We performed 13 GWASs (12 items + sum-score) on sample 1 and applied bivariate LD score regression16,17 on the summary statistics to compute genetic correlations (rg) between all 13 phenotypes (upper triangle Fig. 1; Supplementary Data 1; see Methods for QC and technical details). Between items, rg’s ranged from a low 0.38 (IRR/WORR-EMB; s.e. = 0.048) to a high 0.91 (MOOD/FED-UP; s.e. = 0.030, mean rg = .64), and none of the 95% confidence intervals included 1. These results indicate genetic heterogeneity in the full item set. To replicate these findings, we conducted the same 13 GWASs in sample 2 (lower triangle Fig. 1). The rg’s between the exact same items measured in both samples were all close to 1 (diagonal Fig. 1; all 95% confidence intervals included 1) indicating that the genetic signal is highly similar across the two samples. In addition, the correlation between inter-item rg’s in samples 1 and 2 was 0.97, confirming a highly similar rg pattern across the two samples.
