Understanding Scores

[mbrush]

Link: https://ui.test.transltr.io/main/results?l=Asthma&i=MONDO:0004979&t=0&q=ba7bdf9b-03b0-45ad-811a-255eba0fc321

Look at Valategrast as a result for what may treat Asthma.

• one support path (treats)
• two sources (BTE and Chembl)
• no supporting pubs or clinical trials

A couple issues/questions:

1. If BTE is a primary source, presumably this is an inferred ARA prediction, and there should be additional support paths to show the reasoning BTE used.

2. For the Chembl Statement, exploring this chemical on the Chembl site reveals two associated clinical trials (both phase 2) on which this 'indication' is based:

Chembl record:

Linked CT.gov trial records:

. . . but the UI doesn't report any supporting trials.

As an aside, it would be very useful to the end user if we reported the phase of supporting trials in the UI clinical trials modal - is this in the works?

[andrewsu]

From BTE's result in arax, this is the support graph:

This asthma - has_phenotype - asthma is a type of edge that while correct, is not helpful. So we are planning on removing them in https://github.com/biothings/biothings_explorer/issues/734.

Another issue that this example raises. The UI reports this as the fifth result. BTE reports it as number 252, and as far as I can see, no other ARA is reporting this answer. That seems off, so added the O&O label...

[mbrush]

Thanks for this Andrew. The self-referential edge definitely confuses things here. But if I understand, the BTE 'template' that was used/matched to make this prediction is likely something like:

Chemical - treats -> DiseaseOrPhenotype -phenotype of -> Disease

(but in this case the Phenotype and Disease are the same thing - asthma).

If so, an instance of this metapath/template should show up as a support path in the UI, right? But as you can see in my screenshot above there is no support path shown. What do you think is happening here? Maybe the UI is confused by the self-reference and cuts it? This may be moot for this specific issue if BTE plans to remove inferences like this - but may point to a broader issue to address.

---

The other issue I raised which I'd love your input about is the lack of metadata about the clinical trials on which this 'treats' edge retrieved from Chembl is based. Other KPs that provide 'treats' edges based on clinical trials info aggregated by Chembl provide the max_phase for indication (phase 2 in this case), and also a link to the ct.gov records for the relevant trials. Is this something mychem collects that BTE could pass along?

[cbizon]

Needs a retest on BTE with current version.

[sharatisrani]

Tried re-running, very strange results (but just a UI issue): Same pk gave this:

[sharatisrani]

So as of today, can't see what BTE did, but valategrast is still result #5 overall, via spreadsheet analysis. Improving is the only ARA that came back with that result, and with a score of 1. Thus, clubbing it with issue #845 and will discuss at July 17 O&O.

[andrewsu]

Retest: https://ui.test.transltr.io/main/results?l=Asthma&i=MONDO:0004979&t=0&r=0&q=253d1b7d-433e-4a7b-a7bf-a9d99da070cc

There are two sub-issues in this issue:

• Ordering/ranking: leaving this to @sharatisrani to report back
• EPC: @mbrush: the asthma-has_phenotype-asthma edge is gone now. Are there other EPC issues to address (that won't be addressed by the treats refactor)?