Candidates for DKG extension

[bgyori]

While examining relevant models and data sets, I tried to identify specific entries or subsets of ontologies that could be included in the DKG. NCIT is generally very useful and relevant, and contains (sometimes seemingly exclusively) terms that are needed. But I would say is too large and diverse to include as a whole. Below I'm using OLS link for tree browsing.

NCIT

• [ ] Population group subtree (NCIT:C17005) of NCIT. Includes entries like "At-risk Population", and also terms like "Age Group" or "Race" that are important for stratification. So I would say this subtree is useful as a whole for the epi use case.
• [ ] Rate subtree of NCIT (NCIT_C25636) This is a relatively small subtree with some irrelevant entries but containing "Case Fatality Rate" and "Infection Fatality Rate" so those two terms of the whole subtree could be added.
• [ ] Booster from NCIT (NCIT_C28320) This is an important term, in the "Vaccine Therapy" subtree but that subtree actually just has irrelevant entries other than this one so this single node could be added.
• [ ] COVID-19 Infection from NCIT (NCIT_C171133) This is useful and has a couple of useful subterms. It lives in the Coronavirus Infection->Viral Infection->Infectious Disorder subtree all of which are potentially relevant, though they are relatively large subtrees.
• [ ] Dead (NCIT_C28554), Hospitalized (NCIT_C25179), Quarantine (NCIT_C71902), Discharged Alive (NCIT_C154475) from NCIT. For these terms the surrounding subtrees are not really relevant so it might make sense to including them one by one.
• [ ] Intensive Care Unit (NCIT_C53511) Here we could possibly include the subtree of the parent term Healthcare Facility (NCIT_C21541) since there are other potentially relevant entries:

Not NCIT

• [ ] Uninfected from EFO This seems to be the only place this term is available. Possibly we could include the “disposition” subtree from EFO though this is very large, contains many diseases: https://www.ebi.ac.uk/ols/ontologies/efo/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FBFO_0000016&lang=en&viewMode=All&siblings=false
• [ ] Confirmed case from COVOC This is a supposedly COVID-specific vocabulary but is a bit strange and inconsistent, and doesn't contain that many useful terms.
• [x] An ontology that is probably relevant to include is CIDO: https://bioregistry.io/registry/cido, it contains many relevant terms like names of SARS-CoV2 variants, specific vaccines, various host states, and many other things.

[cthoyt]

So the first thing I'm checking into is using the robot extract command to get subtrees out of NCIT (docs: http://robot.obolibrary.org/extract). I'm trying

robot extract \
    -I http://purl.obolibrary.org/obo/ncit.owl \
    --method MIREOT \
    --output /Users/cthoyt/Desktop/test.owl \
    --branch-from-term "obo:NCIT_C17005" \
    --branch-from-term "obo:NCIT_C25636"

right now and it's making a nice subhierarchy. There might be some chain of commands that can add the parents back to the root, but I don't think this is necesassary.

It appears --upper-term doesn't work without also a --lower-term, so don't use these.

Ideally it's also possible to do multiple upper level terms in a simpler commend, too. This should work somehow with --branch-from-terms but I am not sure what the syntax is. For now, multiple --branch-from-term is fine.

The step after this is to add a custom step to the DKG build that lets you specify some local ontology files instead of looking them up via bioregistry prefix.