The incidence of variants falling into each RegulomeDB category is remarkably stable across individuals particularly when comparing heterozygous and homozygous variant calls. On average, 11.20% of variants fall into ChIP-seq peaks, 21.53% into DNase peaks, 1.29% into eQTLs, 6.15% into footprints, and 39.25% into PWM matches (Supplemental Table S2). Of significance is the difference in rates of annotations for heterozygous and homozygous variant calls (Fig. 2A). In the cases of PWMs, DNase HS sites, FAIRE regions, histone modifications, DNase footprints, and TF ChIP-seq data, the rate of homozygous and heterozygous variants compared to the total count of homozygous/heterozygous variants in each is consistently and significantly different in that homozygous variants in these regulatory annotations are significantly less than those of heterozygous variants (paired, two-tailed t-test; P < 5 × 10−15) (Supplemental Table S2). A genomic background would expect equal fractions of homozygous and heterozygous variants to be covered by regulatory features. Thus, these results indicate a purifying selection is occurring in the RegulomeDB annotated regions.
