Disaggregating the Genetic Overlap Between Educational Attainment and Substance Use Phenotypes

BackgroundEducational attainment (EA) is positively genetically correlated with alcohol and cannabis use, but negatively correlated with alcohol and cannabis use disorders. These opposing associations suggest that shared genetic influences differ by level of substance involvement and for the cognitive (CogEA) and non-cognitive (NonCogEA) components of EA. MethodsTo test this, we examined the shared genetic architecture of EA, CogEA, and NonCogEA with alcohol consumption (AC), alcohol use disorder (AUD), lifetime cannabis use (CanUse), and cannabis use disorder (CUD). We used bivariate causal mixture models, local genetic correlation analyses, and conditional/conjunctional false discovery rate analyses to identify polygenic, regional, and variant-level overlap. ResultsFor variants shared with AC, 48.12% and 52.86% showed concordant effects with CogEA and NonCogEA, respectively, compared to 38.40% and 41.02% for AUD. For CanUse, 71.42% and 65.56% of shared variants were concordant with CogEA and NonCogEA, respectively, versus 37.97% and 42.23% for CUD. Jointly associated loci were identified for each substance-EA pair. AUD and CogEA variants were enriched for expression in the basal ganglia, CUD and CogEA in the substantia nigra, and CanUse and NonCogEA in the cerebellum and its hemispheres. ConclusionsAlthough EA is associated with greater substance use and lower risk for disorder, this pattern reflects a complex mix of concordant and discordant variant effects. CogEA and NonCogEA show partially distinct patterns, particularly for cannabis-related traits, highlighting the importance of disaggregating EA to clarify the genetic architecture underlying its paradoxical associations with substance-related traits.