Figure 3  Large-Effect eQTLs Influence Essential Genes
(A) dN/dS ratio comparing large-effect family cis-eQTLs to population cis-eQTLs. We selected family eQTLs on the basis of their effect sizes relative to population eQTL effect sizes and plotted the distributions of dN/dS ratios. As a comparison, we show the distribution of dN/dS ratios for the most significant cis-eQTL genes identified only in the population (373 unrelated European individuals from the Geuvadis study) given different p value cutoffs. This is further compared to family-level genes that have rare and potentially regulatory variants (within 5 kb of the TSS, within ENCODE TF binding and DNase I hypersensitivity peaks, and with a PhyloP score > 1). We observed that for large-effect cis-eQTLs and family-level genes with a rare variant, a higher proportion were more conserved (described as the percentage of genes with a dN/dS < 0.3; lower dN/dS ratios indicate higher conservation).
(B) Comparison of centrality in the PPI network between large-effect cis-eQTLs in the family and population cis-eQTLs. Centrality is measured by the number of interacting proteins (degrees). Different groups of genes are defined in the same way as in (A). We show proportions of high-connectivity (hub) genes (degree > 10; higher degrees indicate more essential genes) among these groups. We observed that the proportion of high-connectivity genes was greatest for large-effect cis-eQTLs and family-level genes with a rare variant. This suggests that common regulatory variants are less likely to occur at conserved genes. In contrast, family-specific eQTL effects, because they arise from rare variants, can affect conserved genes.