Deep phenotyping and dimensional heterogeneity of alcohol use and misuse

Emerging evidence indicates that distinct genetic influences underlie different dimensions of alcohol use behaviors such as consumption versus problems. However, even when these dimensions are considered, research typically uses low-resolution phenotypes like overall consumption quantity or sum scores of alcohol screening questionnaires. Such limited depth may conceal further genetic distinctions relevant for personalized medicine. In addition, the mechanisms linking genes to alcohol use and misuse remain largely unknown. This study leverages deep phenotyping and item-level analysis to better resolve the genetic dimensions underlying alcohol use and misuse. We also incorporate neuroimaging measures to trace the neural pathways of these dimensions. Survey data and longitudinal electronic health records were collected in adults of European (n=386,971), South Asian (n=9,645), and African (n=7,827) ancestry from the population-based UK Biobank sample. Catalogs of all available data were searched to identify 18 phenotypes related to alcohol use (e.g., consumption habits), misuse (e.g., alcohol use disorder diagnoses), and alcohol-related sequelae (e.g., cirrhosis). We conducted genome-wide association analysis (GWAS) on each phenotype, followed by genomic structural equation modelling (gSEM) to characterize the underlying latent genetic factor structure and perform GWAS on the latent factors. Annotation tools and genetic correlation analyses were employed to interpret GWAS and prioritize candidate genes, tissues, and cell types. Neuroimaging was carried out for a subset of these participants, and measures of structural and functional resting state connectivity were derived. Structural equation modelling revealed four latent genetic factors: Consumption, Problems, Atypical Beverage Preference (indexed by drinking spirits and fortified wine), and Beer Preference (indexed by drinking beer but not wine, drinking without meals, receiving advice from a clinician to cut down, and a recent decrease in drinking). GWAS of the latent factors revealed over 50 novel loci. Most associated loci for Beer Preference were novel, and this factor was not associated with previous robust genetic risk factors such as the ADH genes. Associated genes for all factors were enriched for expression in the brain. Genetic correlations were found with neuropsychiatric phenotypes (Problems) and socioeconomic and anthropometric phenotypes (all factors). Additional analyses will test mediation between the associated genes and alcohol use dimensions via addiction-relevant neurocircuitry. Although two factors, Consumption and Problems, were similar to previous studies with lower resolution phenotypes, this analysis identified two additional genetic factors that have not previously been observed. One of these, Beer Preference, may index genetic risk that is more susceptible to intervention. These results suggest that deep phenotyping is an important tool in the future of psychiatric genetic work, allowing the discovery of clinically relevant genetic factors that might otherwise remain hidden. New data collection efforts should focus not only on larger sample sizes but also on the depth and resolution of the phenotypes measured in order to allow for greater insights into the underlying etiology of complex phenotypes like alcohol use and misuse.