search

monarch-initiative / mondo

Mondo Disease Ontology
http://obofoundry.org/ontology/mondo
Creative Commons Attribution 4.0 International
232 stars 53 forks source link

Logical incompatibilities between MONDO and HPO #356

Closed balhoff closed 3 years ago

balhoff commented 6 years ago

I'd like to use MONDO and HPO together with a reasoner. This doesn't seem to be possible at the moment. Here's an example:

Adenocarcinoma of the colon EquivalentTo Nothing

Not sure if this is more a problem of HPO or of MONDO. But HPO is treating carcinoma as a structure, and MONDO is treating it as a disease. Perhaps HPO just needs a different relation to connect carcinoma to a location.

cc @cmungall @pnrobinson

pnrobinson commented 6 years ago

The HPO and the MONDO terms for an adenocarcinoma are not referring to the same thing, and so the relation shoudl be

MONDO ----has_phenotype--->HPO

even though the term labels may be the same or similar!

-Peter

Peter Robinson Professor and Donald A. Roux Chair, Genomics and Computational Biology The Jackson Laboratory for Genomic Medicine 860.837.2095 t | peter.robinson@jax.org | https://robinsongroup.github.io/ Peter Robinson

From: Jim Balhoff notifications@github.com Sent: Tuesday, October 2, 2018 2:20 PM To: monarch-initiative/mondo Cc: Peter Robinson; Mention Subject: [monarch-initiative/mondo] Logical incompatibilities between MONDO and HPO (#356)

I'd like to use MONDO and HPO together with a reasoner. This doesn't seem to be possible at the moment. Here's an example:

Adenocarcinoma of the colonhttp://purl.obolibrary.org/obo/HP_0040276 EquivalentTo Nothinghttp://www.w3.org/2002/07/owl#Nothing

Not sure if this is more a problem of HPO or of MONDO. But HPO is treating carcinoma as a structure, and MONDO is treating it as a disease. Perhaps HPO just needs a different relation to connect carcinoma to a location.

cc @cmungallhttps://github.com/cmungall @pnrobinsonhttps://github.com/pnrobinson

- You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/monarch-initiative/mondo/issues/356, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AEtuPFyIzNqKpC2v5AwqFaT5mb0Dn0jQks5ug65ngaJpZM4XEmU-.

The information in this email, including attachments, may be confidential and is intended solely for the addressee(s). If you believe you received this email by mistake, please notify the sender by return email as soon as possible.

balhoff commented 6 years ago

I see that, however HPO also uses in its logical definition the NCIT Adenocarcinoma, saying it's part of the colon. And MONDO is equating NCIT Carcinoma (superclass of NCIT Adenocarcinoma) to MONDO carcinoma. So maybe the primary issue is that HPO and MONDO are taking different interpretations of what NCIT is representing.

cmungall commented 6 years ago

Correct, HPO is effectively injecting an inference that NCIT:Adenocarcinoma (and other neoplastic diseases) is a material entity.

This is neither right nor wrong, because NCIT undercommits to BFO

/ NCIT:C7057 ! Disease, Disorder or Finding is_a NCIT:C2991 ! Disease or Disorder is_a NCIT:C3262 ! Neoplasm is_a NCIT:C4741 ! Neoplasm by Morphology is_a NCIT:C3709 ! Epithelial Neoplasm is_a NCIT:C2916 ! Carcinoma is_a NCIT:C2852 ! Adenocarcinoma ***

"disease or disorder" is intentionally left vague.

Similarly, MONDO effectively injects an inference that NCIT:Adenocarcinoma is a disposition/state, which is disjoint from material entity.

We really need to resolve this.

As an aside, when users see things like "Adenocarcinoma has_phenotype adenocarcinoma" they get justifiably confused. At the very least we need consistent labeling to indicate the different meanings of things. I have been suffixing "(disease)" on labels in MONDO that are duplicative of HPO, but this is not ideal.

Possible solutions (not all mutually exclusive):

  1. Treat NCIT 'neoplasm' and subclasses as MEs. This would have implications for MONDO since we either would have a different kind of axiom relating NCIT to MONDO, or the ontological commitment would propagate to MONDO

  2. Ban nested post-composition with NCIT in HPO. There is no need for it, NCIT is already heavily pre-composed, it has adenocarcinoma of colon already

  3. Use MONDO DPs in nested post-compositions with NCIT in HPO (e.g. disease_has_phenotype). This is decidedly odd here as you are using a disease to define a phenotype.

  4. Create a new ontology that is explicitly ME, shadowing NCIT, and use that in HPO logical defs. Users will be delighted with this [sarcasm], as we will have HPO:adenocarcinoma, MONDO:adenocarcinoma, NCIT:adenocarcinoma, PathoMaterialEntity:adenocarcinoma

  5. Remove all neoplasm terms from HPO, and allow use of either NCIT or MONDO in annotation

  6. Follow NCIT and explicitly defer commitment, perhaps using Stefan S's unionOf strategy. This has many issues though.

[btw, love the markdown explanation, can we hook this into ROBOT?]

On 2 Oct 2018, at 14:47, Jim Balhoff wrote:

I see that, however HPO also uses in its logical definition the NCIT Adenocarcinoma, saying it's part of the colon. And MONDO is equating NCIT Carcinoma (superclass of NCIT Adenocarcinoma) to MONDO carcinoma. So maybe the primary issue is that HPO and MONDO are taking different interpretations of what NCIT is representing.

-- You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub: https://github.com/monarch-initiative/mondo/issues/356#issuecomment-426388144

pnrobinson commented 6 years ago

Hi Chris,

consider that for HPO,

Adenocarcinoma is the actual tumor, i.e., a material entity. For MONDO,

Adenocarcinoma is a disease characterized by this tumor as well as potentially other manifestations such as weight loss and pain, and by an etiology (mutations), a time course (progression through stages) and a response to treatment. This can reasonably be modeled as a disposition.

I would argue that the current state of modelling is at least a good start on what we need to do!

balhoff commented 6 years ago

[btw, love the markdown explanation, can we hook this into ROBOT?]

robot explain: https://github.com/ontodev/robot/pull/354

I need to tweak the options and get it merged!

cmungall commented 5 years ago

See #461 for a potential solution

nicolevasilevsky commented 3 years ago

@balhoff can this be closed?

balhoff commented 3 years ago

Yes! Excellent work. :-)