Microbiome

Personalized models of the gut microbiome are valuable for disease prevention and treatment. For this, one requires a mathematical model that predicts microbial community composition and the emergent behavior of microbial communities. In our 2019 paper (arxiv link), we sought a modeling strategy that can capture emergent behavior when built from sets of universal individual interactions. Our investigation revealed that species-metabolite interaction modeling is better able to capture emergent behavior in community composition dynamics than direct species-species modeling.

In order to create species-metabolite interaction models without detailed time-course data, we investigated (arxiv link) the use of genome-scale metabolic models (GEMs) to define interactions between microbes and metabolites. GEMs are used via an optimization procedure called flux balance analysis (FBA) to determine the flux rates of reactions within a cell, including reactions which exchange metabolites with the environment, and the rate at which the cell produces biomass. Using FBA to define a differential-algebraic system for community dynamics is known as dynamic-FBA. We designed an efficient method for computing solutions to dynamic-FBA systems.

I am building a thorough understanding of microbial communities based on models of individual community members and the network of interactions formed within the community. I am using the central concepts of metabolic modeling as axioms for developing an algebraic structure to describe how microbes interact with their environment. My central conjecture is that these relationships form the basis of a topology for microbial communities, with functions that are capable of predicting the results of population perturbations.

I am investigating equivalence and similarity between metabolisms and how this equivalence effects the dynamic-FBA system of a community.

Understanding the population dynamics of the human microbiome will allow us to predict the engraftment of harmful taxa such as C. difficile and design probiotic treatments to prevent these infections. Metabolic modeling allows us to make predictions from limited data by applying biological knowledge to reduce the need for learning parameter values. Community models designed around genome-scale metabolic models (GEMs) can use that structure to make predictions without perfect parameterization, and GEMs can also be used to estimate parameters for simpler models of population dynamics.

I am currently working with Prof. Erida Gjini to use metabolic modeling to parameterize a model for microbial invasion in order to generate predictions of microbial engraftment.