The SNP-based heritability captured by common variants was assessed using linkage disequilibrium score regression (LDSC) for each meta-analysed set of data (25). SNP-based heritability estimates were transformed to the liability scale, assuming population prevalences of 15% for combined MDD, 1% for PGC BD, and 16% for MOOD, and lower and upper bounds of these prevalences for comparison (Supplementary Methods). LDSC separates genome-wide inflation into components due to polygenicity and confounding (25). Inflation not due to polygenicity was quantified as (intercept-1)/(mean observed chi-square-1) (26). Genetic correlations were calculated in LDSC between each analysis and 414 traits curated from published GWAS. Local estimates of SNP-based heritability and genetic covariance were obtained using HESS v0.5.3b (Supplementary Methods and Results) (27, 28).
