Novel Computational Tools for Linking Genotypes to Microbial Community Phenotypes

Author: ORCID icon orcid.org/0000-0001-6492-8180
Biggs, Matthew, Biomedical Engineering - School of Engineering and Applied Science, University of Virginia
Advisor:
Papin, Jason, Department of Biomedical Engineering, University of Virginia
Abstract:

The human microbiome is vital to human health as a metabolic “organ” with essential catabolic and anabolic functions. Development of microbiome-targeted therapies requires mechanistic understanding of how microbes interact with each other and with the host. Mechanistic computational approaches can increase knowledge gains from experiments, infer system properties that are difficult to measure, and accelerate the design of novel therapies. Constraint-based reconstruction and analysis (COBRA) of genome-scale metabolic networks is a powerful mechanistic approach that has been widely applied to the study of metabolism in single species, but is underdeveloped as a tool for studying metabolism in microbial communities. Challenges for applying COBRA methods to the study of microbial communities include a lack of frameworks for integrating other data relevant to the community such as spatial structure, difficulty reconstructing species-specific metabolic networks in the absence of reference genomes, and the burdensome need for months of manual network curation before high-quality predictions can be made.

We addressed these challenges and advanced the ability to simulate microbial communities using COBRA methods on three fronts. First, we created two novel modeling frameworks that integrate genome-scale metabolic networks with additional community information including spatial organization and known microbial interactions. Second, we developed a novel algorithm to assign short metagenomic sequencing fragments to the correct genome using information inherent in metabolic networks, thus increasing the availability of strain-specific genomes for network reconstruction. Finally, we developed an ensemble approach which improves the prediction accuracy of automatically-generated networks, making it possible to generate higher quality predictions without the extensive time requirement of manual network curation.
These three advances pave the way for much broader adoption of mechanistic, multiscale models in microbiome research.

Degree:
PHD (Doctor of Philosophy)
Keywords:
Genome-scale metabolic network, Constraint-based analysis, Ensemble analysis, Microbiome, Multiscale modeling, Pseudomonas aeruginosa, Computational biology, Microbial metabolism, EnsembleFBA, MatNet, SONEC, Altered Schaedler Flora (ASF)
Language:
English
Issued Date:
2016/11/29