We applied S-LDSC (v.1.0.0)3 to partition the common (MAF > 5%) SNP heritability of 111 polygenic traits and diseases. We partitioned heritability using a customized version of the baseline-LD model, accounting for 69 cell-type-nonspecific baseline-LD annotations, and added one or more IMPACT annotations to the model to test for cell-type-specific enrichment. Here, heritability refers to the genetic variation causally explained by common SNPs as defined previously3, as opposed to genotyping array-based SNP heritability61,62. We use three metrics to evaluate how well our IMPACT annotations capture polygenic heritability: enrichment3, the proportion of heritability explained by the top 5% of SNPs3 and per-annotation standardized effect size, τ* (ref.35). Briefly, enrichment is defined as the proportion of common SNP heritability divided by the genome-wide proportion of SNPs in the annotation, for continuous annotations this is the average annotation value across SNPs. τ* represents the average per-SNP heritability of a category of SNPs, where a single SNP may claim membership to one or more categories. τ* has units of heritability and is comparable between traits, annotation and populations, because it is normalized for the
