We conducted a multivariate gene-based GEWIS of SI, testing the interactive effect of SD-related traits. In participants of European descent, we identified multiple genes within the same region of chromosome 16 that showed both significant SD-related interaction effects and a significant association with SI accounting for the possibility of heterogeneous effect sizes due to multivariate SD–gene interactions. Investigating the index variant, rs8052287, we characterized the signal and observed that the multivariate interactions were related to DSM-IV OD, CoD, and ND criterion counts and the severity of polysubstance dependence (i.e., the number of SD diagnoses). These gene-environment interactions account for 98% of the genetic variance within this locus. This is in line with the expected statistical power of the StructLMM method, which has greater power to detect loci with a high fraction of the genetic variance explained by gene-environment interactions27. We could not apply the StructLMM method to the Army STARRS data because of the lack of high-dimensional data on polysubstance dependence. Applying a standard gene-environment test, we did not observe an interaction between rs8052287 and SUDcombined with respect to SI. This may be due to the reduction in the statistical power of standard gene-environment tests compared to the StructLMM approach27.
