Functionality to add to COBREXA

[stelmo]

Here are some potential features/improvements to add:

• Flux coupling analysis
• Eigenreaction analysis (this is great for GPUs!)
• Model reduction, e.g. combine reactions that are stoichiometrically directly coupled
• dynamic flux balance analysis, I have reservations about this since dFBA is tricky once you start thinking about it
• Data integration (very broad)
• Improve model reading performance, especially json and sbml
• Community visualization, e.g. which exchanges are connected to which organisms in the community?
• Model exploration functions, e.g. something that simplifies findall(x -> occursin("biomass", x), reactions(model)) for example. This is most useful if you loaded a model but know nothing about it (e.g. an automatically generated one). Another example, find all the reactions directly linked to metabolite X etc.
• Convergence estimation for sampling - tricky borderline impossible with high dimensionality, will require some thought 🤔
• Growth balance analysis
• Moment + Moma combo?

Anything else @laurentheirendt @htpusa @exaexa?

[htpusa]

Elementary Conversion Modes could be interesting. Originally from [1] but more recently in [2]. See especially the discussion about scaling up the method in the latter.

1. Urbanczik, R., & Wagner, C. (2005). Functional stoichiometric analysis of metabolic networks. Bioinformatics, 21(22), 4176-4180.
2. Clement, T. J., Baalhuis, E. B., Teusink, B., Bruggeman, F. J., Planqué, R., & de Groot, D. H. (2021). Unlocking Elementary Conversion Modes: ecmtool unveils all capabilities of metabolic networks. Patterns, 2(1), 100177.

[exaexa]

if I get it correctly the ECMs are basically eigenvectors in the optimum space that is cut down to exchange reactions only, right?

[laurentheirendt]

The following should be there I think:

dFBA: https://pubmed.ncbi.nlm.nih.gov/12202358/
MOMA: minimization of metabolic adjustment

[stelmo]

So with respect to dFBA, we should implement this algorithm: https://doi.org/10.1002/bit.24748
However, I have serious doubts about dFBA approaches in general, but it is very popular so I guess we should add it :/

[exaexa]

I added some minor ideas (esp. the gene-moma from yesterday, can we call it momament?)

Question 1: "Flux coupling" is the same thing as in CoreModelCoupled right? (Can't we make a "wrapper" coupler model that could work like CoupledModel{StandardModel}? )

Question 2: Eigenreactions is really just finding the eigenvalues right?

[exaexa]

Nice data here: https://www.embopress.org/doi/full/10.15252/msb.20209536

[htpusa]

1. Yes. And yes, I agree that something other than a separate type would make sense.
   Although what one really needs is just adding inequality constraints which could just be a generic functionality.

2. "Eigenreactions" comes from here. It's basically an interpretation of the singular value decomposition of the stoichiometric matrix.

[exaexa]

o nice thanks!