7). The predictors were as shown in Table 2, with the addition of eQTL evidence (best local and distant r2), chromosomes 6 (HLA genes), 19 (which was an outlier in gene density analysis), and X (underrepresented in GWAS), and a blood DNase hypersensitivity / gene conservation interaction (identified in exploratory analyses). eQTL evidence alone (top local and distant SNPs) explained 23.9% of the variation in h2, the full model 32.9%. h2 remained significantly predictive of OMIM/NHGRI disease status except for the smaller sets of NHGRI genes subdivided by immune designation, even while the best local and distant eQTLs were no longer significant. Gene conservation was highly predictive for OMIM status. Gene density was strongly negatively associated with disease status, but this effect was attenuated for OMIM. NHGRI disease status was significantly enriched for chromosome 6, and showed a deficit on chrX, which we attribute to the neglect of chrX in GWAS.62 OMIM showed enrichment of chromosome X, consistent with the importance of X-linked disorders in medical genetics.
