Enhance the polish module

[GwennyGit]

This issue was opened to collect all enhancements for the polish module.

• Change Regex for biomass to first result of biomass.test_biomass_presence
• Modifiy modules to work with models not using BiGG identifiers #36
• Add EC numbers -> Function in SBOAnnotator
• Add functionality to add more identifiers to GeneProducts in polish  #53
• Add functionality to remove NaNs if found in annotations
• Add checks for BiGG identifiers added via the VMH identifiers
• InChI string annotation format broken after refineGEMs.polish #89
• Connect the polish module to the bioregistry API #91
• polish_carvme does not work as expected #8
• Implement ''progress bar'' for slow tasks #10
• polish_carveme.py for package #24
• Option polish_carveme doubles rdf descriptions for sbml:annotation tags #39
• polish needs to define default units and compartment structures #38
• Feature request: Module handling URI patterns #50
• Change add_reac to handle growth|_*biomass\d*_* IDs properly

[GwennyGit]

I just detected that currently the function add_reac in the polish module does not handle the Growth or Biomass reaction properly. This is due to the fact that current_id is shortened - which is necessary for all other reactions - like that: current_id = current_id[2:]. However, if the ID of the reaction is for example 'Growth' the result for current_id will be 'owth'. Thus, the function cannot handle these IDs with the regex 'growth|_*biomass\d*_*' properly. The code needs to be adjusted to handle these exceptions correctly.

[famosab]

At least in my models the growth function has the id id="R_Growth" which would mean that removing the trailing R is still valid and would not break the regex function. We might need to insert a check whether the trailing R is really there before we try to remove it. Or am I missing something?

        # Get ID and remove 'R_'
        current_id = entity.getId()
        if current_id[:2] == 'R_':
            current_id = current_id[2:]

[GwennyGit]

Regarding the ID for growth, this seems to vary between models a lot. We could also just change the regex to include R_ → R_?growth|R_?_*biomass\d*_*. However, the proposed change is also good and might be better as other reactions could also exist without R_.
Looking at your changes I realised one problem with the function change_all_qualifiers. If one has a lab strain model the GeneProtein qualifier should be set to isHomologTo as we discussed before.

[GwennyGit]

As mentioned in these comments #52 (comment) and #52 (comment). It would indeed make sense to add compartment specifications for all reactions which happen in the same compartment within polish.

[draeger]

COBRApy manipulates IDs back and forth. This prefix is only there in the SBML export. Upon import R_ is stripped away.

[famosab]

In libsbml the trailing R is not stripped though. That is sometimes confusing. (The same is true for genes and metabolites).

[draeger]

Of course not. This entire ID manipulation is nonsense and causes chaos. There should be separate fields for every piece of information that the BiGG ID ships.

[GwennyGit]

I think the R_ for reactions is used to clarify that this is a reaction ID within the model. R_ is not part of the BiGG IDs same as M_ for metabolites/species and G_ for groups and GeneProducts.

[famosab]

But for the polishing module we mostly use libsbml, I think we only even use cobrapy for the growth simulations (and a few more small things). So we always need to take care of those trailing IDs (but through string slicing that is not an issue I think). One could argue that they are obsolete since the entity itself holds the information already - but that is something we cannot change here (or shouldn't on our own).

[GwennyGit]

While investigating models from Heinken et al.[1] I realised that improving the function get_curie_set within the polish module could be interesting. As these models contain NaN identifiers and the function get_curie_set assembles a dictionary mapping the database of each CURIE to the correspondingly found identifier it would be helpful to also check for NaNs and remove the according entries.

---

[1]
Heinken, A., Hertel, J., Acharya, G. et al. Genome-scale metabolic re-
construction of 7,302 human microorganisms for personalized medicine.
Nature Biotechnology. issn: 1546-1696. doi:10.1038/s41587-022-01628-0 (Jan. 2023).