Roadmap for MA identifiers

[mihai-sysbio]

While the current process of keeping track of identifiers has worked until now, a better approach should be investigated. Please contribute!

Facts
Currently, MA identifiers are coined and kept track of in the Human-GEM repository. The identifiers in use are listed in the model files, in particular reactions.tsv and metabolites.tsv, while the deprecated identifiers are moved over in the corresponding files. There are also GH Actions in this repository that check the consistency of the identifier use/reuse.

GEMs templated from Human-GEM also use MA identifiers, which can change.

Concerns
As a user of Metabolic Atlas, I want to easily find deprecated identifiers for my version of my model, so that I can understand what changed across versions.
As a model curator, I want to coin new identifiers easily, so that I can introduce new reactions/metabolites.
As a model curator, I want to deprecate identifiers easily, so that I can implement the curation work.

Opportunities
There have been discussions about how to keep track of merged entities here SysBioChalmers/Human-GEM#615

Vision
It's likely that the optimal solution includes setting up a centralised repository for the identifiers.

[haowang-bioinfo]

now it does appear to a good time point to revisit this MA id issue.

GEMs templated from Human-GEM also use MA identifiers, which can change.

so far Human-GEM and five model animal GEMs have been using the MA ids, whose wider application is expected

It's likely that the optimal solution includes setting up a centralised repository for the identifiers.

agree with the idea of setting up a centralised repository, to which the migration plan for MA ids from Human-GEM may be discussed

[mihai-sysbio]

@haowang-bioinfo do you have any ideas for how to address the concerns put forth? or perhaps other concerns which are not mentioned?

As a user of Metabolic Atlas, I want to easily find deprecated identifiers for my version of my model, so that I can understand what changed across versions.
As a model curator, I want to coin new identifiers easily, so that I can introduce new reactions/metabolites.
As a model curator, I want to deprecate identifiers easily, so that I can implement the curation work.

[haowang-bioinfo]

to move forward, a follow-up plan may begin with considering the migration of existing GH actions surrounding MA ids from HumanGEM to a dedicated and centralised repo

[mihai-sysbio]

to move forward, a follow-up plan may begin with considering the migration of existing GH actions surrounding MA ids from HumanGEM to a dedicated and centralised repo

Indeed, that's one of the things to do. However, it does not solve any of the concerns above. And without a plan to tackle these, I'm not keen on introducing changes in a (somewhat) functional system (that we have now for the identifiers).

[mihai-sysbio]

@smoretti and @mpagni12 you probably have some experience with this, any advice?

[smoretti]

This is not an easy task. In MetaNetX we have a specific file to trace deprecated entries:

• their id
• what they became (merge, split, ...)
• when they changed (at which release)
  A graph view or the SPARQL endpoint help to follow their history.

I think the main point is that deprecated ids must not disappear, to always find what they became, and to avoid to reuse them.

Unfortunately I think there is no specific qualifier in SBML to tag them.
https://github.com/combine-org/combine-specifications/blob/main/specifications/qualifiers.md

[mihai-sysbio]

@edkerk any potential concerns coming from the perspective of yeast-GEM, which is not using MA identifiers?

[edkerk]

No concerns, but I like recognize your concerns and share your vision. It would be preferred to also apply MA identifiers in yeast-GEM, but I have no suggestions on how to exactly setup such a centralised repository.

[mihai-sysbio]

On the same topic, @smoretti has also asked if MetAtlas has a centralized listing of all the identifiers. I believe this would be useful, and doable. @smoretti it would be great if you could expand on what would need to be included, or if the raw list of active (non-deprecated) identifiers is sufficient.

[mihai-sysbio]

There was also a question about compartments. For those, there are no identifiers, but we are using the naming convention agreed upon here SysBioChalmers/Human-GEM#341 . Perhaps that's a new feature request - compartment identifiers?

[smoretti]

For compounds, ideally, we would like to have ids, names, cross-references and structures
+ synonyms, formula, mass and charge if possible.

For reactions, ids, equations and cross-references
+ names, synonyms, EC, reversibility and pathways if possible.