Insights from genetically stratified analyses comparing subtypes of alcohol misuse.

In a recent publication, we applied a novel model to address phenotypic heterogeneity in genetic research on alcohol misuse by stratifying individuals based on their patterns of alcohol use behaviours and comorbid psychopathology. In this Commentary, we provide further descriptions of the subtypes of alcohol misuse that emerged from the empirical mixture modelling approach and present new results comparing these groups on sociodemographic characteristics and additional alcohol use outcomes. We take a broad perspective to discuss how these results fit with existing typologies of alcohol misuse and how the results inform future genetic research. Our findings add further evidence that conceptualisations of a binary distinction between 'internalising' (relief-seeking) versus 'externalising' (reward-seeking) subtypes does not fully capture the complexity of alcohol misuse. However, accounting for individual differences in these dimensions is a promising means to reduce heterogeneity and thereby improve power for gene discovery and, eventually, personalised medicine applications. We argue that more detailed, person-specific assessment of alcohol misuse measures, particularly with attention to longitudinal trajectories, is needed to further advance this important line of research.

Investigating genetically stratified subgroups to better understand the etiology of alcohol misuse.

Alcohol misuse (AM) is highly prevalent and harmful, with theorized subgroups differing on internalizing and externalizing dimensions. Despite known heterogeneity, genome-wide association studies (GWAS) are usually conducted on unidimensional phenotypes. These approaches have identified important genes related to AM but fail to capture a large part of the heritability, even with recent increases in sample sizes. This study aimed to address phenotypic heterogeneity in GWAS to aid gene finding and to uncover the etiology of different types of AM. Genetic and phenotypic data from 410,414 unrelated individuals of multiple ancestry groups (primarily European) in the UK Biobank were obtained. Mixture modeling was applied to measures of alcohol misuse and internalizing/externalizing psychopathology to uncover phenotypically homogenous subclasses, which were carried forward to GWAS and functional annotation. A four-class model emerged with "low risk", "internalizing-light/non-drinkers", "heavy alcohol use-low impairment", and "broad high risk" classes. SNP heritability ranged from 3 to 18% and both known AM signals and novel signals were captured by genomic risk loci. Class comparisons showed distinct patterns of regional brain tissue enrichment and genetic correlations with internalizing and externalizing phenotypes. Despite some limitations, this study demonstrated the utility of genetic research on homogenous subclasses. Not only were novel genetic signals identified that might be used for follow-up studies, but addressing phenotypic heterogeneity allows for the discovery and investigation of differential genetic vulnerabilities in the development of AM, which is an important step towards the goal of personalized medicine.