Substance use traits, ranging from initiation to use disorder, are moderately heritable with hundreds of identified risk loci and potential risk genes. Substance use is strongly correlated with externalizing behaviours, which encompass traits related to impulsivity and risk-taking. Given the particularly strong links between externalizing behaviour and substance initiation, externalizing behaviour may represent a target for early intervention before substance use becomes detrimental. However, it is unclear which aspects of substance use are most strongly correlated with externalizing behaviour when accounting for shared genetic variance across substances.
We have investigated initiation and use disorder of alcohol, cannabis, and tobacco, and externalizing behaviour, using Genomic Network Analysis (GNA). GNA estimates relationships between traits conditioning on other traits within the network, producing a final network containing only significant conditional relationships.
We estimated three substance-specific networks, which suggest that externalizing behaviour is conditionally independent from cannabis use disorder, but not alcohol or tobacco use disorder. A combined network containing all three substances identified externalizing behaviour as conditionally independent from all traits except cannabis initiation, and suggests shared genetic overlap as a key driver of the genetic correlation between substance use traits. These networks were used to estimate associations between the unique genetic components of each trait and individual genetic variants or gene expression. We identified conditional gene expression associations with all traits, particularly in tissues involved with the reward system of the brain, including novel associations identified only after conditioning on the network.
Our findings highlight extensive pleiotropy between substance use traits, resulting in their conditional independence from externalizing behaviour. The multivariate approach utilized by GNA identifies genetic components unique to individual traits while accounting for relationships between traits, enabling the complex interactions between traits to be detangled alongside an improved understanding of what makes traits distinct at the genetic level.