Resolving Causal Influences Among Correlated Risk Biomarkers for Coronary Artery Disease

点击翻译按钮获取中文摘要

PROJECT SUMMARY / ABSTRACT Epidemiological studies have shown correlations among numerous biomarkers (defined as measurable indicators of the severity or presence of a disease state) and risk for coronary artery disease (CAD). However, it's unknown whether many of these biomarkers represent causal processes for CAD. Inferring causality of a biomarker with CAD has the potential to identify risk factors that may lead to pathophysiological processes for the development of CAD. Recently, we developed a method, called Multi-Phenotype Mendelian Randomization, that disentangles causal influences for a disease among a set of correlated biomarkers. We applied our method to plasma triglycerides and showed that the effect size of a SNP on triglycerides is linearly related to its effect size on CAD, before and after accounting for the same SNP's potential effect on plasma low-density lipoprotein cholesterol (LDL-C) and/or high-density lipoprotein cholesterol (HDL-C). This finding has since been validated by other studies. Together, these results suggest that plasma triglycerides may capture causal processes that may promote atherosclerosis and CAD. We propose to expand on our prior work by inferring causal relationships between a wide range of 32 cardiometabolic traits, 245 metabolites and >2,000 clinical phenotypes from electronic medical records with subclinical CAD endophenotypes. In Aim 1, we will evaluate current Mendelian randomization methods and refine the approach to allow for detection of pleiotropy (or detection of violation of a basic assumption of Mendelian randomization), which can improve statistical properties of these methods. In Aim 2, we will infer causality of a wide range of 32 cardiometabolic and 245 metabolite traits with subclinical atherosclerosis and cardiac structure and function endophenotypes for CAD. In Aim 3, we will perform a novel framework called Phenome-Wide Mendelian Randomization to infer causality of CAD traits with >2,000 clinical phenotypes from electronic medical records (EMR). The proposal is innovative because we are utilizing novel approaches for causal inference, along with a detailed repository of cardiometabolic traits, metabolites, EMR clinical phenotypes, and subclinical CAD disease traits. We propose to use the following resources: 1) new causal inference approach that accounts for pleiotropy; 2) extensive set of cardiometabolic traits (32 in total) and metabolites (245 in total); 3) subclinical CAD outcomes (42 subclinical atherosclerosis and 54 cardiac structure and function traits); and 4) EMR phenotypes from large-scale Mount Sinai's BioMe Biobank and UK Biobank (>2,000). This proposal has the potential to reveal new causal risk biomarkers for subclinical CAD disease outcomes. It can provide new avenues for the development of new therapeutics for the prevention and treatment of CAD.