We calculated genetic correlations using LDSC regression 10 and the online software LD hub (http://ldsc.broadinstitute.org/) 65. LDSC regression leverages linkage disequilibrium (LD) information for each genetic variant such that the χ2 statistic for that variant includes the effects of all loci in LD with that variant. This can be extended to the analysis of genetic correlations between traits if the χ2 statistic is replaced with the product of 2 z-scores from 2 traits of interest 66. Genetic correlations were calculated between the three datasets that contributed to the meta-analysis. Genetic correlations were also calculated between our meta-analysed results (23andMe_307k, UK Biobank and PGC_139k) for depression and 234 behavioural and disease related traits. P-values were false discovery rate (FDR) adjusted67 and correlations reported if PFDR < 0.01.
