μBialSim: Constraint-Based Dynamic Simulation of Complex Microbiomes.

Denny Popp,Florian Centler

doi:10.3389/fbioe.2020.00574

Abstract

Microbial communities are pervasive in the natural environment, associated with many hosts, and of increasing importance in biotechnological applications. The complexity of these microbial systems makes the underlying mechanisms driving their dynamics difficult to identify. While experimental meta-OMICS techniques are routinely applied to record the inventory and activity of microbiomes over time, it remains difficult to obtain quantitative predictions based on such data. Mechanistic, quantitative mathematical modeling approaches hold the promise to both provide predictive power and shed light on cause-effect relationships driving these dynamic systems. We introduce μbialSim (pronounced “microbial sim”), a dynamic Flux-Balance-Analysis-based (dFBA) numerical simulator which is able to predict the time course in terms of composition and activity of microbiomes containing 100s of species in batch or chemostat mode. Activity of individual species is simulated by using separate FBA models which have access to a common pool of compounds, allowing for metabolite exchange. A novel augmented forward Euler method ensures numerical accuracy by temporarily reducing the time step size when compound concentrations decrease rapidly due to high compound affinities and/or the presence of many consuming species. We present three exemplary applications of μbialSim: a batch culture of a hydrogenotrophic archaeon, a syntrophic methanogenic biculture, and a 773-species human gut microbiome which exhibits a complex and dynamic pattern of metabolite exchange. Focusing on metabolite exchange as the main interaction type, μbialSim allows for the mechanistic simulation of microbiomes at their natural complexity. Simulated trajectories can be used to contextualize experimental meta-OMICS data and to derive hypotheses on cause-effect relationships driving community dynamics based on scenario simulations. μbialSim is implemented in Matlab and relies on the COBRA Toolbox or CellNetAnalyzer for FBA calculations. The source code is available under the GNU General Public License v3.0 at https://git.ufz.de/UMBSysBio/microbialsim.

Highlights

Microbial communities are ubiquitous in nature, thriving in diverse habitats ranging from the deep subsurface (Dutta et al, 2018) over digestive tracts of higher animals (Gould et al, 2018) to the upper troposphere (Deleon-Rodriguez et al, 2013)
The mathematical modeling of microbiomes holds the promise to move from observation to a more quantitative understanding of microbiome dynamics and underlying mechanisms (Song et al, 2014; Widder et al, 2016; Bosi et al, 2017; Succurro and Ebenhöh, 2018)
Simulations using Matlab’s ODE solver ode15s and the augmented forward Euler method lead to identical results (Figure 2)

Summary

Introduction

Microbial communities are ubiquitous in nature, thriving in diverse habitats ranging from the deep subsurface (Dutta et al, 2018) over digestive tracts of higher animals (Gould et al, 2018) to the upper troposphere (Deleon-Rodriguez et al, 2013). They are selforganizing entities which both modulate the environment they are embedded in, as well as their own constituents in terms of abundance of individual member populations. Most analyses based on such data remain observational in nature and cannot be used to derive quantitative predictions. The mathematical modeling of microbiomes holds the promise to move from observation to a more quantitative understanding of microbiome dynamics and underlying mechanisms (Song et al, 2014; Widder et al, 2016; Bosi et al, 2017; Succurro and Ebenhöh, 2018)

Methods

Results

Conclusion