NTR: supplies blood to; drains blood from

[dosumis]

Make new subproperties of supplies/drains: "supplies blood to" / "drains blood from" with domains 'arterial blood vessel' and 'venous blood vessel'' respectively (these classes cover the microvasculature as well as arteries and veins.).

Background: These will replace usage of supplies and drains in Uberon, where they have been used extensively but with domain and range added locally - illegally restricting the meaning of these relations & being the the potential source of problematic inferences where drains and supplies are used outside of an anatomical context (e.g. for rivers).

[cthoyt]

does this apply to ticks, mosquitos, and/or vampires?

[dosumis]

I guess tempting to 'drains blood from' for feeding ticks, mosquitos and vampires, but domain restricts it to vertebrate(?) veinous vessels (veins, venules)

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 domains can not fully be added.

[wdduncan]

I've submitted #701 to address consistency for domains and/or ranges.

[wdduncan]

Notes for RO Call:

Are just intending to constrain the current RO drains so that it only involves blood? If so, are you okay with syringe also being used with this relation?

Label suggestion: vessel drains blood / vessel supplies blood

[dosumis]

Decision punted to @cmungall :

Either
(a) - Roll back from all this complexity and just promote much more general use of drains and supplies. I just reviewed a bunch of tickets on drains/supplies/channels - I worry that the complexity is unsustainable. There is a very small amount of positive inference in Uberon coming from the domain of each of these terms (a hand full of inferences) but it would be no big deal to add these manually. The downside is losing some consistency checking in Uberon, but are we really that worried about these being used for something dumb? The inconsistencies we have found all look defensible to me if we allow for a more general usage/meaning (e.g draining calyces of the kidney).
(b) Add as requested but with more specific names - 'vessel drains blood' & 'vessel supplies blood'. These were requested in the RO call as without them drains_blood might be applies to medical devices or mosquitos.

[wdduncan]

Any update on this?

[cmungall]

I favor b, with names "vessel for draining blood from" and "vessel for supplying blood to"

or another naming convention that might be more scalable is "(domain) genericRelation (range)" [including the parentheses]

E.g. "(vessel) drains (blood)"

I am wondering if we want to revise the shoreline and just include these in uberon with uberon IDs They are not reusable outside uberon or a descendant of uberon (e.g ZFA). This feels more modular.

[wdduncan]

I like the naming convention (domain) genericRelation (range), but I worry it will be off putting to user.

[wdduncan]

Note: (vessel) drains (blood) does not contain "from" or "to". So, it may not be clear what the intent is.

E.g.:
(vessel) for draining (blood) from
However, if "(blood)" is the domain indicated by this syntactic convention, it obfuscates that there is an organ from which the blood is being drained.

To make this more specific you would something like:
(vessel) for draining blood from (organ)

This seems awkward. Perhaps that may be of use for more general relations.

The sentiment on RO call is to not implement the syntactic convention.

Minor suggestion is to add an "is" to the front of the label:
"is vessel for draining blood from" and "is vessel for supplying blood to"

Another suggestion is to define a general vessel drains relation and then check for consistency using GCIs in UBERON.

[balhoff]

Can Uberon folks provide an example of the consistency checking that would be lost by just using the existing RO drains? There may be a reasonable way to implement the constraints in Uberon using GCIs or rolification.

[cmungall]

I would like to avoid adding more things that are hard to mentally reason about. I think the decision to use specific relations is a good one. We can include a more specific and anatomically relevant text definition, examples, and other metadata. There are many other characteristics that separate rivers from blood vessels.

We do need to address design patterns of genericity and specificity more broadly across RO, but we should do this in another ticket.