Modeling Human Gut Microbiome Community Structure in Healthy and Diseased States

? DESCRIPTION (provided by applicant): Historically, bacteria were either seen as pathogenic if they caused disease or benign if they lived commensally with the host. In many cases, however, single organisms have not been identified that are consistently associated with disease, but rather it is thought that bacterial community structure and bacteria- bacteria interactions differ between healthy and diseased individuals, otherwise known as dysbiosis. Naturally, investigating the microbiomes of healthy and diseased individuals using systems biology methodology could lead to insight into the processes that underlie dysbiosis. To that end, methodology has been developed to identify co-occurrence networks from microbiome data. However, to date this methodology has primarily been applied to healthy microbiome datasets and we still lack an understanding of the common properties of dysbiosis. To address these gaps, I propose to build gut microbiome co-occurrence networks for 34 immune diseases and 133 quantitative phenotypes using ~2,500 individuals from the United Kingdom Adult Twins Registry (TwinsUK), for which microbiome and phenotype data has previously been collected. First, I will identify disease-associated bacteria using generalized and linear mixed models that take into account relatedness between individuals. Next, I will use a simulation framework to compare six microbiome co- occurrence network software packages to quantify the consistency of co-occurrence networks for individuals either i) with and without disease or ii) from opposite tails of quantitative phenotype distributions. Using these networks, I will identify modules of co-occurring bacteria that are differentially abundant in disease, examine general network statistics (such as modularity and diversity), and examine the properties of disease-associated nodes as determined in Aim 1 (such as degree, betweenness, and closeness centrality). Additionally, I will take advantage of the twin relationships in our data to identify heritable network characteristics and examine co- heritability of pairs of bacteria. Finally, this will be the first study to compare these network statistics across a range of diseases in order to develop a model of how microbiome network structure differs in health and disease. Additionally, I will compare underlying network structure across phenotypes to identify diseases that share similar gut microbiome alterations. Understanding the gut microbiome at this level will be beneficial for designing therapies to target the microbiota and ultimately improve human health.