search

cdanielmachado / carveme

CarveMe: genome-scale metabolic model reconstruction
Other
145 stars 49 forks source link

Reactions without GPR in carveme 1.5.0 #120

Closed dotPiano closed 3 years ago

dotPiano commented 3 years ago

Hi Daniel,

Thanks for the last release, I see many improvements that significantly reduce the need for manual postprocessing.

However, the release also added several reactions without GPRs, which carveme seem to be able to add to a model "for free". An example is ITP:oxaloacetate carboxy-lyase (ITPOXAL). I see this added to several models before gapfilling, even if there is no genomic evidence for it.

This doesn't seem to be carveme's fault directly, the issue is rather in the model that is missing the GPR. Skimming through the universe model it seems like this is not the only case. Do you have an idea about how to deal with this?

Cheers, Mattia

cdanielmachado commented 3 years ago

Hi Mattia,

CarveMe always produces gapless models, i.e, any pathway is either fully included or excluded, depending on the amount of genetic evidence for the respective reactions (finding this fine balance is one of our main features).

This is context-independent and is different from "traditional" (context-dependent) medium-specific gap-filling (i.e., add a minimum number of reactions to enable growth in a given medium).

Therefore, there will always be a fraction of reactions without GPR in a model, even without medium-specific gap-filling. To be honest, I don't think this is something that can be "solved". It will slowly get better as better genome annotations come along.

dotPiano commented 3 years ago

Hi Daniel,

Thanks! Just for clarification: when you say "any pathway is either fully included or excluded" do you mean that carveme internally has the notion of a pathway and tries to make a complete one if only some of the reactions are present? Reading the original paper I understood that this is not the case. Did I misunderstand/did things change meanwhile?

cdanielmachado commented 3 years ago

You're correct. CarveMe does not use any textbook definition of pathway. The carving MILP problem requires that every reaction can carry flux. Based on this constraint, it simultaneously maximizes the presence of reactions with genetic evidence and minimizes the presence of reactions without genetic evidence. So whenever there is a "topological" pathway with mixed cases, it will either include or exclude all reactions.

dotPiano commented 3 years ago

Thanks a lot for the clarifications!