Cumulative effect of multi-exposures on complex traits and diseases: derivation and applications of the polyexposure risk score

Abstract

Human diseases and traits are influenced by a combination of genetic and environmental factors. Genetic risk scores, which reflect the aggregate genetic predisposition from many markers across the genome, are being developed and updated for disease prediction and risk stratification. However, environmental risk scores are still limited to a handful of exposures selected a priori, and it is unclear whether exposures cumulatively impact disease risk burden. We lack a systematic and unbiased approach to construct a non-genetic environmental score that considers the entire “exposome”. In current-day biobank and high-dimensional epidemiologic cohort ‘omic data, there exist new opportunities to systematically consider the effects of multiple exposures. In the following dissertation, we introduce the polyexposure risk score (PXS), a weighted sum of the effects of many non-genetic exposure factors, and describe the regularization-based method used to extract non-redundant and independent exposures for the PXS. We develop an R package, PXStools, which includes a set of analytical tools that allow users to efficiently build PXS as well as to conduct other exposure-related analyses. In addition, we develop and validate PXS for several common diseases and traits in participants of the UK Biobank and demonstrate the abilities of the PXS to improve the prediction and risk stratification of diseases. Taken together, we provide a roadmap for studying the exposome and the comprehensive effects of the environment on complex phenotypes.