How to check the mass_balance when BiGG metabolites have multiple formulas

[dtmvandenberg]

Dear Moritz and Matthias,

Do you know how to ensure that reaction.check_mass_balance() does not give false positives when the metabolites come from the BiGG database and have multiple formulas, like f.i. phosphate (pi) has as formula HPO4;HO4P.
I have made a model using CarveMe and checked the metabolites and mass balance of reactions as such:

Kind regards,

Dennis

[Midnighter]

Hmm, I had actually not noticed this before and it's certainly not SBML compliant. This will also present a problem for memote.io which just looks at the formula directly. The only thing I can immediately think of, is to pre-process metabolite formulas and pick one of them. You could also get fancy and let OpenBabel/RDKit/ChemAxon give you a formula at specific pH but that seems like overkill.

[cdiener]

I think this is a bug in CarveME. @cdanielmachado is that intentional?

[dtmvandenberg]

Thanks for the info, I will try to write a script to do this automatically then and post it at the CarveMe issue. Maybe @cdanielmachado can then use it to curate his universe files.

[cdanielmachado]

I initially decided not to exclude any information obtained from BiGG, including the multiple formulae and charges per metabolite. In retrospective, this was a bad decision. I am currently working on a new release that gets the formula and charge from metanetx instead. Not sure if it is more reliable than BiGG but at least the values are unique.

[dtmvandenberg]

Ok than I will just curate them manually in my model instead of trying to do it generally for all metabolites as that will probably become a spaghetti of if statements ;)

[Midnighter]

@cdanielmachado one of the major inputs for MetaNetX compounds is ChEBI. So I hope the information is good but I have no solid info on it. As part of the work for equilibrator, I've written a pipeline that parses the tables from MetaNetX and puts the information in a database structure (see https://github.com/opencobra/cobra-component-models/ for the models). Maybe that can jump start your work?

[cdanielmachado]

Thanks, I will take a look.