In each individual dataset, a statistically significant association was seen between polygenic risk score for schizophrenia and the substance of ascertainment. Specifically, tobacco use disorder was most strongly associated with schizophrenia polygenic risk score in COGEND, the dataset ascertained to study nicotine dependence (Figure 2A, pseudo R2 range 0.42%–5.8%, minimum p<9.5×10−20), alcohol use disorder was most strongly associated with schizophrenia polygenic risk score in COGA, the dataset ascertained to study alcohol dependence (Figure 2B, pseudo R2 range 0.57%–2.1%, minimum p<7×10−6), and stimulant use disorder was most strongly associated with schizophrenia polygenic risk score in FSCD, the dataset ascertained to study cocaine dependence (Figure 2C, pseudo R2 range 1.7–3.5%, minimum p<9×10−5). Cannabis use disorder was associated with polygenic risk score only in FSCD, the dataset ascertained to study cocaine dependence. These results led to statistically significant associations between schizophrenia polygenic risk score and tobacco, alcohol, cannabis, and stimulant use disorder in the mega-analysis (Figure 2D, tobacco use disorder pseudo R2 range 0.5%–3.1%, minimum p=2×10−18; alcohol use disorder pseudo R2 range 0.04%–0.4%, minimum p=4×10−4; cannabis use disorder pseudo R2 range 0.09%–0.5%, minimum p=0.001; stimulant use disorder pseudo R2 range 0.06%–0.7%, minimum p=5×10−5).
