Statistical Methods for Large Scale Microbiome Studies of Cardiovascular Disease Risk

点击翻译按钮获取中文摘要

PROJECT SUMMARY Cardiovascular diseases (CVD) are the leading cause of death globally, affecting nearly half of all adults in the United States and responsible for a quarter of all deaths each year, with rates continuing to rise. The microbiome offers a unique paradigm for investigating and eventually mitigating the global burden of CVD. Aided by high-throughput sequencing technology, microbiome profiling studies have found bacterial communities to be related to CVD risk factors, including hypertension and systolic blood pressure. This knowledge is invaluable both from the perspective of better understanding etiology and from the perspective of therapeutic development, as the microbiome is inherently modifiable. However, although many studies have demonstrated possible relationships, the specific bacterial taxa associated with CVD risk factors, as well as the manner in which they are related, are poorly understood. Recently, large scale microbiome profiling studies of hundreds to thousands of individuals have been conducted within existing, on-going cohort studies. These studies offer a unique opportunity to achieve more thorough investigation of the impact of microbiomes on CVD risk factors. Unfortunately, the statistical and computational approaches for analyzing these studies are lacking. This proposal aims to fill critical gaps in the methodological literature by addressing four major areas. Specifically, we aim to develop comprehensive suite of statistical tools for (1) addressing batch effects in microbiome studies – increasingly problematic as studies get bigger; (2) improved identification of individual taxa associated with CVD risk factors; (3) conducting mediation analysis and understanding the relative role of microbiota and exposures on risk factors; and (4) assessing the role of the microbiota as an effect modifier. These approaches are all based on rigorous prior data emphasizing the importance of the problems as well as the limitations or absence of existing strategies. Our work is motivated by and will directly enable analyses in three of the largest, and richest microbiome profiling studies around: few studies have the combination of sample size and richness of covariates as the CARDIA, MEC, and SOL microbiome studies. Consequently, our methods have the potential for accelerating understanding of the role of microbes in CVD and facilitate development of therapies and strategies for stemming the rising CVD epidemic.