PROJECT SUMMARY / ABSTRACT The human body is home to a complex community of microorganisms (“microbiome”) that differs in composition between people, with numerous correlates to cardiovascular disease (CVD). Any two people will harbor different strains of a given species, which can be more genetically different than a human and chimpanzee with <60% of their genes shared. Even within a single person, each microbiome species may be a complex mixture of strains with different genomes and functional capabilities. This striking within-species genetic diversity has functional consequences for CVD, because gene loss and gain modify how strains process our diet, metabolize drugs, and stimulate inflammation. Hence, a population genetic approach is essential for revealing causal links between the microbiome and CVD. We have compiled a deeply phenotyped cohort of ~50,000 individuals with metagenomic sequencing of their gut microbiomes. This dataset includes ~8,000 people with atherosclerosis, thousands with measurements of heart function and metabolic health, and hundreds with acute coronary syndrome. This cohort is a unique and ideal setting to perform a well-powered CVD metagenome-wide association study (MWAS). Several barriers must be overcome before MWAS can be deployed at this scale. First, we must reduce the infeasible computational cost of genotyping thousands of microbiome species across ~50,000 people. Second, to ensure that statistical tests for associations do not have high false positive rates we need statistical models that adjust for microbial population structure within and across hosts. The goal of this proposal is to create a research toolbox to address these challenges as well as to identify putative mechanistic links between microbiome and CVD. We will develop data structures and query algorithms for accelerated genotype estimation and mixed effects models for accurate association tests. All code and methods will be open source and designed to be easily extended to other microbiome cohorts. Applying these tools to our cohort, we aim to identify specific microbial genes and pathways responsible for known associations between microbes and CVD. We also expect to discover new associations that were missed because cohorts were too small or they were analyzed with methods that ignore differences in gene content across strains. These findings will be used to identify microbial biomarkers for CVD diagnosis and personalized treatments or to design microbiome targeted drugs, prebiotics, and probiotics to treat heart disease.

PROJECT NARRATIVE This project aims to transform microbiome population genetics so that association studies can be performed at the same level of resolution and scale as in human genetics. We will produce a computationally-efficient method for genotyping microbiome species in situ using metagenomics and statistical models for associating microbiome genes with host traits while accounting for microbial population genetic structure. Applying these tools to our deeply phenotyped cohort of ~50,000 individuals, we will test the hypothesis that cardiovascular disease is linked to microbial genes that vary across strains and alter how gut microbes influence human physiology.