Microbial Simulation Environment

The Microbial Simulation Environment (MSE) is a MATLAB toolbox for the simulation of genome-scale metabolic models (GEMs) in the natural environment. It was designed to relax the static parameterizations of cell physiology typical of standard flux balance analysis (FBA) simulations, by allowing for cellular acclimation processes (nutrient transport, photoacclimation, macromolecular allocation) which are important under environmental conditions which depart from 'optimal' laboratory growth conditions.

The impact of physiological acclimation on fitness can be substantial. Below is a histogram of the relative increase in growth rates due to acclimation (see MSE Pipeline section below) for a collection of 69 strains of a cyanobacterial genus Prochlorococcus along a meridional section of the North and South Atlantic Oceans:

Citation

Coming soon (with any luck)

MSE Pipeline

This distribution includes a complete dataset for you to try out, since there is quite a bit of fairly specialized experimental data that are needed. MSE is compatible with the PanGEM Toolbox, which generates strain-specific GEMs from a pangenome. The MSE pipeline is implemented in several sequential steps:

1. Load a strain GEM.
2. Parse environmental parameters. These currently include temperature, light (spectral downwelling irradiance), and nutrients (ammonia, nitrite, nitrate, phosphate).
3. Specification of the GEM. Assigns initial physiological parameters from experimental data collected during exponential growth in 'optimal' laboratory growth conditions (i.e., nutrient replete, optimal temperature, optimal light). These data include cell size, macromolecular compositions, metabolomics, DNA and RNA base stoichiometry, maximum photosynthetic rates, transporter abundances from quantitative proteomics, growth and non-growth associated maintenance ATP requirements, and pigment concentrations.
4. Specification of constraints. Constraints include upper and lower bounds on cell size, macromolecular compositions, and pigment concentrations and pigment ratios. Upper bounds on cell size set upper bounds on nutrient transporter abundances, based on available membrane surface area.
5. Nutrient transport optimization (OptTrans). Adjusts membrane transporter abundance and cell size to maximize growth rate. Based on the model of Casey and Follows, 2020.
6. Photoacclimation (PhysOpt) - optimally adjusts pigment content to maximize growth rate, based on the in situ downwelling irradiance spectrum and the absorption spectra of each light-harvesting or photoprotective pigment, subject to constraints on their concentrations and photosystem ratios. Macromolecular allocation (also part of PhysOpt). optimally adjusts the macromolecular composition of biomass to maximize growth rate. The relative abundance of protein, lipid, carbohydrate, RNA, cell wall, pigments, and several intracellular dissolved metabolite pools are allowed to vary within some experimentally defined bounds, subject to the substrate uptake rate constraints determined in the nutrient transport acclimation step.
7. $l_{1}$-norm FBA. Solves the FBA problem (max $c^{T}v$, s.t. $Sv=0$, $v^{lb}\leq v < v^{ub}$), stores the function value $f$, and then minimizes the euclidian of the flux vector $v$ subject to $Sv=0$, $c^{T}v = f$ and $v^{lb}\leq v < v^{ub}$.
8. Adjusts growth rate and fluxes for temperature effects. A parameterization of growth rate dependence on temperature is applied, specifying the optimal growth temperature predicted by a machine learning algorithm called TOME which requires only the translated genome sequence.

Additionally, there are quite a number of post-run clean-up steps, sanity checks and various analyses included to help you make sense of the output.

Installation

Dependencies: RAVEN 2.0*, Mosek 9.

1. Clone the repository to a local directory (using your SSH key): git clone [email protected]:jrcasey/mse.git
2. Add to MATLAB path: addpath(genpath(path/to/mse/));
3. MSE is typically run in batch. Start by specifying paths and other options with the script toolbox/wrappers/BatchSetup.m
4. The outer wrapper for running �MSE is the script toolbox/wrappers/mse.m. This wrapper is called by the shell script job2.sh, which is in turn called by the SLURM array loop Job_loop.sh. OR, if you would just like to run a batch locally, you can edit the sbatch --array line in Job_loop.sh to store the variable SLURM_ARRAY_TASK_ID, which is read by mse.m. OR, if you just want to run a single simulation, use the function /toolbox/wrappers/runMSE.m

*You will need to sign up for a free academic license for Mosek. Not painless, unfortunately. If you plan to run mse_AMT on a cluster, you will need to install on the server, place the license in the Mosek directory, and make sure the solver is working by running the included test test/mosekdiag.m. Also, please be sure to check for any licensing requirements with Mosek before installing on a server.

Please send me your feedback!