Biolink definitions for clinical edge attributes

[karafecho]

The TCDC has standardized on the following edge attributes for the TCDC Curated Query Service: log_odds_ratio, log_odds_ratio_95_ci, total_sample_size, infores id, URL to wiki page. We now need to define the first three attributes in Biolink Model. I have proposed definitions below. These are adapted from the descriptions provided in https://www.ncbi.nlm.nih.gov/books/NBK431098/, just so we have supporting evidence in whatever definitions we agree upon.

This discussion thread is intended to stimulate comment and reach consensus. Please comment on or before Friday, April 28.

[karafecho]

log_odds_ratio: The logarithm of the odds ratio, or the ratio of the odds of event Y occurring in an exposed group versus the odds of event Y occurring in a non-exposed group.

Updated, more explicit definition, 07.07.2023:

log_odds_ratio: The natural logarithm of the odds ratio, or the ratio of the odds of event Y occurring in an exposed group versus the odds of event Y occurring in a non-exposed group.

[karafecho]

@sierra-moxon : Please see the above updated definition. I can reopen the ticket, if you can help me find it. :-)

[nlharris]

this may be the ticket: biolink/biolink-model#1301

[karafecho]

Thanks, Nomi!

[karafecho]

log_odds_ratio_95_ci: The ninety-five percent confidence range in which the true log odds ratio for the sample population falls.

[CaseyTa]

COHD by default calculates 99% CI to reduce the false-positives rate given the potentially large number of edges. If others are giving 95%, then we provide that also (we'll probably do it in addition to 99% CI if we can do so unambiguously).

[karafecho]

ICEES KG provides 95% CI.

[GitHubbit]

Multiomics EHR Risk is fine with whatever alpha the committee decides on, since we have yet to implement this. That said, I believe the packages we use default to 5%...
-Kamileh

[CaseyTa]

Sounds good, we will comply with 95.

[karafecho]

FWIW, I think it's fine to keep the 99% CI, too. While we agreed on a small set of edge attributes with standardized names, I don't think anything will break by having additional edge attributes.

[karafecho]

total_sample_size: The total number of patients or participants within a sample population.

[karafecho]

Consensus reached, 04.28.2023