Leaky metabolites in Xomics contextualized models

[Dana Luong]

Hi,

I've generated models from transcriptomic data from a human cohort using XomicsToModel. 

After manual inspection of the models, I saw that instead of taking up glucose from the environment, all glucose was being produced by the following reaction: 

I identified the reaction r1393 as contributing to a mass imbalanced loop, which I think caused a leak in the metabolite g1p. After closing this reaction, the model took up glucose using the exchange as expected.

Running the checks findStoichConsistentSubset and findThermoConsistentFluxSubset  on the Xomics contextualized model did not identify the reaction r1393 as an issue. Does findStoichConsistentSubset include mass-imbalanced reactions in the output model if they are part of a topologically closed loop, still satisfy S*v=0? How would you suggest handling these cases in the output models?

I also ran the numerical stability tutorial on my model - the model passes as well-scaled by 'checkScaling', but I do notice that there are 6 orders of magnitudes between the largest and smallest stoichiometric coefficients. This range is driven by reaction HMR_5160 (10696.0 h2o[c] in reactants) and biomass_maintenance (0.002914 pglyc_hs[c] in reactants). 

What is the typical acceptable numerical range for the S matrix; could this be causing issues in the Xomics optimization steps?

Thank you in advance for your help!

Dana