Model balancing

Model balancing is a computational method to determine plausible kinetic constants and metabolic states in kinetic metabolic models. It integrates flux, metabolite, protein, and kinetic constant data, using prior distributions for all these variables, and computes the joint posterior mode. Model balancing can be run in matlab or python. Data tables can be provided in SBtab format, models can be provided in SBML or SBtab format.

Matlab

To run model balancing in matlab, please adjust the demo scripts in matlab/demo to your model and data. For help, please type help model-balancing in matlab, or see the documentation at www.metabolic-economics.de/model-balancing/matlab-doc/.

Python

For using model balancing in python, please refer to our Read The Docs page, and the code itself is found on our GitLab repository

Example models

Example model and data files can be found in the folder resource-data. The folder resource-data/models contains files for four different example models. Each model comes as an SBML (.xml) and an SBtab (.tsv) file. The folders examples/results/[MODEL_NAME] contains balancing results for a large number of model balancing problems with these models: artificial data for all 4 models, experimental data for the E coli model (folder e_coli_noor_2016).

How to balance your own model

To balance your own model, please have a look at the matlab demo script demo_cmb_experimental_data.m. You need to prepare three input files in SBtab format, describing

1. the network structure and metabolite constraints (tables Reaction, Compound, Position (optional), Parameter (optional))
2. kinetic data (table ParameterData)
3. state data (tables MetabolicFluxData, MetaboliteConcentrationData, EnzymeConcentrationData)

For an example, please have a look at the files artificial_network_true.tsv, artificial_kinetic_data.tsv, and artificial_state_data.tsv in the folder ./resourcedir/models/branch_point_model/data/.

To balance your model in matlab, you can then use the code from demo_cmb_experimental_data.m. Please note that there are many more settings that you can change. For an overview, type help cmb_default_options in matlab.

To balance your model in python, you first need to convert your model and data into a JSON file. This can be done in matlab and is shown in demo_cmb_experimental_data.m. For an overview of the JSON format, type help cvxpy_problem_data_structure in matlab.

Installation

The following installation instructions are only for the Matlab version (also see the file INSTALLATION in this directory). For Python, please refer to our Read The Docs page.

• SBML toolbox (optional - needed only if SBML files are used; version needs to match your matlab version and requires a matching version of libSBML)
• Clone the following GitHub repositories
  • matlab-utils - utility functions
  • metabolic-network-toolbox - metabolic network toolbox
  • sbtab-matlab - SBtab toolbox
  • enzyme-cost-minimization - enzyme cost minimization toolbox
• Make sure all the directories and subdirectories are included in your Matlab path.

The code was tested with Matlab R2019b on Linux.

License

This package is released under the GNU General Public License.

Contact

Please contact Wolfram Liebermeister and Elad Noor with any questions or comments.

References

Liebermeister W. and Noor E. (2021), Model balancing: in search of consistent metabolic states and in-vivo kinetic constants bioRxiv doi:10.1101/2019.12.23.887166v2

www.metabolic-economics.de/model-balancing/