Add documentation describing when to add terms to CL vs species-specific ontologies

[cmungall]

Here is a strawperson proposal, we should reach consensus and then add some prominent docs

1. new mammalian cell types should be added directly to CL. Groups annotating mammalian systems should annotate to CL terms
2. for non-mammalian vertebrate cell types in model systems, existing ssAOs should either
   • a import a generic vertebrate CL subset and add species specific subclasses for cell types that are species specific
   • b maintain all cell types in their ssAO and commit to syncing generic vertebrate cell types with CL
   • e.g. XAO:0005097 ! blastomere D1.1.2
3. for non-vertebrate metazoan cell types in model systems, existing ssAOs should maintain all cell types in their ssAO and commit to syncing generic vertebrate cell types with CL
4. for non-metazoan eukaryotic cell types, a kingdom-specific AO should be used (PO, FAO). This ksAO will likely import or sync with the most general CL terms only (cell, native cell, in vitro cell, ...)
5. for metazoan non-model cell types, these can be housed in CL, but groups wishing to manage more than a few terms in CL need to learn CL DPs, attend calls, etc

The most difficult decision here is sync vs import. In the principle of incrementalism, given that the two main non-mammalian certebrate AOs already have generic cell types, the immediate strategy should be to sync, with ongoing conversations about when and how to swap out homologous terms with CL IDs directly. Both strategies have multiple pros and cons from the side of users, ontology maintainers, curators, database maintainers, precision, etc. Tooling for both strategies is also progressing, albeit slowly.

Note the decision about when to include cells derived from clustering of scSeq data is mostly orthogonal to this.

[cmungall]

Regarding option 2b, currently this syncing is not happening. See #2271. You can see the different strata of different CL eras in XAO and ZFA...

[ybradford]

@cmungall I believe we are doing a flavor of a and b above. We have ZFA terms that xref CL terms like: ZFA:0009031 (osteoblast) CL:0000062. In this instance the CL term def has changed compared to what we have in ZFA, but still seems generic enough to be xrefed.

Some terms have changed the textual definition in CL to be non generic and we believe we should drop the xref to CL term: ZFA:0009047 (osteoclast) CL:0000092.

We have terms that are only in ZFA: ZFA:0005929 xantholeucophore

I'm not sure we can import - we are editing in OBO edit - so I think importing is not 'easy'. I believe this is why we are xref-ing to CL instead of importing CL terms properly.

What is XAO doing?

[dosumis]

@ybradford - might be a good time to revisit this. Damien and others have developed new tools new we can take advantage of & we have more editors. I'd like to aim to have all cells in ZFA at least nest under some reasonably specific term in CL and for mappings to be kept up to date as far as possible. With this, we can more reliably handle data annotated with ZFA terms using composite metazoan. This is something we cover in the new CL paper we're working on.

We can potentially keep the mappings in a separate file rather than as xrefs (@gouttegd can detail options).

To start this - could you provide a list of cases where you believe xrefs in ZFA are no longer valid given changes to CL? We could work with you to review these.

[ybradford]

@dosumis we can provide a list, it will take a little time as we have to manually look at the terms. Will post here when we have it.