Inferred Edge Should Match Fully Qualified Query Graph Edge - MVP2

[sstemann]

I dont fully understand why the predicates differ so frequently, when I read them in the UI, I'm not sure I understand how substantially different to the reasoned answer

"Tamoxifen Causes Decreased Activity Or Abundance Of Arid1a" vs "Tamoxifen Affects Decreased Activity Or Abundance Of Arid1a"

then the more general "Tamoxifen Affects Arid1a"

https://ui.test.transltr.io/main/results?l=ARID1A%20(Human)&i=NCBIGene:8289&t=2&r=0&q=1d650508-5292-4faf-a52c-21954c89fb36

When I look at it in the ARAX GUI ARAGORN

• Lookup via TMKP

  

• Inferred via Prediction - Affects Decreased Activity Or Abundance Of

  

The support graph of the prediction - looks to be a TMKP lookup

ARAX

• lookup via TMKP

BTE

• lookup via TMKP

  

• the other edge is inferred, prediction Affects

  

all but one of these SG edges are TMKP

Unsecret

• lookup via TMKP

  

• Inferred 1 - Causes Decreased Activity Or Abundance Of

  

• Inferred 2 - Causes Decreased Activity Or Abundance Of

  

[sstemann]

@sierra-moxon i'm not sure if any one thing is actually wrong here, but do you have guidance?

[sierra-moxon]

I believe these are all the same result and need to be merged.

How do we get to a merge?

• @cbizon - do you know why the inferred edges here do not have qualified_predicate: biolink:causes?

In the back of my mind, it seems a little odd that we have inferred edges that seem to be just reiterating the lookup edges. Am I misunderstanding these results?

[sstemann]

i believe ARS treats these edges as unique and doesnt merge them but defer to @MarkDWilliams @gprice1129

[gprice1129]

The predicate variation (especially Affects Decreased Activity Or Abundance Of) seems to be a mistake from the ARA/KP. We will look into why the edges in the first screenshot are not being compressed.

[dnsmith124]

After reviewing the pk again in the UI the edges in the first screenshot are now being properly compressed

[sstemann]

i see some compression but i dont particularly understand how these are substantially different that they need to be displayed this way (my question my really be - do the ARAs really have different intentions here or are they just non-standardized for some other reason)?

[sierra-moxon]

from TAQA: lets try this in CI and see if it works - we think BTE was fixed to use the qualified_predicate in the query. e.g. we think BTE, Aragorn will respond to a query with qualified_predicate (but we need to test via ARAX and make sure no ARAs are always left out).

then we need the UI to pass the qualified_predicate to the ARs.

Sui: too many predicates saying the same thing, eg "chemical affects expression of x" - gets mapped differently in the UI? In two-hop answers, we have down and up-regulated (can mean expression, activity, etc), a-upregulates-b- upregulates-c)

[colleenXu]

For BTE:

• as described here, BTE should now be able to accept queries with qualified_predicate set in Fugu.
• we now have a fix on CI to set the inferred edge qualifier set correctly

[sierra-moxon]

retested on CI today, far fewer predicates, more merging. did not spend too long evaluating results but the original issue to me, is fixed.

[sstemann]

this is in prod