search

NCATSTranslator / testing

Materials and tools for testing Translator components
1 stars 9 forks source link

Disease (HP:0001657) related_to NamedThing related_to Gene (NCBIGene:388591) #130

Open sstemann opened 2 years ago

sstemann commented 2 years ago

We'll be doing this one asynchronously - please take a look at how your tool responded and provide a comment below and Mark or I will be touch Friday afternoon. Tagging all ARAs just to ensure you are notified. Thank you!

Query: D.1_RNF207-long-qt.json PK: ff64b065-33cf-4fc8-ac16-9e46fdda5731 Control: KCNH2 NCBIGene:3757 https://www.sciencedirect.com/science/article/abs/pii/S1547527121022219 - it took a team 10 days to do sequencing to identify a variant in RNF207 and then 5 years to explain how it worked. Results Tracking Sheet

image

MarkDWilliams commented 2 years ago

So, this one came to us by way of Raj, who was one of our external reviewers for the most recent run-through of the demo. The idea started with this tweet https://twitter.com/euanashley/status/1447769055771824130 and ended with this paper https://www.sciencedirect.com/science/article/abs/pii/S1547527121022219

The controls that we were hoping to find is hERG/KCNH2 which does seem to be showing up in the ARAX, Aragorn, and BTE's result sets.

finnagin commented 2 years ago

Looking at ARAX's results it has KCNH2 as the top result which is great, but also the 2nd through 17th result which is probably not ideal. This is due to the descendant conflation where the descendants of the two id specified nodes are being seen as separate results from their ancestor node. This can be fixed by resubmitting the query with is_set: true then it will combine all the descendants into one result like so. In general setting name or id pinned nodes to have is_set: true will combine all the conflated ids for those nodes into 1 result.

cbizon commented 2 years ago

For ARAGORN, we get KCNH2 also as the top result. There are a number of individual results where a subclass of LongQT syndrome has been returned from a KP as having a direct link to KCNH2 as well, and these have been rolled up into the same answer, which is nice.

Room for improvement: because LongQT is a subclass of itself, and because there's a direct link fro RNF207 to LongQT, there are a bunch of answers like (RNF207)-(longQT)-(longQt) that are not very appealing.

sstemann commented 2 years ago

@pg427 @brettasmi @webyrd do you have any feedback on this ticket?