To evaluate if the effects of the top variants identified in the PTSD GWAS were specific to PTSD, we conditioned PTSD on MDD, and MDD plus BPD plus SCZ using the multi-trait conditional and joint analysis (mtCOJO)24 feature in GCTA to regress out the effects of correlated traits based on external GWAS summary data. MDD was selected here as the main psychiatric trait because of the high co-morbidity and genetic correlation of depressive symptoms and PTSD (rg = 0.80 for depressive symptoms and rg = 0.62 for MDD; see Supplementary Data 3). Publicly available summary statistics were supplied as program inputs: Bipolar cases vs. controls for BPD, and MDD2 excluding 23andMe for MDD (both from https://www.med. unc.edu/pgc/results-and-downloads); Schizophrenia: CLOZUK + PGC2 meta-analysis for SCZ (http://walters.psycm.cf.ac.uk/). The effect of each psychiatric disorder on PTSD was estimated using a generalized summary-data based Mendelian randomization analysis of significant LD independent psychiatric trait SNPs (r2 < 0.05, based on 1000 G Phase 3 CEU samples), where the threshold for significance was set to P < 5 × 10−7 due to having less than the
