support proteins field that match genes

[edkerk]

Description of the issue:

COBRA has protein and proteinNames fields that match the genes field in size (which means that each gene can be associated to one protein). These are read from the SBML file from the fbc_geneProduct.fbc_name tag that is currently not considered by RAVEN. It would be convenient to support such a protein field as well.

proteins is an identifier field, which means that (if the field exists) it cannot be empty. COBRA solves this by filling it with COBRAprotein1 etc. if no value was found in the SBML file. proteinNames has then the same content, but without filling empty entries with COBRAprotein1 etc. As the proteins field is currently not used for anything (COBRA nor RAVEN), there is little reason to support this at the moment, while it is mostly redundant in comparison with proteinNames. Therefore, only the proteinNames field will be implemented. To ensure cross-compatibility, we will stick to the same field names.

I hereby confirm that I have:

• Checked that a similar issue does not already exist or is already discussed.

[IVANDOMENZAIN]

In PR #559, functions importModel and ravenCobraWrapper return a MATLAB model structure that includes the field proteinNames as a row vector, however, all vector fields are column vectors in RAVEN. Should be corrected for consistency.

[IVANDOMENZAIN]

I've imported the same .xml model file, containing information stored in the SBML field fbc:geneProduct.fbc:name, using both readCbModel from the latest release, and also importModel in RAVEN (updated devel branch 04cdcc9). COBRA returns a MATLAB structure with this information in the field .proteins and does not add proteinNamesfield.
In contrast, RAVEN returns this information in the field .proteinNames and does not add .proteins field to the model structure. This discrepancy causes that, when an SBML model imported to MATLAB with COBRA is then transformed to RAVEN (using the ravenCobraWrapper function) this information gets lost.

[edkerk]

Thanks for reporting, then I misinterpreted COBRA, I now see that they indeed never add the .proteinNames field to model. With this in mind, let's also use the proteins field in RAVEN. ravenCobraWrapper can then add/remove COBRAProteinXXX entries when converting model variants.

[IVANDOMENZAIN]

@edkerk I agree with the suggested approach, will fix as new PR.