Open TranslatorIssueCreator opened 3 months ago
from improving agent:
the predicate chain is the following:
LAMA3 Regulates Increased Activity Or Abundance Of
CDK6 Has Increased Expression Affected By
Acetylcholine
I do not understand why LAMA3 is the answer to the question while CDK6 is not?
there are in the top ranked a bunch of genes as valid answers where CDK6 in an intermediate node
reviewing the ticket in triage: @sandrine-muller - is LAMA3 a valid answer to this query?
@suihuang-ISB @brettasmi - is this something you could look into? We are curious why there isn't there a lookup returned for CDK6 in the results? wouldn't the direct result score higher than this multi-hop result?
@dnsmith124 @gprice1129 - we also noticed that MAPK in this query result has 282 valid paths. This greatly slows down opening the result details to the point it looks like my interface is broken...it does eventually open though! @bill-baumgartner - just making sure that all these edges should be returned for MAPK in the result set?
I don't think it is a valid answer to the query (IMO). I went to ARAX UI to make sure we were not messing up the predicates some how, and this is just bad logic. It is from improving agent and I don't think it should be an answer that is coming back. I don't know who to tag on that team? @sierra-moxon do you know who to ask about this?
Note as an aside that the composed predicates in the support path Sandrine pasted above are not reading correctly - likely because of the issue I documented here concerning KPs not including all the qualifiers necessary for the UI to compose the right edge label.
LAMA3 regulates increased activity or abundance Of
CDK6 has increased expression affected by
Acetylcholine
should be:
LAMA3 CAUSES increased activity or abundance of
CDK6 has increased expression CAUSED by
Acetylcholine
I agree with Andy, I think CDK6 should be an answer but not the other genes like LAMA3.
I'll take a look at this; thanks for reporting. It looks like the directionality of the relationships in our query logic might not be coming through correctly.
Hi @brettasmi - yep, I think so too. Part of the problem is from enforcing 2 hops (gene1/protein)--[Xregulates]-->(gene2)--Xregulates]-->(gene3) and then getting something wrong with the directionality of the edge.
But independently, I was not able to find in SPOKE the edges CDK6 - Lama3, and that between Acetylcholine and CDK6. (Obviously everything that stimulates proliferation will upregulate CDK6 anyway - so I do not see the point of the query here).
@mbrush : Not sure if t is is a biolink/predicate qualifier issue. At least it should not be. I hope not. The direction of edges is so profound and flows into reasoning, even into the actual grpah query, so early in the answering process that it should not be something that the UI "composes" based on qualifiers.
Type: Bug Report
URL: https://ui.test.transltr.io/main/results?l=Acetylcholine&i=PUBCHEM.COMPOUND:187&t=3&r=0&q=27730ac2-b19e-4edd-b146-7a79551965b0
ARS PK: 27730ac2-b19e-4edd-b146-7a79551965b0
Steps to reproduce:
what gene may be upregulated by acetylcholine? search for LAMA3
Screenshots: