Univariate GCTA-GREML46 analyses were used to estimate the proportion of variance explained by all common (MAF > 1%) SNPs for each of the depression phenotypes. A relatedness cutoff of 0.05 was used in the generation of the genetic relationship matrix, as including close relatives inflates heritability estimates47. This did not alter the sample size in UKB due to previous sample filtering, however in GS:SFHS this reduced the sample size by 38.5–58.4% (Supplementary Table 15). In GS:SFHS, the first 20 PCs were fitted as fixed effects. In UKB, batch, recruitment centre and the first 8 PCs were fitted. Univariate Linkage Disequilibrium Score regression (LDSR), implemented in LD Score (v1.0.0.)36, was applied to GWAS summary statistics to evaluate the proportion of inflation in the test statistics caused by confounding biases, such as population stratification, relative to genuine polygenicity. This method also provides an estimate of SNP-based heritability. Pre-computed LD Scores were used, estimated from the European-ancestry samples in the 1000 Genomes Project43. To obtain heritability estimates on the liability scale, sample and population prevalence estimates were used. Sample prevalence estimates were calculated
