Incorporating additive genetic effects and full LD information to discover genome-level gene-environment interactions with summary statistics of complex traits

Uncovering environmental factors interacting with genetic factors to influence complex traits is important in genetic epidemiology and disease etiology. Existing methods examining gene-environment (G\(\:\times\:\)E) interactions either test G\(\:\times\:\)E interaction for each genetic variant individually, ignoring correlations with additive genetic effects, or use only partial information of the Linkage Disequilibrium (LD), leading to potential loss of statistical power to uncover interacting factors. In this paper, we introduce BiVariate Linkage-Disequilibrium Eigenvalue Regression for Gene-Environment interactions (BV-LDER-GE), a novel statistical method that detects the overall contributions of G\(\:\times\:\)E interactions in the genome using summary statistics of complex traits. BV-LDER-GE harnesses both correlations with additive genetic effects and full LD information to enhance the statistical power to detect genome-scale G\(\:\times\:\)E interactions. Extensive simulations demonstrate that the BV-LDER-GE is more powerful than existing methods while the type-I error rate is well-controlled. When examining 151 environmental covariate-phenotype (E-Y) pairs from the UK Biobank data, BV-LDER-GE identified 28 (80%) more statistically significant interacting environmental covariate phenotype pairs (E-Y pairs) than existing methods.