F7. POLYGENIC SCORES ACROSS ENVIRONMENTAL CONTEXTS: HOW INTERSECTIONALITY AFFECTS PREDICTIVE ACCURACY FOR PSYCHIATRIC CONDITIONS

• Published in: European neuropsychopharmacology Vol. 87; p. 209
• Main Authors: Cudic, Mihael, Tubbs, Justin, Smoller, Jordan W.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.10.2024
• ISSN: 0924-977X
• DOI: 10.1016/j.euroneuro.2024.08.418

Polygenic Scores (PGS) have shown promise for clinical risk stratification, but their utility for psychiatric conditions is limited partially due to the complex interplay between genetics and the environment. Recent evidence has begun to indicate that PGS predictiveness varies considerably across demographic and environmental contexts. However, current work often overlooks intersectionality, where individuals belong to multiple contexts simultaneously. In this study, we test for differences in PGS variance explained on the liability scale (R²) across intersectional contexts in the All of Us Cohort (n=105,570) for three psychiatric conditions known to be largely influenced by the environment: Alcohol Use Disorder (AUD; n=6,030), Generalized Anxiety Disorder (GAD; n=32,828), and Major Depressive Disorder (MDD; n=21,828). Specifically, we use PRS-CS to calculate PGS from the latest available GWAS summary statistics and examine variations in PGS R² across overlapping intersections of sex, age, deprivation, income, education, smoking, and drinking behavior. As hypothesized, R² from PGS exhibited substantial variation across different environmental intersections. For example, the overall R² for MDD was 1.62%, which was higher among individuals in low deprivation contexts (1.81%), and among males (1.65%). However, when we consider the intersectional context of males with low deprivation, the MDD PGS R² was higher (2.01%) than either context in isolation. In general, we find widespread significant interactions between PGS, sex, and other environmental factors, with the variability in R² being consistently higher among males than females. For instance, R² values for GAD in males ranged from 0.06 % to 1.77%, while for females, these values spanned from 0.07 % to 1.00%. Additionally, R² values across environmental contexts showed only moderate correlation between males and females, with correlation coefficients for AUD, GAD, and MDD being 0.45, 0.26, and 0.51 respectively. In some cases, we find divergent directions of gene-environment interactions. In the case of AUD, we observed an increase in the PGS R² in females who smoke versus non-smokers, whereas the PGS R² was lower in men who smoke compared to non-smokers. These results highlight the importance of considering intersectional contexts when modeling PGS in psychiatry, where R² can vary widely in individuals with different combinations of demographic and environmental factors. Future clinical applications should account for these unique intersections to inform patient-specific risk that is more tailored to an individual's context.