Why is this one scored 5.0?

NCATSTranslator / Feedback

A repo for tracking gaps in Translator data and finding ways to fill them.

7 stars 0 forks source link

Why is this one scored 5.0? #783

Closed sstemann closed 3 months ago

sstemann commented 4 months ago

https://ui.test.transltr.io/main/results?l=Maturity-onset%20Diabetes%20Of%20The%20Young&i=MONDO:0018911&t=0&r=0&q=f8de3997-f479-49c7-b9ed-b15dc9db9ae4

I have no filters on the UI, but substance "Name Unknown" (https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL511749/) with 2 pieces of evidence has a 5.0?

Looks like it's coming from Improving Agent

cbizon commented 3 months ago

I reran this one now that NN is updated and I no longer see NameUnknown. @sstemann do you agree?

sierra-moxon commented 3 months ago

from relay: we think this is fixed, removing octopus tags and review tags. Mark puts Nameunkown as a default. NN does not return a name, none of the ARAs return it, just return the CURIE.

MarkDWilliams commented 3 months ago

The NameUnknown problem should be resolved by this PR https://github.com/NCATSTranslator/Relay/pull/625

sstemann commented 3 months ago

while the NN issue is resolved - I'm still unclear about why this particular result is a 5.0 and ranked 2. It is only returned by Improving Agent, so I would have thought confidence/sugeno would have taken into account the other results that are shared by multiple ARAs and ranked this below that.

@sharatisrani

sstemann commented 3 months ago

this result is gone from Test and Prod so hard to investigate