search

NCATSTranslator / Feedback

A repo for tracking gaps in Translator data and finding ways to fill them.
7 stars 0 forks source link

MVP2 - Unsecret #903

Open sstemann opened 1 month ago

sstemann commented 1 month ago

I was testing #762 and noticed that Unsecret has results in the ARAX GUI which dont seem to be filter-able in the UI so not sure they made it?

https://ui.test.transltr.io/main/results?l=Lepirudin&i=CHEBI:142437&t=4&r=0&q=32af1bc0-75ce-41c1-8cd8-da94829642cc

image

dnsmith124 commented 1 month ago

@gprice1129 any insight here?

gprice1129 commented 1 month ago

Looking into it

gprice1129 commented 1 month ago

I reached out to @kaiwenho @cbizon and @webyrd on Slack. But also posting this here:

When we ingest the analysis node bindings to determine the anchor points to use when generating paths for a graph, we first try to use the query_id, if not then we use the id. Once these anchor points are set, if we can't find a path between the anchors in a graph, we disregard the graph. The reason we implemented this behavior was due to how ARAGORN returned analysis graphs. IIRC their graphs included the query_id as an endpoint instead of the id.

The reason no Unsecret answers show up is because the query_id is supplied, but the analysis graphs use the id field for the endpoint instead. So the UI anchors the endpoint on the query_id but none of the Unsecret graphs have this CURIE.

This seems unresolvable from our side unless we implement a special case for either ARAGORN or Unsecret (which I am not a fan of doing). Can we standardize the relationship between the node binding IDs and the graphs?

gprice1129 commented 6 hours ago

It looks like the query_id special case is no longer needed. This will be removed in Guppy and will resolve this issue.