search

NCATSTranslator / minihackathons

MIT License
5 stars 5 forks source link

Discuss/Decide whether to combine queries C.2 & C.3 #173

Closed vgardner-renci closed 3 years ago

vgardner-renci commented 3 years ago

In the July 29 hackathon, this was discussed. Creating issue to pin it down.

rtroper commented 3 years ago

See issue #34 for a schematic showing the three separate queries for workflow C. Below is a revised schematic illustrating how queries 2 and 3 could be combined.

image

The three dashed lines, above, represent published research "evidence" links between the original disease and three hypothetical results within the chemical substance results. Technically, the actual query structure is better represented below.

image

tursynay commented 3 years ago

No real movement, @rtroper to follow up with team

rtroper commented 3 years ago

Given issues we're having, generally, getting results back for query C.3 (two curie-specified nodes and related_to for the predicate), I suggest we consider not adding query C.3 to the workflow. Instead, I propose relying on normalized google distance edges (and future similar edges that come from overlay operations) for published research evidence. For example, see below:

image

image

Having these kinds of edges with PubMed links to research evidence is, I think, a nice feature of Translator that we should showcase.

rtroper commented 3 years ago

After further discussion, we plan to drop query C.3 (as described in the comment above) and rely on overlay edges (e.g. normalized google distance edges currently provided by expander agent). This provides links to published research results supporting the associations.