abnormal and normal phenotype groupings

[ValWood]

Add grouping terms for "normal and abnormal phenotype". Sorry I didn't realise we didn't have these.

(in pathogen host interaction, pathogen only and host only)

@jseager7 is this something you could help with? It might be easier to do in the flat file?

[jseager7]

@ValWood I can take care of adding the groupings. It should be easy enough to template this in with tools like ROBOT.

As long as PHIPO is structured so that normal and abnormal terms have some (not necessarily direct) parent term with 'normal' or 'abnormal' in the term name, it should be possible to automatically query the relevant parent terms and then just make those a subclass of the new normal / abnormal grouping terms.

Otherwise, I might need some help identifying terminology that implies normality or abnormality. For example, 'increased', 'decreased' and 'abolished' all seem to imply abnormality, so they could be used as a fallback if there's no clear parent term.

[jseager7]

Note that my intended solution would add extra hierarchies to PHIPO, probably at the top level, like this:

I think this follows the rector normalisation pattern that Chris Mungall promoted, only without it being automatically derived from logical definitions.

Note that this would aggregate all normal / abnormal phenotypes regardless of whether they're single species or pathogen-host. If we don't want to do that, we could add additional subclasses for 'normal / abnormal single species phenotype' and so on.

[ValWood]

I think it might be better to have only abnormal pathogen host interaction phenotype and abnormal single-species phenotype

to keep these branches fully disjoint.

[ValWood]

Otherwise, I might need some help identifying terminology that implies normality or abnormality. For example, 'increased', 'decreased' and 'abolished' all seem to imply abnormality, so they could be used as a fallback if there's no clear parent term.

This seems like a good start. If any are missed they can be added later. If we see any other phrases specifying normality/abnormality without these terms as parents we can add them to the ticket.

I see that we used to have "abnormal tissue phenotype" and "abnormal cell phenotype" which were obsoleted (I think I assumed the generic parents existed). But this made me wonder if you could find all of their descendants from an old ontology version?)

[jseager7]

Oh, I missed the fact that we already have abnormal and normal grouping terms under single species phenotype, but there's nothing in them at the moment:

There doesn't seem to be any corresponding grouping under pathogen-host interaction phenotypes.

Do you want to have the normal / abnormal groupings as subclasses of the single species / pathogen-host branches (option 1), or do you want them as a separate hierarchy (option 2), like below?

[jseager7]

But this made me wonder if you could find all of their descendants from an old ontology version?)

Probably. I'd just have to reset my local repository to a version before they were obsoleted and then hopefully I'd be able to extract the hierarchy using Protege.

[jseager7]

Do you want to have the normal / abnormal groupings as subclasses of the single species / pathogen-host branches (option 1), or do you want them as a separate hierarchy (option 2), like below?

Having said that, there's no reason why we can't have them in both places at once (e.g. have 'normal single species phenotype' be a subclass of both 'single species phenotype' and 'normal phenotype'). That's probably what would happen once we're able to use logical inference anyway.

[ValWood]

I think option 1. The 2 branches would not be disjoint otherwise. @CuzickA do you agree?

[jseager7]

The 2 branches would not be disjoint otherwise.

I think in OWL we can assert that the branches are disjoint if there's any risk of ambiguity in logical inference, but I'm not sure if that could pose problems for the OBO format.

[CuzickA]

I thought we were trying to move away from normal and abnormal terms? Where will the presence and absence terms fit in?

[ValWood]

That's true sorry, we probably won't have normal or abnormal in the phi branch. It will still exist in the single species branch for things which are clearly abnormal (i.e where a process does not work properly, like "abnormal chromosome segregation").

The best approach might be just to request a config addition for 'severity' to the highest term possible whenever we see the need to use it. So this ticket can close as the terms exist already in single species branch.