Multi-trait Analysis of GWAS (MTAG) of Substance Use Traits Identifies Novel Genetic Loci and Phenomic Associations

IntroductionA large majority of genome-wide significant (GWS) loci identified for substance use traits (SUTs) in genome-wide association studies (GWAS) have been for alcohol and smoking-related phenotypes. GWAS of opioid use disorder (OUD) and cannabis use disorder (CUD) have lagged those of the two historically legal substances. MethodsWe applied multi-trait analysis of GWAS (MTAG) to 2,888,727 single nucleotide polymorphisms (SNPs) common to GWAS of four SUTs (OUD, CUD, alcohol use disorder [AUD], and smoking initiation [SMK]) in European-ancestry (EUR) subjects. We calculated polygenic risk scores (PRS) for the four traits in an independent sample (i.e., the Yale-Penn sample; N=5,692 EUR) and examined the power increment for each set of MTAG-GWAS summary statistics relative to those of the input GWAS. ResultsMTAG increased the effective sample size for all four SUTs, which showed high pairwise genetic correlations. After clumping, MTAG identified independent GWS SNPs for all 4 traits: 41 SNPs in 36 loci (including 5 novel loci not previously associated with any SUT) for OUD; 74 SNPs in 60 loci (including 4 novel loci) for CUD; 63 SNPs in 52 loci (including 10 novel loci) for AUD; and 183 SNPs in 144 loci (including 8 novel loci) for SMK. In PRS analyses in the Yale-Penn sample, the MTAG-derived PRS consistently yielded more significant associations with both the corresponding substance use disorder diagnosis and multiple related phenotypes than each of the 4 GWAS-derived PRS. ConclusionsMTAG boosted the number of GWS loci for the 4 SUTs, including identifying genes not previously linked to any SUT. MTAG-derived PRS also showed stronger associations with expected phenotypes than PRS for the input GWAS. MTAG can be used to identify novel associations for SUTs, especially those with sample sizes smaller than for historically legal substances.