Should we be using species neutral PRO terms?

[jamesamcl]

For example

https://www.ebi.ac.uk/ols4/ontologies/oba/classes/http%253A%252F%252Fpurl.obolibrary.org%252Fobo%252FOBA_2056762

we are using the human PRO term so we end up with "level of protein S100-A8 (human) in cerebrospinal fluid"

But what if OBA gets used in a mouse study - do we make new terms for each species or should we be using the species neutral PRO term?

Important to note that in PRO only the species speciifc terms have uniprot xrefs whcih is a requirement for the GWAS Catalog. So if we do switch to the species neutral ones we will need to inject these as species specific annotations

CC @eks-ebi

[jamesamcl]

@matentzn can we do something similar to what we did in upheno with injecting the species specific annotations perhaps?

[matentzn]

From @cmungall:

This is an important and fundamental Q that touches on many uses and misuses of ontologies, I will write more shortly, but this may require a full paper!

some contextual stuff from uberon:

in an ideal world we would have precise atlases for each species and could point to discrete unambiguous entities. In this world we could just generalize Uberon automatically.
But we are not here yet. The difference between 2 typicaly ssAOs for closely related structures is 99% random arbitrary ontological stuff and 1% biology.

oba related:

In contrast, genes and the proteins they encode are crisp, discrete, very well defined entities, and the relationships between them in species and across species are generally known (or knowable) precisely
In database terms, the species-neutral groupings may not be stable, leading to practical tracking issues when using these as primary identifiers. This is maybe slightly less of a problem if you use PRO species-neutral groupings as a convenience to avoid saying mouse v human, and that may well be the main use case for OBA right now…

I cant exactly deduce a course of action from @cmungall comment, and I think he is also not entirely sure himself. Given this:

In contrast, genes and the proteins they encode are crisp, discrete, very well defined entities, and the relationships between them in species and across species are generally known (or knowable) precisely

I would personally say that perhaps it is best we

1. Add an SOP to OBA to include the species neutral terms alongside any species specific terms for grouping
2. Add species specific proteins to OBA as needed

I am happy to be contradicted though.

[jamesamcl]

This comes from a GWAS catalog request so there are really two questions here:

1. Should we include the species neutral terms in OBA?

2. Should we include the species specific terms in OBA?

If 1 is yes then we can use them in the GWAS Catalog providing we can inject uniprot xrefs. This would give us cleaner names and the potential for cross species integration.

I think 2 becomes irrelevant for the GWAS Catalog if 1 is true, because the species neutral terms are at least as specific as the EFO terms they are replacing which did not specify “(human)”

As far as I know we don’t have any use cases for levels of mouse proteins in humans, so the context the term is used in (the GWAS catalog) tells us the species

[matentzn]

As far as I know we don’t have any use cases for levels of mouse proteins in humans, so the context the term is used in (the GWAS catalog) tells us the species

If that is so, I think we can for now defer the decision. Either way I think we need the species neutral terms in OBA. And at least for PRO, we already have the UNIPROT mappings in there; https://github.com/obophenotype/pro_obo_slim