Mapping and characterizing caQTLs in liver tissue. (A) Liver chromatin accessibility data generated by ATAC-seq. (B) caQTL mapping strategy testing variants within 1 kb of ATAC peak centers. (C) caPeaks (FDR < 5%; purple) and non-caPeaks (FastQTL beta-adjusted P > 0.5; gray) overlapping chromatin states from liver tissue from the NIH Epigenomics Roadmap. (D) Number of caQTLs identified after downsampling read depth or sample size to various percentages of full depth. Mean (bars) and standard deviation (error bars) for five technical replicates. Approximate number of read pairs used in each downsampled percentage: 17 million, 34 million, 51 million, and 68 million. Number of samples used in each downsampled percentage: 27, 55, 82, and 110. (E) TF motifs more often disrupted by caQTL variants than non-caQTL variants. Log odds ratios and 95% confidence intervals. Only motifs with P < 1.8 × 10−4 and odds ratio (OR) ≥ 2 are shown. Full results are shown in Supplemental Table 8. (F) Comparison of minor allele frequencies (MAFs) in TOPMed African (AFR) and European (EUR) populations for the lead variants of the 156 caQTLs that are not polymorphic in the individuals genetically similar to European populations. (G) caQTL for caPeak165117 with lead variant rs6758168 not detected in individuals genetically similar to European populations. The G allele, associated with lower chromatin accessibility, was observed in individuals genetically similar to African (blue) and admixed (red), but not European populations. Frequencies correspond to individuals in this study. The boxplots represent ATAC peak counts normalized for covariates. (H) Sequence logo motif for HNF4A, which is disrupted by rs6758168 (position shown by arrow). The G allele is predicted to result in a weaker motif match.
