PATO quality node page on next.monarchinitiative.org shows NSAID to pathway associations

[kevinschaper]

This is odd, but maybe fine? Does this look like something odd in phenio @matentzn or is it expected? (I'm assuming the d2p and g2p associations are expected)

http://next.monarchinitiative.org/node/PATO:0000001?associations=biolink:ChemicalToPathwayAssociation

[kevinschaper]

Here's a snippet of the api response that populates this table:

      "subject_namespace": "CHEBI",
      "subject_category": "biolink:PhenotypicQuality",
      "subject_closure": [
        "CHEBI:50906",
        "CHEBI:35842",
        "CHEBI:52217",
        "CHEBI:33232",
        "CHEBI:67079",
        "BFO:0000002",
        "BFO:0000017",
        "CHEBI:35472",
        "BFO:0000020",
        "CHEBI:35475",
        "CHEBI:23888",
        "BFO:0000001",
        "PATO:0000001",
        "BFO:0000023"
      ],
      "subject_label": "non-steroidal anti-inflammatory drug",

[matentzn]

Nope this makes no sense at all (the fact that this drug shows up as a quality). This should be moved to the PHENIO repo and fixed asap by Harry; most likely, PATO quality has a faulty top level wiring from one of the participating ontologies.

[caufieldjh]

I will transfer to PHENIO and check on this

[caufieldjh]

PATO:0000001 has a root in BFO in PHENIO:

OK, that's fine.

But here's the path from CHEBI:35475 to the PATO root in PHENIO:

Getting suspicious about that BFO <-> CHEBI connection at role, and also not clear on how that class gets assigned a subclass of the PATO root (yet)

[caufieldjh]

It's very odd that this is (as per the Monarch API response) the only drug that's a PhenotypicQuality, since there are other CHEBI terms in the same neighborhood, but here's one clue: others, like CHEBI:65265, may have biological role in addition to application:

[kevinschaper]

It might be that it's the only drug with an association that we show.

Also, I knew this issue would need those visualizations from oak but I didn't know how to make them, @caufieldjh, could you add the commands you used to generate them if you still have that handy?

[caufieldjh]

What's even stranger is that in KG-Phenio, CHEBI:35475 is biolink:Drug and CHEBI:65265 has a category of biolink:biological_role_mixin, so they clearly undergo different category assignment but neither of them is PhenotypicQuality.

Are these derived from PHENIO or KG-Phenio, and which version?

[caufieldjh]

Also, I knew this issue would need those visualizations from oak but I didn't know how to make them, @caufieldjh, could you add the commands you used to generate them if you still have that handy?

Oh yeah, something like the following works:

runoak -i sqlite:obo:phenio viz PATO:0000001 CHEBI:65265 -p i,p --no-view -o phenio_pato_antacid.png

Add -S your_custom_style.json to modify the style. You can base if off the default in src/oaklib/conf/obograph-style.json

[kevinschaper]

Are these derived from PHENIO or KG-Phenio, and which version?

I believe it's 20230822, since this is my build from 2023-08-24, but I can't be 100% confident that there isn't some kind of timezone shenanigans going on that would have let me get a 20230825 phenio into a 2023-08-24 monarch-kg. (I mean, probably not, unless you've moved your build process to New Zealand 🤔 )

-- edit to add, it would be great to capture version info for sources, maybe even just as a node/edge property

[cmungall]

I skimmed. I think there are two things going on here:

• pato:quality == chebi:role. Ignore the labels. Pretend it's called "property" or "attribute". Or make a cob PR
• chebi has two hierarchies, things and roles/properties. Their naming conventions are poor. There are many things in the role branch that are named like they are things but refer to a role

So this is just funny names AFAICT

[matentzn]

@cmungall What should we do practically here? should we merge chebi role and pato quality explicitly in phenio? That seems a bit.. strong?

[caufieldjh]

The original link doesn't work but this does: https://monarchinitiative.org/PATO:0000001#chemical-to-pathway