Pervasive Gene–Environment Interactions Reshape Polygenic Risk Score Utility Across Exposures
核心洞察
A large-scale analysis reveals pervasive interactions between polygenic risk and environmental exposures across 2,249 disease–exposure pairs in the UK Biobank, fundamentally challenging assumptions of uniform genetic risk effects.
The study demonstrates that polygenic risk score (PRS) performance varies substantially across exposure strata, with significant implications for risk stratification accuracy and clinical implementation.
Researchers found that incorporating exposure-stratified PRS models can identify individuals who may derive greater benefit from interventions, potentially informing more effective precision prevention strategies.
A comprehensive investigation into the interplay between polygenic risk scores (PRS) and environmental exposures has uncovered pervasive gene–environment interactions that could fundamentally alter how genetic risk prediction is deployed in clinical practice. The study, drawing on data from the UK Biobank, systematically evaluated interactions across thousands of disease–exposure combinations, revealing that the predictive performance of PRS is not uniform but instead varies markedly depending on an individual's exposure context.
The findings arrive at a critical juncture for genomic medicine, as PRS are increasingly being evaluated for clinical implementation in screening, prevention, and treatment stratification for complex diseases including cardiovascular disease (搜索), diabetes (搜索), and cancer (搜索). Yet the study's results suggest that current approaches—which typically assume consistent genetic risk effects across all individuals—may overlook substantial heterogeneity driven by modifiable and non-modifiable exposures.
Pervasive Interactions Across the Phenome
The research documented significant gene–environment interactions across an expansive landscape of 2,249 disease–exposure pairs. These interactions were not limited to a handful of well-studied relationships but appeared broadly across the phenome, suggesting that context-dependent genetic risk is a fundamental feature of complex disease architecture rather than an exception.
This finding aligns with growing recognition in the field that PRS performance can vary substantially within ancestry groups, as highlighted by Mostafavi and colleagues, and that risk factors affecting polygenic score performance operate across diverse cohorts. The pervasive nature of these interactions carries profound implications: a PRS calibrated in one population or exposure context may not perform equivalently in another, potentially leading to misclassification of risk and suboptimal clinical decision-making.
Implications for Risk Stratification and Clinical Utility
One of the most clinically relevant findings concerns how exposure-stratified PRS models can refine risk prediction. The study demonstrated that accounting for gene–environment interactions enables more precise identification of individuals at elevated risk within specific exposure strata. This has direct implications for the number needed to treat (NNT)—a clinically useful measure of treatment effect—by potentially enriching for individuals most likely to benefit from preventive interventions.
Prior work has already shown that PRS can identify subgroups with greater burden of atherosclerosis (搜索) and greater relative benefit from statin therapy in the primary prevention setting. The new findings extend this principle by demonstrating that such benefit may be further modulated by environmental context, suggesting that integrated risk models incorporating both polygenic risk and exposure information could substantially improve the efficiency of preventive strategies.
The research also resonates with evidence that highly elevated polygenic risk scores are better predictors of myocardial infarction (搜索) risk early in life than later, and that modification of coronary artery disease (搜索) clinical risk factors by coronary artery disease PRS can inform more tailored approaches across the life course.
Methodological Considerations and Challenges
The study highlights important methodological challenges in detecting and modeling gene–environment interactions. Prior work has documented type 1 error inflation and power loss in GxE PRS models, and the current research underscores the need for robust statistical frameworks that can reliably identify true interactions while accounting for potential confounding.
Tradeoffs in modeling context dependency in complex trait genetics remain a central challenge. The study's authors note that different analytical approaches—from genome-wide by environment interaction studies to gene–context interaction methods—offer complementary insights but also carry distinct limitations. Calibrated prediction intervals for polygenic scores across diverse contexts represent one promising avenue for conveying the uncertainty inherent in context-dependent risk estimates.
Equity and Generalizability Concerns
The findings carry significant implications for health equity. As Martin and colleagues have cautioned, clinical use of current polygenic risk scores may exacerbate health disparities if PRS performance varies systematically across groups defined not only by genetic ancestry but also by environmental exposures that correlate with social determinants of health.
The study's demonstration of pervasive exposure-dependent PRS performance adds another dimension to ongoing efforts to improve cross-population transferability, portability, and calibration of PRS. Research from the eMERGE network and the All of Us Research Program has emphasized the importance of diverse cohorts for developing more generalizable polygenic risk scores, and the current findings suggest that diversity in environmental exposures must be considered alongside genetic diversity.
Toward More Effective Precision Prevention
The research points toward a future in which PRS are not deployed in isolation but are integrated with comprehensive exposure information to generate context-sensitive risk estimates. This vision aligns with multimodal models that integrate PRS with clinical, molecular, imaging, environmental, lifestyle, and wearable-derived data—an approach that may ultimately prove more clinically informative than genetic risk scores alone.
By demonstrating that gene–environment interactions are pervasive rather than exceptional, the study provides a compelling rationale for rethinking how polygenic risk information is communicated to patients and clinicians. Rather than presenting PRS as a fixed, intrinsic risk, context-dependent risk estimates may better reflect the dynamic interplay between genetic susceptibility and modifiable exposures—potentially empowering more effective and equitable precision prevention strategies.