A microbial community growth model for dynamic phenotype predictions

Abstract

1AbstractMicrobial communities are increasingly recognized as key drivers in animal health, agricultural productivity, industrial operations, and ecological systems. The abundance of chemical interactions in these complex communities, however, can complicate or evade experimental studies, which hinders basic understanding and limits efforts to rationally design communities for applications in the aforementioned fields. Numerous computational approaches have been proposed to deduce these metabolic interactions – notably including flux balance analysis (FBA) and systems of ordinary differential equations (ODEs) – yet, these methods either fail to capture the dynamic phenotype expression of community members or lack the abstractions required to fit or explain the diverse experimental omics data that can be acquired today.We therefore developed a dynamic model (CommPhitting) that deduces phenotype abundances and growth kinetics for each community member, concurrent with metabolic concentrations, by coupling flux profiles for each phenotype with experimental growth and -omics data of the community. These data are captured as variables and coefficients within a mixed integer linear optimization problem (MILP) designed to represent the associated biological processes. This problem finds the globally optimized fit to all experimental data of a trial, thereby most accurately computing aspects of the community: (1) species and phenotype abundances over time; (2) a linearized growth kinetic constant for each phenotype; and (3) metabolite concentrations over time. We exemplify CommPhitting by applying it to study batch growth of an idealized two-member community of the model organisms (Escherichia coliandPseudomonas flourescens) that exhibits cross-feeding in maltose media. Measurements of this community from our accompanying experimental studies – including total biomass, species biomass, and metabolite abundances over time – were parameterized into a CommPhitting simulation. The resultant kinetics constants and biomass proportions for each member phenotype would be difficult to ascertain experimentally, yet are important for understanding community responses to environmental perturbations and therefore engineering applications: e.g. for bioproduction. We believe that CommPhitting – which is generalized for a diversity of data types and formats, and is further available and amply documented as a Python API – will augment basic understanding of microbial communities and will accelerate the engineering of synthetic communities for diverse applications in medicine, agriculture, industry, and ecology.