search

NCATSTranslator / minihackathons

MIT License
5 stars 5 forks source link

What is the correct predicate for the ingest of COHD into RTX? #165

Closed vgardner-renci closed 3 years ago

vgardner-renci commented 3 years ago

https://arax.ncats.io/?source=ARS&id=597d0a89-013c-47cd-a296-ada21af6fa15

image

image

vgardner-renci commented 3 years ago

@dkoslicki @edeutsch @CaseyTa request for update

edeutsch commented 3 years ago

I do not know what this means or the answer to the question..

dkoslicki commented 3 years ago

@vgardner-renci I think this is something to ask the data modeling group. IIRC, COHD uses the correlated_with predicate, but since there are at least three distinct kinds of edges (paired concept frequency, log observed/expected, and chi square p-value), we thought it best to be more specific.

CaseyTa commented 3 years ago

What is the timeline for resolving this issue?

If this is for immediate ingest of COHD, I would suggest working with the correlated_with predicate. Currently, COHD decides if there is an edge between a pair of concepts based on significance of the log observed/expected metric. The other metrics (chi-square, relative frequency) are currently tagging along as additional edge attributes on this edge.

There is more discussion on how to better model the "knowledge" from the clinical KPs in https://github.com/NCATSTranslator/minihackathons/issues/47. In the near future, we will adopt the temporary predicate has real world evidence of association with, which will replace the correlated with predicate while maintaining the same behavior described above. Distinct predicates to differentiate the various metrics available in COHD and other clinical KPs have not yet been established.

tursynay commented 3 years ago

Leave it until what is showing up in RTX is the right predicate

tursynay commented 3 years ago

@edeutsch will follow up

vgardner-renci commented 3 years ago

@edeutsch request for update

tursynay commented 3 years ago

@vgardner-renci to check in with Eric

vgardner-renci commented 3 years ago

@edeutsch can you comment?

tursynay commented 3 years ago

Update from Amy Glen, this is close to completion

amykglen commented 3 years ago

Ok, we now use the predicate biolink:has_real_world_evidence_of_association_with for all COHD edges - example: https://arax.ncats.io/?r=23658

So I think this issue is resolved.