part_of consistencies

[mjy]

Some notes from discussion with @teleaslamellatus

We discussed insuring the following "principles":

• part_of relationship should be the finest localization
• If a class is a parent of another through an is_a relationship then the parent can not have a part_of relationship.
• when I have a class, then I should be able to see where it fits in the partonomy

[teleaslamellatus]

I have discussed this with Serg and he agreed in this rule, sort of, he thinks we should actually based our ontology on topological "relationships". So we discussed that basically there are only a very few topological relationships between anatomical structures on the gross morphology level:

1. Bodyparts are defined usually by sclerites and conjunctivae (and adjacent epithelial cells including gland cells, seta related cells)
2. Sclerites are connected via conjunctiva or muscles
3. Sclerites/conjunctivae (area) and muscles are connected to the brain via muscles

so object property relations are:

sclerite/conjunctiva A is attached to muscle AB sclerite/conjunctiva B is attached to muscle AB sclerite/conjunctiva A is connected to sclerite/conjunctiva B by muscle AB

sclerite A is_connected to conjunctiva AB sclerite B is_connected to conjunctiva AB sclerite A is_connected to sclerite B (by conjunctiva AB)

eventually every sclerites are connected to each other by all of the in-between structures (sclerites or conjunctivae or muscles)

sclerite A is articulated by sclerite B via articulation AB (this is defined as an anatomical cluster composed of articulating surfaces of the two sclerites).

So I do have some immediate question to ask here:

Part_of should be an object property applied only for the endpoints (i.e. a seta is the part of a particular sclerite or conjunctiva) and I think are crucial for correctly modeling "fusions" in the cuticle.

Bodyparts are defined by the surrounding sclerites, i.e. bona field boundaries can be observed in most cases only on the cuticle. The internal structures of a bodypart do have usually fiat boundaries (like muscles arising from the tibia are tibial muscles, but only their portion that is in-between the distal and proximal tibial margins are part_of the tibia.

Glands are nothing else but modified epidermal cells that corresponds usually with a somewhat modified cuticular surface.

Nerves are composed of neurons (excitable cells: cells whose cell membrane contains voltage dependent ion channels) and glial cells and attached to the brain and a muscle, gland, or sensilla.

Fore and hindgut (including Malpighian tubules, rectal pads etc. are part of the epithelium, so every structures on that are just like anything on the cuticle.

Midgut is endodermal, part of the endothelium.

Muscles are composed of muscle cells (second type of excitable cells).

so again, we should get together (Serg, Aaron, Matt, me (students) and make this basic ontology within 2 days!

[mjy]

We should get some feedback from @balhoff et al. before leaping towards major refactors I think?

[balhoff]

Not knowing much about insect anatomy, this sounds reasonable to me. :-) But I don't understand @mjy's second bullet above ("If a class is a parent of another through an is_a relationship then the parent can not have a part_of relationship").

In general I think there is a lot of value in staying compatible with Uberon, which is bringing a lot of animal anatomy together in one model. But I don't know whether anything you're proposing would prohibit that (just make sure 'part of' works the same way).

cc @cmungall

[mjy]

@balhoff thanks; the second point was a transcription of an idea @teleaslamellatus was working out, so it's interesting you are confused at it. Perhaps getting more to the point- OBO barked at you when you had redundant part_of relationships for a term, does this really matter in a modern "Uberon" world?

[cmungall]

sorry I may need more context. It seems the issue is whether redundant axioms are OK? Redundancy is locally fine. It's conventional to remove this, e.g. robot reduce

[teleaslamellatus]

Lets use a simple example to explain what our question is with part_of:

femur (bone in human leg) part_of leg subClassOf bone (is_a bone in .obo) leg part_of body

bone part_of body

A while ago the .obo version of HAO have been opened with obo edit which marked part_of relationships for classes that inherits part_of relationships trough is_a (subClassOf) as redundancies. So, it was then decided that these part_of relationships need to be erased (in my example that would be femur part_of leg relationship). So, we do have plenty of classes without part_of relationships.

We would like to fix this because part_of leg is a more precise relationships than part_of body an thus, we should resuscitate it.

We see two solutions for that:

1. recreate these relationships and accept that there are redundancies Q: is it a problem if there is a redundancy in part_of relationships in an ontology?

2. create part_of only for "endpoint classes". "endpoint class" means a class that do not have subClassOf children (like femur in humans). So bone part_of body would be erased and could be later inferred from the part_of relationships of its subClassOf children ( i.e. femur).

Thanks!

István

On Tue, Mar 13, 2018 at 2:03 PM, Chris Mungall notifications@github.com wrote:

sorry I may need more context. It seems the issue is whether redundant axioms are OK? Redundancy is locally fine. It's conventional to remove this, e.g. robot reduce

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/hymao/hao/issues/1#issuecomment-372762253, or mute the thread https://github.com/notifications/unsubscribe-auth/AI1_JW8C_8q-q40d2j47lh8zXCbLSjkyks5teAn1gaJpZM4SnJWD .

-- István Mikó PhD Research Associate Department of Entomology Pennsylvania State University 501 ASI Building University Park, PA 16802 USA

[balhoff]

It's not a problem. You should have 'femur part_of leg' and also 'bone part_of body'. I would not adhere to point 2—there are lots of cases for existential restrictions on superclasses.

[teleaslamellatus]

Great!!!!!

Thanks Jim!

i

On Tue, Mar 13, 2018 at 10:05 PM, Jim Balhoff notifications@github.com wrote:

It's not a problem. You should have 'femur part_of leg' and also 'bone part_of body'. I would not adhere to point 2—there are lots of cases for existential restrictions on superclasses.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/hymao/hao/issues/1#issuecomment-372880271, or mute the thread https://github.com/notifications/unsubscribe-auth/AI1_JeOZD-ajI7tX71rMHoiFb98gv64mks5teHsDgaJpZM4SnJWD .

-- István Mikó PhD Research Associate Department of Entomology Pennsylvania State University 501 ASI Building University Park, PA 16802 USA