Testing Gene-environment Interactions Without Measuring the Environment

摘要

Both genetic variations and environmental factors play key roles in shaping the etiology of complex disease. Dissecting the interplay between genetics and environmental factors may provide insights into disease development and shed light on treatment strategies. However, studies focusing on gene-environment (GxE) interactions only had limited success, in part due to a lack of large datasets with both genetic data and robust environmental measures. Here, we seek a solution for a challenging problem: can we test GxE interactions without measuring the "E"? We propose a novel statistical framework to test GxE interactions by using polygenic scores (PGS) as proxies for the environmental factors. PGS has gained popularity in GxE research and many studies used PGS as the "G" component. Through theoretical and numerical analyses, we demonstrate that the inference of interaction remains valid after replacing the "E" component with its PGS. We applied our method to three large, independent genetic datasets for autism spectrum disorder (ASD; n = 7,805 probands) to investigate the interaction of genetics and birth weight on ASD risk. We used a PGS derived from the UK biobank (n = 205,475) as the genetic proxy for birth weight. Meta-analysis revealed a significant negative interaction between genetics and birth weight (P = 4.6e-4), suggesting that a higher birth weight may buffer the genetic risk of ASD. As a negative control, we also applied our approach to 3,243 healthy siblings of ASD probands and did not identify any interaction (P = 0.65). We believe this method has great potential for advancing our understanding of complex disease.