Sponsored by:
Detecting polygenic gene-environment interactions (GxE) can provide crucial insights into the interplay between genes and environments and how they jointly shape human complex phenotypes in the population. The common approach used to identify the polygenic GxE is to test the interactions between polygenic score (PGS) and environment for the outcome of interest (i.e., PGSxE). However, PGSxE has several drawbacks, including the lack of robustness to the widespread sample overlap and loss of predictive power when constructing the PGS, which prevents the accurate estimation of polygenic GxE effects. Here, we address these limitations by introducing PIGEON, a novel framework to detect polygenic GxE effects that only requires GWAS and SNPxE summary statistics. PIGEON provides unbiased estimates for polygenic GxE effects and places an upper bound on the effects obtained from the PGSxE analysis. PIGEON is robust to sample overlap, population stratification, and can debias the polygenic GxE effects driven by G-E correlation. We further propose a PIGEON-based scanning approach for a hypothesis-free search of polygenic GxE. We showcased the superior performance of PIGEON over PGSxE through extensive simulations and real-data analysis. We validated our approach by replicating the polygenic G x Education reform interactions for health-related outcomes in UK Biobank. We further applied PIGEON to scan for polygenic GxSex effects for 530 traits in UK Biobank and identified 337 significant results. Overall, PIGEON represents a robust, unbiased, privacy-preserving, and computationally efficient method for detecting polygenic GxE and may have broader applications in future gene-environment studies.