search

obophenotype / upheno

The Unified Phenotype Ontology (uPheno) integrates multiple phenotype ontologies into a unified cross-species phenotype ontology.
https://obophenotype.github.io/upheno/
Creative Commons Zero v1.0 Universal
76 stars 17 forks source link

new pattern 'misaligned X' #267

Open sbello opened 6 years ago

sbello commented 6 years ago

We have a few terms in the MP that use the PATO term 'misaligned with'. We may want to add this to our standard patterns. See for example MP:0030541 misaligned incisors equivalence axiom: 'has part' some ('misaligned with' and ('inheres in' some 'incisor tooth') and (towards some 'incisor tooth') and ('has modifier' some abnormal))

pnrobinson commented 6 years ago

One probably minor issue is that (I think) the definition does not imply that one incisor is misaligned with the other incisor, and so I do not think we gain anything by having both "inheres in" and "towards" -- I think it would be essentially the same to simply say ('inheres in' some 'incisor tooth') and not try to define exactly how the incisor is misaligned.

sbello commented 6 years ago

The term in the MP specifies misalignment of the upper and lower incisors so I think the inclusion of 'inheres in' and 'towards' makes sense but we might want to be more specific and change 'incisor tooth' to upper jaw incisor and lower jaw incisor.

Another example of this pattern is strabismus? I currently have this as: 'has part' some ('misaligned with' and ('inheres in' some 'camera-type eye') and (towards some 'camera-type eye') and ('has modifier' some abnormal))

I think having the specification of towards here is important. But maybe I'm missing something?

pnrobinson commented 6 years ago

what I mean is that the logical definition does not imply that the tooth or eye of the "inhere is" is different from the tooth or eye in "towards", and thus it does not provide additional information above and beyond saying simply ('inheres in' some 'camera-type eye') unless I am missing something.

sbello commented 6 years ago

Ah, I read the addition of towards to specify what the first thing was aligned with. So adding towards incisor tooth implies that the incisor alignment is in relation to some incisor and not to say a molar or the jaw in general. Same thing for eye, you're aligning the eyes to each other rather than to some other facial feature.

pnrobinson commented 6 years ago

My understanding is that one "some" knows nothing about the other "some", and so the towards could be related to any other tooth or eye and not necessarily the other one. If one wrote inheres in tooth towards nose it would be better defined...but hopefully Chris or Seb can weigh in here about this? @cmungall @drseb

cmungall commented 6 years ago

I recommend avoiding relational qualities and 'towards' here. They are awkward, and often the direction is arbitrary. We've been successful doing this for representing things like inc/dec distance between eyes using classes like UBERON:0013678 "anatomical line between inner canthi". This may seem like punting the problem, but it makes some sense as the design patterns for representing things spatially already exists in the anatomy ontology.

This is a bit more challenging. If we do want to have OWL equivalence axioms here (and it's not clear this buys us any more than having the classes be subclasses of an axiomatized term like "abnormality of the incisor"; remember the goal is not to have equivalence axioms for everything, only for where it makes sense) then it's important to be accurate, maybe @markengelstad can help us here.

This may sound a bit bonkers or overkill, but if we want to do this right, then I think having a separate sub-ontologies of alignments would help (analogous to and building on the "anatomical line" hierarchy in uberon). These could be very rigorous, and state precisely which two axes/lines are being compared, relative to individual structures (or collective structures like dentition).

Maybe it's better to not have an equivalence axiom and simply say SubClassOf has-part some ('orientation' and inheres_in some <structure> and has-modifier some abnormal)

sbello commented 6 years ago

For current MGI purposes the part of this we really need is the relation to the UBERON anatomy term. @cmungall is there an acceptable way to just specify the anatomical structure without specifying the rest of the equivalence axiom?

cmungall commented 6 years ago

Can you explain the dependency more? Is having an axiom on the superclass ok if it's too the most precise uberon term?

On Fri, May 11, 2018, 07:36 Sue Bello notifications@github.com wrote:

For current MGI purposes the part of this we really need is the relation to the UBERON anatomy term. @cmungall https://github.com/cmungall is there an acceptable way to just specify the anatomical structure without specifying the rest of the equivalence axiom?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/obophenotype/upheno/issues/267#issuecomment-388382763, or mute the thread https://github.com/notifications/unsubscribe-auth/AADGOSjcjY4BaJ_gbzPZnmf8bQzV1htIks5txaH8gaJpZM4T6ePj .

sbello commented 6 years ago

We have implemented links and comparison matrices between gene expression and phenotype data. See http://www.informatics.jax.org/gxd/phenogrid/MGI:109583 The links are currently only made when the MP term has an UBERON term in the logical definition (and the UBERON term has a cross-reference to EMAPA). We don't consider the presence of an UBERON term associated with an MP term's parent since this would not always be an accurate thing to do. This was done to avoid needing complex rules for when we could or could not use the UBERON term associated with the parent. Maybe we could talk more at the next phenotype editors call?