This repository contains MATLAB and Python scripts for metabolic network analysis and performing Flux Balance Analysis (FBA) along with example metabolic models and experimental flux data.
The repository is hosted privately at BitBucket and mirrored publicly at GitHub.
.m files: MATLAB script files. May be opened/executed with MATLAB or GNU Octave.
.mat files: MATLAB data files. To open a .mat file in GNU Octave, use the load command: "load "
.py files: Python script files. Most/some files are written for python 2.7 and must be updated for use with Python 3.
The MATLAB scripts were developed for a half-semester project at the Norwegian University of Scienc and Technology (NTNU) in fall 2012. The scripts can be used to compare the results of Flux Balance Analysis (FBA) with experimentally determined flux data, calculate the euclidean distance between experimental and computed fluxes and determine which model parameters (combinations of objective function and constraints) which are most consistent with the experimental data. The MATLAB scripts will not be developed further, as they are superceded by the Python scripts.
The collection of Python scripts expand on the functionality of the MATLAB files, and also include implementations of several non-related methods previously only available as MATLAB scripts or not available as open source code. The scripts were written starting in in Fall 2013. Most scripts requires the free FBA package CobraPy and a linear optimization solver supported by that package, such as Gurobi. NumPy and/or SciPy are also required for most of the scripts.
A description of some of the scripts and functions follows below:
optreqanalysis.py: Used to evaluate the minimal achievable distance between a FBA solution and a set of experimental fluxes, for various objective-optimality requirements with respect to the FBA solution. Can process several models and conditions (objectives, constraints) in one run. Receives as first argument a Python dictionary containing one or several CobraPy models, along with a reaction map for each model that specifies which reactions in the model correspond to which reactions in the experimental data. The third argument is a list of experiments for which experimental data exists, and for which the computations should be made. The second argument is a list of objective functions which should be used.
Loads one or several reaction maps relating reactions and flux values in an experimental data set to appropriate reactions in the model(s) to be used.
Dependencies:
QPmindist.py
Reads from:
expdata.mat
QPmindist.py:
Uses quadratic programming in conjunction with flux balance analysis (FBA) to determine the minimal euclidean distance between a set of flux values and the solution of an FBA problem, for a given requirement of objective-optimality for the FBA solution.
Dependencies:
compdist.py (test block only)
extractflux.py
Gurobipy
Cobrapy
Reads from:
reactionmaps.mat (test block only)
Compdist.py:
Loads flux values from stored experimental data and compares with a supplied flux vector (the first argument). Returns the euclidean distance between the two vectors.
Dependencies:
Experimental data.
loadData.py
extractflux.py
constrainfluxes.py:
The function constrainfluxes takes as input a COBRA model and constrains one or more reactions with a de ned corresponding experimental reaction to limits specified by the experimental value and by a tolerance relative to the experimental uncertainty.
Which fluxes to constrain is specified by the optional field "fluxconstrainset", which is a list containing the reactions to be constrained. By default, all fluxes for which experimental data is available are constrained. Default relative tolerance is 1.
Input:
*A CobraPy model
*Experimental fluxes
*experimental errors
*reaction map
*error tolerance (optional)
*fluxconstrainset (optional)
Output:
*Modified CobraPy model
extractflux.py: The function extractflux takes as input a flux vector from FBA result and returns a subset of those reactions as a list or dictionary with fluxes corresponding to the experimental reactions.
Used by:
QPmindist.
Convertrxmaps:
Loads legacy reaction maps from a .mat file.
exportreactionIDs.py:
Writes all reaction IDs in a CobraPy model to a text file.
Fluxreport.py:
The function fluxreport takes as input a vector of fluxes corresponding to the set of experimental reaction rates, and returns both sets of fluxes, the difference between each corresponding flux, the uncertainty in the experimental fluxes and the difference between the fluxes divided by that uncertainty.
loadData.py:
Loads flux values from stored experimental data and reaction maps which define which experimental flux values correspond to which model reactions for specific models.
Dependencies: "xml" package (included with Python)
For the scripts comparing experimental and computed flux values, the following data must be supplied :
Experimental data: Experimental flux values must be supplied in .xml format. See expdata.xml for example data.
Reaction maps:
Reaction maps define which experimental flux values correspond to which model reactions for specific models, pairing publications and models. Reaction maps are loaded from .txt/ or .xml files by the script loadData.py. See reactionmaps.xml for an example file.
SBML Models
The following models are included in the repository in the SBML directory:
In addition to the changes described for each model, for those experimental reactions described in the original project report which had several corresponding model reactions, token metabolites and reactions were added so that the ?ux through each ’token’ reaction would give the sum of the model reactions. This was done to allow those sums to be used directly during optimization
"SCHUETZR.xml" ("SCHUETZ Revised"): A revised version of the metabolic model of E. coli central carbon metabolism used by Schuetz et al.
The .xml ?le containing the model by Schuetz et al. was edited by by changing metabolite and reaction identi?ers to the format expected by the COBRA toolbox, to allow the model to be processed by the program.
In addition, identical duplicate reactions were removed to simplify the computational work and reduce the potential for ?ux loops in the solutions.
"SCHUETZR_notokens.xml" SCHUETZR model without token reactions.
"ECME.xml" ("E. Coli Core Expanded"): Expanded version of E. coli core model downloaded from http: //systemsbiology.ucsd.edu/Downloads/EcoliCore. The E. coli core model was edited to add reactions present in the SCHUETZR model and the experimental data set but missing in that model.
"iJO1366.xml" and "iJO1366b.xml": Original E. coli genome-scale metabolic model iJO1366 used by Jorth et al., and a version modified by the addition of token reactions. Being so changed it is referred to as iJO1366b.
Citations: Robert Schuetz, Lars Kuepfer, and Uwe Sauer. Systematic evaluation of objective functions for predicting intracellular ?uxes in Escherichia coli. Molecular Systems Biology, 3(119), 2007. Jeffrey D Orth, Tom M Conrad, Jessica Na, et al. A comprehensive genome-scale reconstruction of Escherichia coli metabolism—2011. Molecular Systems Biology, 7(535), 2011.
Please refer to the project report in Report directory for further details.