To evaluate the causal effect of SI and CPD on BMI, schizophrenia and educational attainment (EA), we conducted Mendelian randomization (MR) analyses using three complementary approaches available in MR-Base [30]: inverse variance weighted regression [31], MR-Egger [32, 33], and weighted median [34]. We used both the previously reported smoking-associated SNVs and the SNVs from the current report (as provided in Tables 1–3 and Supp. Table 3) as instrumental variables. The BMI [35], schizophrenia [36] and educational attainment [37] data came from previously published publicly available data. To assess possible reverse causation, we also used outcome associated SNVs as instrumental variables and conducted MR analyses using SI and CPD as outcome. We considered P < 0.05/3 = 0.017 as statistically significant (Bonferroni-adjusted for three traits).Table 1Association results for SNVs identified in single variant association meta-analyses and taken forward to replication are provideddbSNP ID (Exome ID)Chr:PosEA/OAGeneConsequenceTraitDiscovery stageReplication stageNEAFDoEP-valueBeta (SE)P-valuers141611945 (exm118559)1:161771868G/AATF6MissenseCPD128,7460.0065%MAC = 9+2.95 × 10−70.184 (0.169)*P = 0.276 in African American samplesrs1190736 ** (exm1659559)X:136113464A/CGPR101MissenseCPD (PY)99,037 (96,824)46.6% (47.0%)-1.40 × 10−11 (4.98 x 10−9)−0.028 (0.0041)−0.027 (0.0049)−0.028 (0.0073)All samples: 8.20 x 10−12 (2.70 x 10−11)Males only:
