Relationships between phenotypes and diseases (General discussion)

[matentzn]

I would like to collect all ways in which diseases and phenotypes can and should be related. It comes up more and more now, and we should decide on a common strategy.

The most general one is diseaseX: 'has phenotype' some phenotypeY. This simply means, I assume, but correct me if I am wrong, that diseaseX is in some unspecified way associated with phenotypeY. So this could be anything from phenotype always/often/sometimes occurs when disease occurs to diseaseX is causally upstream of phenotypeY.

Peter has made some more suggestions on the Mondo tracker, and we have discussed this before in an email thread.

Peter suggests this:

defined_by_primary_finding some HP:001

And I think this relation is critical. ((You know me, I dont care about the labels, but to avoid mingling biology and clinical jargon here, I suggest to use has defining phenotype instead of defined by primary finding.))

@pnrobinson continues to suggest disease considering_during_current_encounter_only_one_aspect some HP:..., which is important for clinical decision making, but not that important for uPheno, which is concerned with cross species inference. However, @pnrobinson you say:

The second option would reflect clinical but not biological reality. The first option would be convenient but would be inaccurate for neurogenic bladder, which often/usually is not the primary finding.

Would this be alleviated by abstracting away from "finding" and instead using "has defining phenotype"? So should these then be separate relationships ("has defining phenotype" and "has primary finding"? Or do you think they are close enough?

@pnrobinson further continues to say:

disease_state_considered_as_phenotypic_feature some MONDO:...

Is there a risk here that we do a lot of double curation? p2d vs d2p? Can we get another argument why we need both?

Please all, can you think of the ways in which phenotype and disease can be connected. What would be useful for your database, portal or analysis use cases? What would be useful for AGR (@sbello), what is useful for Monarch (@cmungall), for HPO (@pnrobinson)?

[nicolevasilevsky]

I would love to see this get implemented, I don't have any strong feelings about the best way to implement this though, I defer to @pnrobinson and @cmungall.

[cmungall]

cc @mbrush

do you want to catalog all of the different ordo relations here?

I wonder if some of these things are not overloading the relation, and should be part of the quantifier (hard to do directly in OWL since they are not all-some)