FAQ


Answers to Frequently Asked Questions (FAQ) are listed below.

If you need support, please feel free to post your question in our ImageLink.

Installation

How may I remove a legacy installation?

If you have an existing installation of a legacy COBRA Toolbox on your system, please remove the installation directory from your MATLAB path.

|warning| The following commands will delete your cobratoolbox directory and all of its contents

>> CBTDIR = fileparts(which('initCobraToolbox.m')); % get the directory of the COBRA Toolbox
>> rmpath(genpath(CBTDIR)); % remove the directory from the path
>> savepath % save the new path
>> delete(CBTDIR,'s') % delete the installation directory

I cannot update the COBRA Toolbox using updateCobraToolbox(). Why?

Whenever the update of the COBRA Toolbox fails when running updateCobraToolbox(), there is a chance that major restructurations happened recently. For instance, new submodules might have been added.

Another reason for a failing update might be when the local version of the COBRA Toolbox is too old. In that case, the easiest is to reinstall (reclone) the COBRA Toolbox or the fork after having backed up the current version.

In case the update fails because of changes in the COBRA Toolbox, please contribute your changes first by following the instructions here <https://opencobra.github.io/cobratoolbox/stable/contributing.html>.

In case you do not want to contribute your changes and are familiar with git, you may also type (beware, your changes will be lost!):

$ git stash # stash all potential changes
$ git add --all  # add all files first to stage
$ git reset --hard HEAD  # hard reset the repository

When running git submodule update, the following error message appears. What should I do?

No submodule mapping found in .gitmodules for path 'external/lusolMex64bit'

remove the cached version of the respective submodule by typing:

$ git rm --cached external/lusolMex64bit

Note: The submodule throwing an error might be different than external/lusolMex64bit, but the command should work with any submodule.

On Windows, MATLAB R2016b crashes with CPLEX 12.7.1. Why?

When you experience an unexpected crash of MATLAB R2016b when running:

>> changeCobraSolver('ibm_cplex')

or

>> initCobraToolbox

after having installed CPLEX 12.7.1, the solver might not be correctly installed (see this issue).

In order to fix this issue, follow these steps:

  • Uninstall all older versions of CPLEX (e.g., 12.6.3)
  • Uninstall CPLEX 12.7.1
  • Restart your computer
  • Install CPLEX 12.7.1. You will be prompted to install Microsoft Visual C++ 2013
  • Download this software package and install Microsoft Visual C++ 2013 (x64)
  • Finish the installation of CPLEX 12.7.1
  • Restart your computer
  • Start MATLAB and the above commands again

On Linux, MATLAB Suddenly crashes without any error

This can happen due to some MATLAB versions shipping broken libraries, in particular libssl.so.1.0.0. To fix this, you will have to replace the matlab library by the system library as follows:

$ sudo mv <MATLAB_ROOT>/bin/glnxa64/libssl.so.1.0.0 <MATLAB_ROOT>/bin/glnxa64/libssl.so.1.0.0.old
$ sudo cp /lib/x86_64-linux-gnu/libssl.so.1.0.0 <MATLAB_ROOT>/bin/glnxa64/libssl.so.1.0.0

where <MATLAB_ROOT> is the directory of your MATLAB installation.

Parallel programming

When running code in a parfor loop, solvers (and other global variables) are not properly set.

This is an issue with global variables and the matlab parallel computing toolbox. Global variables are not passed on to the workers of a parallel pool. To change cobra global settings for a parfor loop, it is necessary to reinitialize the global variables on each worker. The toolbox offers two helper functions for this purpose, which also take care of pathes, getEnvironment() and restoreEnvironment(), which can be used as in the below example.

environment = getEnvironment();
parfor i = 1:2
    restoreEnvironment(environment);
    changeCobraSolver(solver, 'LP', 0, -1); %third argument is printLevel, fourth argument is validation Level.
    % additional code in the parfor loop will now use the currently set solver
    optimizeCbModel(model);
end

By requesting the current environment (global variables and path) before the parfor loop and assigning it to a local variable, that variable is passed on to the workers, which can then use it to set up the environment. dqqMinos and quadMinos use the file system to input and output solutions. Therefore, they can currently not be used in any function that uses parfor, as this would cause concurrency issues between different workers.

Github & Contributing

How may I update a submodule?

When updating a submodule, please consider updating the submodule itself in the opencobra/cobratoolbox repository. Below is an example of how to update the tutorials submodule:

$ cd fork-cobratoolbox # directory of your cloned fork
$ git checkout develop
$ git fetch upstream # upstream must be configured to point to opencobra/cobratoolbox
$ git merge upstream/develop
$ git checkout -b update-submodule
$ cd tutorials
$ git pull origin master # pull the latest changes from the master branch of COBRA.tutorials
$ cd .. # change back to the root
$ git add tutorials
$ git commit -m "Updating the tutorials submodule"
$ git push origin update-submodule

Then, proceed to open the PR to the opencobra/cobratoolbox repository.

What do all these labels on issues and PRs mean?

A comprehensive list of labels and their description for the issues and pull requests is given here.

General

After loading a model, I get errors when using it with toolbox functions. What can I do?

If you used load('filename.mat') to load your model, you may encounter unexpected errors. Please only use readCbModel('filename.mat'). Many models stored in a MATLAB format (.mat) contain outdated data structures, which are no longer compatible with the COBRA Toolbox. The readCbModel() function tries to convert these models to the current format and will inform you whether this was successful or not.

If the readCbModel() call was unsuccessful, please use load again to load your model struct and run verifyModel(model) to determine which fields in the model are problematic. You can then either try to correct the fields, or remove them, if they are not necessary for your analysis.

If this does not solve your problem, feel free to report an issue as described here.