PUTTING POLYGENIC SCORES IN CONTEXT: HOW INTERSECTIONAL FACTORS AFFECT RELATIVE AND ABSOLUTE GENETIC RISK

The clinical utility of polygenic scores (PGS) is known to vary when training and test samples differ in ancestry. Recent work also suggests that variation in demographic and environmental contexts can affect PGS performance. However, little attention has been given to the ubiquitous phenomenon of intersectionality—where individuals simultaneously belong to multiple demographic and environmental contexts—and its impact on relative and absolute genetic risk. We examined how lifetime odds ratios (ORs) and absolute risk (AR) for high-PGS individuals differ across intersectional contexts in the UK Biobank (UKB; n = 375,054, British-European-like ancestry), with a particular focus on major depressive disorder (MDD). Additional phenotypes studied included atrial fibrillation, coronary artery disease, type 2 diabetes, hypercholesterolemia, asthma, and obesity. PGS were computed using PRS-CS and current GWAS summary statistics. We analyzed 106 two-way intersections across sociodemographic factors (sex, age, income, smoking, alcohol intake, and deprivation). Replication was conducted in the All of Us Research Program (AoU; African-like: n = 36,552; European-like: n = 99,477). Odds ratios for major depression varied substantially across sociodemographic contexts, with even greater variability across intersectional combinations. In two-way intersections, the average maximal OR variation was 56% across all phenotypes. Although absolute risk differences were more moderate after adjusting for baseline prevalence within a context, intersectional effects were still observed. For example, high-PGS individuals with low income had an average 1.0 percentage-point lower AR across phenotypes when using a context-aware versus context-unaware model. For major depression specifically, the AR was 3.1 percentage points lower among those also reporting low alcohol intake in UKB (3.6 points lower in AoU European-like ancestry). Results were broadly consistent across cohorts, with the strongest replication for depression risk observed in the European-like All of Us sample. We show that intersectional contexts can substantially shape relative genetic risk. Future clinical applications may need to consider incorporating contextual effects to enhance the precision and equity of patient-specific genetic risk assessments.