search

obophenotype / ABA_Uberon

Ontologized versions of the ABA structureGraphs - with mappings to Uberon
0 stars 2 forks source link

unsats #16

Open shawntanzk opened 2 years ago

shawntanzk commented 2 years ago

Looks like unsats are a bit more complicated will try using robot explain to figure out some stuff tagging @dosumis

1) no unsats in uberon_slice.owl 2) no unsats in sources_merged.owl 3) unsats when merged the above two 

sh run.sh robot merge --input uberon_slice.owl --input sources_merged.owl --output tmp.owl
sh run.sh robot explain -i tmp.owl -M unsatisfiability --unsatisfiable random:10 --explanation unsats.md

olfactory tubercle SubClassOf Nothing

uncinate fasciculus SubClassOf Nothing

taenia tectum SubClassOf Nothing

posterior commissure SubClassOf Nothing

islands of Calleja SubClassOf Nothing

scattered islands of Calleja SubClassOf Nothing

optic tract SubClassOf Nothing

major island of Calleja SubClassOf Nothing

olfactory bulb SubClassOf Nothing

indusium griseum SubClassOf Nothing

Axiom Impact

Axioms used 9 times

Axioms used 7 times

Axioms used 4 times

Axioms used 3 times

Axioms used 2 times

Axioms used 1 times

shawntanzk commented 2 years ago

wierdly enough when i run sh run.sh robot explain --input tmp.owl --reasoner ELK --axiom "hindbrain SubClassOf 'part of' some brainstem" --explanation hindbrain-brainstem.md theres no explanation, not sure where we found that, cause i cant replicate it with aba_uberon too

shawntanzk commented 2 years ago

uncinate fasciculus unsat fixed in uberon

See details here [uncinate fasciculus](http://purl.obolibrary.org/obo/UBERON_0003044) SubClassOf [part_of](http://purl.obolibrary.org/obo/BFO_0000050) some [pontine tegmentum](http://purl.obolibrary.org/obo/UBERON_0003023) seems wrong Uncinate fasciculus definition - "Pathway arising from the fastigial nucleus to the vestibular nuclei, reticular formation, motor neurons of brainstem and cervical spinal cord" but has "part of" pontine tegmentum - which has [part of](https://www.ebi.ac.uk/ols/ontologies/uberon/properties?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FBFO_0000050) some [pons](https://www.ebi.ac.uk/ols/ontologies/uberon/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FUBERON_0000988) seems like a legit uberon issue, not sure why this didnt get picked up with uberon? does uberon not have the disjoint [Nothing](http://www.w3.org/2002/07/owl#Nothing) EquivalentTo ([part_of](http://purl.obolibrary.org/obo/BFO_0000050) some [forebrain](http://purl.obolibrary.org/obo/UBERON_0001890)) and ([part_of](http://purl.obolibrary.org/obo/BFO_0000050) some [brainstem](http://purl.obolibrary.org/obo/UBERON_0002298))?
shawntanzk commented 2 years ago

posterior commissure SubClassOf Nothing

error coming from uberon and dhba classifying it differently:

DHBA: posterior commissure SubClassOf part_of some white matter of midbrain

UBERON: posterior commissure SubClassOf part_of some diencephalic white matter

On research, hard to say which is actually correct given the commissure connects midbrain and diencephalon structures

shawntanzk commented 2 years ago

Olfactory bulb:

I think this is the issue: white matter SubClassOf part_of some neocortex

I'm guessing that there is white matter in the allocortex too

but also this seems wrong: olfactory bulb, left SubClassOf part_of some olfactory nerve, left

dosumis commented 2 years ago

Thinking about this - these inconsistencies must be coming from a clash between the partonomy coming from JSON and spatial disjointness axioms in Uberon expressed above as

Nothing EquivalentTo (part_of some forebrain) and (part_of some brainstem) [uberon.owl]

(I have a vague recollection of having been through this last time I worked on this. Or maybe I was using a version of Uberon that lacked the spatial disjointness.)

I think we need to (a) strip out disjoints for now so that we can generate accurate reports for review. (b) Slowly chip away at the reasons for these inconsistencies. They are, I think, likely to indicate that some relationships in ABA JSON files should not be mapped to part_of.

We could also use the approach we developed for HubMap - using UberGraph queries to test the validity and type of each relationship. This would be a lot more work though, so we'd need to consider carefully if we want to take it on.

Note - for ease of reviewing, maybe we should use longer display names for the ABA terms by including species in brackets. The inconsistency reports above are confusing because it's hard to tell ABA and Uberon terms apart - many have the same names.

dosumis commented 2 years ago

These are good examples I think

white matter SubClassOf part_of some neocortex

I'm guessing that there is white matter in the allocortex too

-> overlaps , or we make the term PBA_294022044 'white matter of neocortex'

but also this seems wrong: olfactory bulb, left SubClassOf part_of some olfactory nerve, left

part_of -> connected_to ?

shawntanzk commented 2 years ago

(a) strip out disjoints for now so that we can generate accurate reports for review.

will do this for now

(b) Slowly chip away at the reasons for these inconsistencies. They are, I think, likely to indicate that some relationships in ABA JSON files should not be mapped to part_of.

This sound a bit too fun and things we can spend too many hours puzzle-solving :p

dosumis commented 2 years ago

Another good example

uncinate fasciculus SubClassOf part_of some pontine tegmentum seems wrong

Uncinate fasciculus definition - "Pathway arising from the fastigial nucleus to the vestibular nuclei, reticular formation, motor neurons of brainstem and cervical spinal cord" but has "part of" pontine tegmentum - which has part of some pons

seems like a legit uberon issue, not sure why this didnt get picked up with uberon? does uberon not have the disjoint Nothing EquivalentTo (part_of some forebrain) and (part_of some brainstem)?

-> uncinate fasciculus SubClassOf [overlaps]() some pontine tegmentum

Also we should add the spatial disjointness axiom to Uberon if it's missing.

dosumis commented 2 years ago

These kinds of issues are inevitable in a hierarchy built for annotating an atlas.