LARSEN-DISSERTATION-2017.pdf (5.76 MB)
Modeling Host-Microbiome Interactions
thesisposted on 2018-02-08, 00:00 authored by Peter E Larsen
Humans are not individual organisms. Rather, we are complex, dynamic ecosystems, comprised of interacting communities of human cells and a wide variety of microorganisms. Without these symbiotic microbial communities, our existence, and perhaps the existence of all complex forms of life, would be impossible. Recently, the profound impact of these communities on human health has begun to be understood, made possible by the advent of ultra-high throughput sequencing. In many regards, opening these communities to investigation has only highlighted the tremendous diversity in microbial populations from host to host and over time in the same host. This diversity makes associating specific microbial species with their effects on host health difficult. Here, we propose a set of integrated computational biology tools that link bacterial sensor networks, microbiome community interactions, and microbiome metabolome to host health. Using previously published genomic and metagenomic datasets from bacterial, human, and mouse-model microbiome experiments, we have generated computational models that span multiple biological scales of host-microbiome interactions: the interaction between an individual bacterium’s metabolome and transportome with their host, a microbiome community’s metabolome influence on host health, and a dynamic model of the interactions between microbiome community, host, and host diet. Together, these elements will be used to generate a comprehensive model, iMOUSE, which can be used to reliably predict the results of biological experiments in silico and to identify an optimal diet for a host based on the host’s current microbiome. These approaches necessitated the development of novel methods for metabolomics and transportomic modeling, accurate prediction of a microbiome’s community enzyme function profile from its community structure, and the creation of a unique computational method for predicting dynamic microbiome community interactions. From these analyses, we come to the important conclusion that host-microbiome interactions are not best described as interactions between individual community members, but rather host-microbiome interactions are an emergent property of the microbiome community. The computational models we have developed form a cohesive and comprehensive toolbox for understanding host-microbiome interactions and developing microbiome-based therapies for disease or to promote health that can be expanded far beyond the specific datasets analyzed here.
Degree GrantorUniversity of Illinois at Chicago