Metabolites mediate genetic effects on disease in the Canadian Longitudinal Study on Aging

Abstract

Understanding the biological mechanisms linking genetic variants to disease risk is essential for advancing precision health. We have developed a causal mediation analysis framework, the C-MAPLE (Causal Mediation Analysis of Pathways Linking Exposures) method, to identify disease-causing pathways for which the effect of genetic variants is mediated through metabolites to impact age-related diseases. Unlike Mendelian randomization, our approach is robust to horizontal pleiotropy, and models multiple mediators and interactions between genetic variants and metabolites simultaneously. To ensure robust model selection, we incorporate least absolute shrinkage and selection operator (LASSO) with stability selection, which can effectively select relevant mediators even in the presence of unmeasured confounding. We also introduce a dynamic adjustment to the number of bootstrap trials to reduce computational burden during uncertainty estimation. Applying this novel framework to the Canadian Longitudinal Study on Aging, we identified 190 potential causal links involving 108 genetic variants, 176 metabolites, and 6 age-related diseases. Our method and findings highlight the utility of causal mediation analysis in uncovering metabolite-mediated genetic mechanisms. This method, combined with large-scale population data sets, has the potential to revolutionize the identification of targets for downstream clinical research, and the development of personalized disease prevention, interventions, and therapeutics.