CL is asserting that the OBI:0000070 'assay' class is an annotation property

[jamesaoverton]

One of OBI's most important terms is the OBI:0000070 'assay' class.

In two of CL's component files, 'assay' is being declared as an owl:AnnotationProperty:

• https://github.com/obophenotype/cell-ontology/blob/master/src/ontology/components/bgo-cl-comp.owl#L85
• https://github.com/obophenotype/cell-ontology/blob/master/src/ontology/components/wmbo-cl-comp.owl

These were introduced in PRs #3464 and #3169.

This is a very bad form of axiom injection, because OWL treats classes and annotation properties is completely different ways.

[jamesaoverton]

@gouttegd tracked this down to a DOSDP template that asserts 'assay' as an annotation property: https://github.com/Cellular-Semantics/whole_mouse_brain_ontology/blob/main/src/patterns/dosdp-patterns/taxonomy_class.yaml#L37

[balhoff]

Just want to note that CoPilot describes this as "illegal OWL punning" in its PR description. However it is technically perfectly fine to pun a class and an annotation property in OWL. But of course you might not want to.

[matentzn]

Just to be clear where the fix needs to go, not #3579, but here (as @jamesaoverton points out above)

• URGENT: Please get rid of the punning that makes OBI_0000070 an annotation property! Cellular-Semantics/whole_mouse_brain_ontology#121
• URGENT: Please get rid of the punning that makes OBI_0000070 an annotation property! Cellular-Semantics/hmba_basal_ganglia_ontology#3

[matentzn]

here is my sledge hammer thread: #3580

does not need to be merged but if all else fails within a week...

[shawntanzk]

Requestor: Multiomics-Analytics-Group/prefer_ontology#137

PS I like how its one of the things that's coming from BICAN which I worked on before that is screwing up one of the things I am working on now, ahhh full circle :p

[dosumis]

We have quite a bit of class-AP punning. While, as Jim points out, this is legal OWL, some libs dont like it and it may be against the spirit of OBO in at least some cases (e.g. here) Happy to sub this for an actual property. Any suggestions?

[jamesaoverton]

@dosumis I think you would be very unhappy if another OBO project was punning CL:0000000 'cell' in this way, preventing other projects from importing your terms.

some libs dont like it

We're talking about how OWLAPI handles this punning.

it may be against the spirit of OBO in at least some cases (e.g. here)

Axiom injection is perfectly legal OWL, but we've been working very hard to stop doing that in OBO, because it makes integration difficult or impossible.

Happy to sub this for an actual property. Any suggestions?

I can't tell what you're trying to say with these axioms, so I don't know what other property to suggest. RO has a number of "evidence" properties. Or you could mint a new property that means exactly what you intend. If that requires discussion, you could remove these problematic axioms in the meantime.

[dosumis]

@jamesaoverton - sorry. My last post was too glib.

Agree with you on axiom injection in general. I haven't thought of this type of punning in that way but probably should have - esp if it causes OWL-API problems.

I am already working on editing these axioms to remove the use of the assay term. We are due a new release so will be fixed within the next few days.