Handling ontology versions used in an ARC

[Brilator]

Hey all, I'm opening the discussion in this repo as it's about ontologies. However, it's probably more relevant, for tools / implementations.

To my knowledge, the top-most section (ONTOLOGY SOURCE REFERENCE) of an ARC's isa.investigation.xlsx is currently not tackled by any tool.

The section could look like something this:

ONTOLOGY SOURCE REFERENCE
Term Source Name	PECO	UO
Term Source File	http://purl.obolibrary.org/obo/peco.owl	http://purl.obolibrary.org/obo/uo.owl
Term Source Version	releases/2020-08-21	releases/2020-03-10
Term Source Description	Plant Experimental Conditions Ontology	Unit ontology

It summarizes the use of different ontologies across the ARC, i.e. within any of the isa.* files (including isa.investigation.xlsx and the isa_study and isa_assay sheets, respectively).

Now, referencing the correct ontology can possibly become tricky (and I didn't try to find out how ISA originally handles this).
Ontology versions may be diverging

• annotating an ARC via Swate or ARCitect with some time in between, in which ontologies were changed in Swate DB
• reusing an assay from another ARC annotated with another ontology version

[Brilator]

One solution might be to use a URL to the ontology term that includes the ontology version
e.g. not just http://purl.obolibrary.org/obo/EO_0007231

Unfortunately, trying to find an example, I run into a lot of dead links in ontobee and OLS...

[Brilator]

@Freymaurer
@HLWeil
@JonasLukasczyk
@kdumschott

[kdumschott]

@Hannah-Doerpholz

[Freymaurer]

The section could look like something this:

We could collect all TSR from all term accross all ISA files and auto complete this section. The necessary information is written in SwateDB, BUT i am not sure how to fix dead links and ontology versioning.

Maybe this is the point at which we host our own ontology search + versioning.

[kdumschott]

It may be a better idea to reference the github repo containing the ontology file and not just the ontology file itself, since ontology versioning results in dead links. Not sure how this would work automatically though.

[Brilator]

This assumes, all ontologies come from the gh repo.
You might however not use any of the tools for annotation (e.g. do it by hand) and / or ingest data (e.g. assays) from other sources.

[Brilator]

Plus, you would still run into version issues.
Good example from @eik-dahms: what if a taxonomy (Species, Organism) changes? We'd have to update all existing ontology annotations in outdated ARCs.

[eik-dahms]

At least for ncbi they seem to track everything they change so you would get something like this:

https://www.ncbi.nlm.nih.gov/taxonomy/12

Which is fine because you can find the current id.

[HLWeil]

I don't think versioning necessarily leads to dead links. Even the contrary might be the case (as Dominik and Eik stated).

In principle, versioned ontologies should have releases whose URIs must never change. By this, pointing to an ontology term in a specific version of an ontology should be permanently working.

The thing is, we could implement this, but how do other ontologies handle this?

[HLWeil]

I see one problem here in though:

Currently, in the Annotation Tables of Assay and Study Files, we use full links. This somehow clashes with the whole Ontology Source References and versioning concept, as for the mapping we would need to deconstruct the link, which feels a bit counterintuitive.