Predicted target genes of caPeaks. (A) Four methods used to link caPeaks to genes: transcription start site (TSS) proximity, promoter capture Hi-C from HepG2 and liver tissue, coordination of caQTLs for distal and promoter peaks, and colocalization of caQTLs with liver tissue eQTLs. (B) A HMGCL TSS-proximal caPeak (peak4713) for a caQTL colocalized with an HMGCL eQTL. The caQTL lead variant (rs58035855) is shown in each plot by a purple diamond, and colors represent linkage disequilibrium (LD) r2 values from 1000G Europeans. Gray points represent variants that were not in the 1000G LD reference panel. (C) A caPeak linked to CDO1 by Hi-C. Read depth–normalized ATAC signal tracks averaged by genotype for rs6863733, which is the lead variant in the 1 kb and 1 Mb caQTL analyses. (D) The distribution of log10 distances between caPeaks and the 5′-most TSS of linked genes. “HepG2 Hi-C 1” and “HepG2 Hi-C 2” indicate Hi-C results from two independent studies. (E) Percentage of caPeak target genes (purple) and non-caPeak genes (gray) enriched in liver tissue or other tissues. The non-caPeak target genes are any gene detected in the Human Protein Atlas that was not linked to a caPeak. (F) Logistic regression results testing if links between peaks and TFs are more likely to be supported by each of the linking methods relative to non-TF genes. Odds ratios and 95% confidence intervals, with P-values above. (G) Logistic regression results testing if coordinated caQTLs are more likely than noncoordinated caQTLs to colocalize with GTEx liver tissue eQTLs. Odds ratios and 95% confidence intervals for enrichment using sets of coordinated caQTLs with exactly two peaks, exactly three peaks, exactly four peaks, and five or more peaks.
