This project combines the graph structure of metabolic model, and the learning abilities of neural networks to model the behavior of the metabolic system of bacteria (ecoli or putida). The graph structure to create the neural model comes from the genome scale metabolic model. The data used to learn can be generated by flux balance analysis, using the Cobra python library, or comes from experiments.
The Artificial Metabolic Networks (AMN) is a general framework proposed by Léon Faure, Bastien Mollet, Wolfram Liebermeister and Jean-Loup Faulon in the article A neural-mechanistic hybrid approach improving the predictive power of genome-scale metabolic models - 2022.01.09.475487v3.full.pdf, with the corresponding repository. Here we refactor the model AMNWt of this works and explores other model also using the graph structure of metabolic model.
The exploratory notebook train evaluate model could be an easy start to run the code for the first start.
-
Clone the git (how to clone a git repository)
-
Install a distribution of conda if not already installed (how to install conda)
-
Import the environment
environment.yaml(stored at the root of the repository) with the following command:
conda env create -n <local-env-name> --file environment.yml