Item-Level Genome-Wide Association Study of the Alcohol Use Disorders Identification Test in Three Population-Based Cohorts.

Genetic relationships between AUDIT items. Path diagram of the best fitting genetic confirmatory factor model for AUDIT, as estimated with Genomic Structural Equation Modeling. All parameter estimates are standardized, and standard errors are presented in parentheses. The genetic components of items and factors (denoted by g) are inferred variables that are represented as circles. Regression relationships between variables are represented as straight one-headed arrows pointing from the independent variable(s) to the dependent variable(s). Covariance relationships are depicted as curved two-headed arrows linking two variables. The variances for factors are represented as a two-headed arrow connecting the variable to itself, as are the residual variances for individual items (denoted by u). As item 6 was included via factor extension, its parameter estimates are illustrated using dashed lines.

Genetic correlations between latent AUDIT phenotypes and other complex traits. Bar charts of the genetic correlation (rg) results for three AUDIT phenotypes: Consumption (green), Problems (blue), and Frequency Residual (gray). Point estimates and corresponding standard errors (SEs) are displayed for select phenotypes related to substance use, psychopathology, impulsivity, cognition, and socioeconomic factors. Full results are reported in Table S8.

Multivariate genome-wide association analyses for the latent genetic factors. Miami plot for the two latent genetic factors: Consumption (top) and Problems (bottom). Approximately independent lead SNPs are labeled with a white diamond. For each lead SNP, the neared gene is labeled. Additional symbols convey findings from additional biological annotation; filled symbols indicate that the gene was identified in the corresponding pipeline, while empty symbols indicate that the gene was not. The y-axis refers to the significance on a -log10 scale, the x-axis refers to chromosomal position, the horizontal dotted line marks suggestive significance (p = 1E-5), and the horizontal dashed line denotes genome-wide significance (p = 5E-8).

Associations between Consumption and Problems PRS and selected alcohol-related phenotypes. Bar charts of the variance explained by Consumption and Problems PRS for various clinical and quantitative measures of alcohol use. Values correspond to the proportion of variance explained the outcome (R2 or pseudo R2 depending on the use of linear or logistic regression; see Supplementary Section 7 for more details). Results for the independent UK Biobank subsample are presented on the left, while results for the independent COGA cohort are presented on the right. Please note that the COGA models are not directly comparable to those from the UKB models, as mixed-effect models were used in COGA. Please also note that the R2 for each PRS is calculated from a single PRS model and, as such, the values not independent (due to shared variance between PRSs). Complete results are available in Tables S32–S35.

Phenome-wide association study of polygenic risk scores for Consumption (left panel) and Problems (right panel) against 1,338 diseases available in the biobank from Vanderbilt University Medical Center, BioVU. PheWAS of both Consumption and Problems revealed positive genetic associations with alcohol use disorders. Problems was positively genetically associated with multiple psychiatric conditions, whereas Consumption was counterintuitively negatively associated with metabolic conditions. Importantly, most of the associations disappear when adjusting for alcohol use disorders diagnosis (non-significant associations are highlighted in gray).