search

opencobra / cobrapy

COBRApy is a package for constraint-based modeling of metabolic networks.
http://opencobra.github.io/cobrapy/
GNU General Public License v2.0
467 stars 218 forks source link

FBA/FVA in Cobra (Matlab) vs Cobrapy #15

Closed Rubbert closed 11 years ago

Rubbert commented 11 years ago

hi everyone,

I have noticed a small glitch between the implementations of the cobra toolbox in python and Matlab. After removing several genes, metabolites and reactions from a yeast model ( always taking into account that it should be possible to create biomass ) I am able to get a solution (FBA) in python but not in Matlab.

What I do more specifically: I take a standard yeast model (SBML file) and import it into Matlab. I modify the lower bounds to match my medium and remove all dead end reactions using the standard functions. next, I export this model as an SBML file and import it into python. And then I use this model to determine which reaction should be removed, and remove them using the code below. Next, I perform an FBA on both the model before deleting the reactions, metabolites and genes and after deleting. Finally,I write out an SBML file

# Original model
model = cobra.io.sbml.create_cobra_model_from_sbml_file(os.path.join('data','iMM904_NADcorrected_pappEtAl2011_adj_and_remove.xml'))
model.optimize(solver='glpk')
print 'optimal f full model '+str(model.solution.f)

# Remove
for reaction in removed_reactions: model.reactions.__getattr__(reaction[2:]).remove_from_model(model)
for gene in removed_genes: model.genes.__getattr__(gene).remove_from_model(model)
for metabolite in removed_metabolites: model.metabolites.__getattr__(metabolite[2:]).remove_from_model(model)

# New model calculations
model.optimize(solver='glpk')
print 'optimal f partial model ' + str(model.solution.f)
FVA_result = cobra.flux_analysis.variability.flux_variability_analysis(model, fraction_of_optimum=1.0, number_of_processes=15)

# Write to file
write_cobra_model_to_sbml_file(model, os.path.join('data','iMM904_examo_%s.xml' % description))

Outputs: optimal f full model 0.287998303658 optimal f partial model 0.0635426076692

as you can see both the full model and the model in which I deleted some reactions, metabolites and genes are able to create biomass. The partial model is only able to do it less efficiently.

I import the SBML file in Matlab using the code below:

initCobraToolbox

reduced_model = readCbModel(['C:\Master_CNCZ_new\trunk\CMBI\src\EXAMO_scripts\data\iMM904_examo_WT_MDM34.xml']);
old_model = readCbModel(['C:\Master_CNCZ_new\trunk\CMBI\src\EXAMO_scripts\data\iMM904_NADcorrected_pappEtAl2011_adj_and_remove.xml']);

FBAsolution_old = optimizeCbModel(old_model);
FBAsolution_red = optimizeCbModel(reduced_model);
FBAsolution_old.f
FBAsolution_red.f

Outputs: ans = 0.2880 ans = 0

as you can see the partial model gives the exact same result in both python in Matlab. However, the model in which I deleted some genes, reactions and metabolites is not able to generate a solution in Matlab. If I try to do an FVA, it generates an error in Matlab but works fine in python.

Am I doing something wrong in exporting the SBML file from python, or could there be some other cause?

Any help would be greatly appreciated,

cheers, Rob

here's what I'm using: Matlab R2013a Python 2.7 cobra.version: 0.2.1

Initilization Matlab: Warning: LP solver Gurobi not useable: gurobi.m not in Matlab path

In changeCobraSolver at 138 In initCobraToolbox at 42 In fa_fba_analysis at 1 Warning: LP solver Gurobi not useable: gurobi_mex not in Matlab path In changeCobraSolver at 133 In initCobraToolbox at 42 In fa_fba_analysis at 1 Warning: LP solver CPLEX through Tomlab not usable: tomRun.m not in Matlab path In changeCobraSolver at 113 In initCobraToolbox at 42 In fa_fba_analysis at 1 LP solver set to glpk successful Warning: MILP solver Gurobi not useable: gurobi.m not in Matlab path In changeCobraSolver at 175 In initCobraToolbox at 55 In fa_fba_analysis at 1 Warning: MILP solver Gurobi not useable: gurobi_mex not in Matlab path In changeCobraSolver at 170 In initCobraToolbox at 55 In fa_fba_analysis at 1 Warning: MILP solver CPLEX through Tomlab not usable: tomRun.m not in Matlab path In changeCobraSolver at 160 In initCobraToolbox at 55 In fa_fba_analysis at 1 MILP solver set to glpk successful Warning: QP solver Gurobi not useable: gurobi.m not in Matlab path In changeCobraSolver at 231 In initCobraToolbox at 67 In fa_fba_analysis at 1 Warning: QP solver Gurobi not useable: gurobi_mex not in Matlab path In changeCobraSolver at 224 In initCobraToolbox at 67 In fa_fba_analysis at 1 Warning: QP solver CPLEX through Tomlab not usable: tomRun.m not in Matlab path In changeCobraSolver at 195 In initCobraToolbox at 67 In fa_fba_analysis at 1 Warning: qpng solver has not been fully tested - results may not be correct In changeCobraSolver at 205 In initCobraToolbox at 67 In fa_fba_analysis at 1 QP solver set to qpng successful Warning: MIQP solver gurobi not usable: gurobi.m not in Matlab path In changeCobraSolver at 262 In initCobraToolbox at 78 In fa_fba_analysis at 1 Warning: MIQP solver gurobi not usable: gurobi_mex not in Matlab path In changeCobraSolver at 255 In initCobraToolbox at 78 In fa_fba_analysis at 1 Warning: MIQP solver CPLEX through Tomlab not usable: tomRun.m not in Matlab path In changeCobraSolver at 248 In initCobraToolbox at 78 In fa_fba_analysis at 1 MIQP solver set failed CB map output set to svg successful

danielhyduke commented 11 years ago

Strange behavior. Please post this and the sbml files to the help page. http://groups.google.com/group/cobra-pie