Reported primary source (Drug Gentral) does not support Result (Baclofen )

[mbrush]

https://ui.test.transltr.io/main/results?l=Cerebral%20Palsy&i=MONDO:0006497&t=0&q=52bea588-5893-48b2-8561-16309ab0a1ed

Search for 'What may treat Cerebral Palsy', and go to Baclofen Result . . . there are actually two separate Baclofen Results - as reported in #623.

The first Baclofen Result reports DrugCentral as a primary source, but I could find no info in Drug Central connecting Baclofen to Cerebral Palsy: https://drugcentral.org/drugcard/282?q=baclofen .

Note also that there is a Baclofen variant (Arbaclofen) reported as a result in Translator, also with Drug Central as a primary Source - and here to I find no connection in the source. (in fact, there is not even an Arbaclofen record in Drug Central).

[sierra-moxon]

Could it be that we're making an inference based on indication (from DrugBank) for treating Muscle Spasticity of Cerebral Origin?

I know we've discussed this before for other diseases that have multiple forms, but there are also multiple kinds of CP -- spasticity is a "sign/symptom" of one/some of those...
https://www.ebi.ac.uk/ols4/ontologies/mondo/classes/http%253A%252F%252Fpurl.obolibrary.org%252Fobo%252FMONDO_0000396?lang=en

It would be completely incorrect to offer "baclofen" as a treatment for hypotonic cerebral palsy.

[mbrush]

Could be . . .might be worth exploring how the node normalizer handles Muscle Spasticity of Cerebral Origin. and whether it treats it as equivalent to 'cerebral palsy', or to a more specific subtype like 'spastic cerebral palsy'.

This would explain the provenance - but also surfaces an issue I've raised before about difficulty in finding how knowledge in an edge appears in its primary source - in cases where the node normalizer results in the name displayed in the UI being different from the name used in the source.

If the user has no provenance for how the node normalizer does these mappings they will often be unable to find evidence of the edge in the reported primary source.

[sierra-moxon]

@gaurav - apologies if this is not a node-normalizer issue; but do you think my explanation of why we see baclofen associated with cerebral palsy in this instance is plausible?

[gaurav]

I re-ran this query, and I don't think the connection is being made because of node normalization -- instead, it looks like the connection is being made though "spasticity" as a phenotype of cerebral palsy. What do y'all think?

This would explain the provenance - but also surfaces an issue I've raised before about difficulty in finding how knowledge in an edge appears in its primary source - in cases where the node normalizer results in the name displayed in the UI being different from the name used in the source.
If the user has no provenance for how the node normalizer does these mappings they will often be unable to find evidence of the edge in the reported primary source.

That is a good question! There are two types of provenance we would think about in this case:

• Explaining that we understand a particular DrugCentral identifier to mean the same thing as a CHEBI identifier, but that we could be wrong there. This is something we could do in the UI with the clique information we have now, but would require additional information to be added to an already busy user interface.
• Explaining that in some cases we're doing this with "drug-chemical conflation", where we believe a bunch of closely related chemicals should be considered to mean the same thing. This is also something we could do in the UI with the information we have now, but would require additional information to be added to an already busy user interface.
• Explaining why we believe that a particular DrugCentral identifier to mean the same thing as a CHEBI identifier. This is information we don't have now, but we're thinking about it at Publish intermediate files or other information to explain how we came to our cliquing decisions TranslatorSRI/Babel#319 -- this is probably something that will happen outside of current Translator funding.