has_skeleton - add domain and range

[dosumis]

has_skeleton - RO:0002551 should have:
domain: UBERON:0000475 ! organism subdivision
range: UBERON:0010912 ! subdivision of skeleton

See obophenotype/uberon#2453 (comment)

[cmungall]

uberon query -q "SELECT * FROM edge WHERE predicate='RO:0002551' AND object NOT IN (SELECT subject FROM entailed_edge WHERE predicate='rdfs:subClassOf' AND object='UBERON:0010912')"

A few exceptions:

subject	subject_label	predicate	predicate_label	object	object_label
UBERON:0000376	hindlimb stylopod	RO:0002551	has skeleton	UBERON:0015052	femur endochondral element
UBERON:0003822	forelimb stylopod	RO:0002551	has skeleton	UBERON:0015053	humerus endochondral element
UBERON:0005473	sacral region	RO:0002551	has skeleton	UBERON:0006075	sacral region of vertebral column
UBERON:0006071	caudal region	RO:0002551	has skeleton	UBERON:0006076	caudal region of vertebral column
UBERON:0015875	heel	RO:0002551	has skeleton	UBERON:0001450	calcaneus

the femur/humerus axioms look odd but they reflect the fact we have avoided naming subdivisions where there is only a single bone. We could revisit this decision but if we add a hindlimb stylopod subdivision of skeleton we should make it clear that curators should avoid this term for annotation and instead use the appropriate femur term.

The heel situation is similar. The "heel skeleton" is just the calcaneous (not sure if this holds for all species with a "heel")

The vertebral column ones are a bit odd, they are a consequence of the distinction between skeletal system (includes joints, such as intervertebral joints) and skeleton (no joints).

unfortunately resolving this will take a bit of work. We could choose to have parallel subdivisions (one including joints, one excluding). Or we could choose to center on the most inclusive one - but this would be a huge refactor, and may introduce disconnect with FMA.

[ghost]

Of note- this issue is dependent on the completion of #695 as there is currently no Uberon import to RO. Without the import, the requested domain and range can not fully be added.

[wdduncan]

@bvarner-ebi The issue of whether RO needs to explicitly include classes mentioned as domains/ranges is important. See #701.

[wdduncan]

Does this issue need to be add to the RO agenda for 03/28 or can it be handled off line?

[wdduncan]

Notes from RO call:

• Add definition for relation.
• Rename to make domain/range more clear. E.g.: has skeletal subdivision

[dosumis]

@cmungall - you comment above is a bit worrying as it looks like this causes some problems and we don't have a clear plan to fix. OTOH - this axiom is still in Uberon, so moving here would just continue the status quo.

Suggestion for how to proceed: If you are happy to keep this domain, could you add a ticket to Uberon with some details of the edits needed to fix any problems it causes. Otherwise maybe we should drop this ticket and drop the axiom from Uberon?

[gouttegd]

Otherwise maybe we should drop this ticket and drop the axiom from Uberon?

FYI: If we drop the axioms in Uberon (so that we use the existing range and domain restrictions defined in RO), the only inference we lose is the following:

nose SubClassOf: 'organism subdivision'

which is currently (with the Uberon-injected range domain) inferred because

nose SubClassOf: has skeleton some 'nasal skeleton'

This does not seem like a great loss to me. I am in favour of closing here and, in Uberon, removing the injected domain and range constraints for 'has skeleton'.