GWAS variant enrichment analysis

We used LD score regression (48, 111) to estimate genome-wide enrichment for GWAS traits using annotation sets from single-cell chromatin accessibility from the heart or lung (54) or bulk DNase hypersensitivity sites for tissues from ENCODE (3, 5, 17). For bulk DNase-seq datasets, peak annotations were merged across biological replicates from the same tissue type. We compiled published GWAS summary statistics for cardiovascular diseases (49–53), other diseases (112–122), and nondisease traits (123–132) using the European subset from transethnic studies where applicable. We created custom LD score files by using peaks from each cell type or tissue as a binary annotation. As background, we used baseline annotations included in the baseline-LD model v2.2. For each trait, we used LD score regression to estimate enrichment coefficient z scores for each annotation relative to the background. Using these z scores, we computed two-sided P values for enrichment and used the Benjamini-Hochberg procedure to correct for multiple tests within each set of annotations.