Botulinum toxin type a shows up twice in the result set for "What treats Cerebral Palsy", one page apart with different scores for each result

[sierra-moxon]

https://ui.test.transltr.io/results?l=&t=&q=c572dfd1-1a94-4074-bb5a-6bd0b27cc045

page 1 result with 30 evidence

page 2 result with 1 evidence

papers do not appear to overlap

[sierra-moxon]

from TAQA:
behind the scenes, these have different identifiers (different curies) - so UI can't merge.
Chris B: could be a mistake in NN, they should be synonymous, or they could be really different and people give them the same name. That would take some more work, could be a "conflation", etc.

[sierra-moxon]

from TAQA: this could be a NN issue.
http://identifiers.org/umls/C0006050
http://identifiers.org/unii/E211KPY694

[gaurav]

MESH:D019274 should connect these two cliques together (since it is connected to UMLS:C0006050 in the first clique and is itself part of the second clique), so yup, it does look like a NodeNorm issue. Let me look into what's going wrong here.

[gaurav]

I think the issue here is that we don't currently use MeSH identifiers when combining protein identifiers -- we only use NCIT, UMLS, UniProtKB and the protein ontology. I worry that doing that might overcombine concepts -- I'll discuss this with @cbizon and get back to you.

[cbizon]

I think that's a valid concern. We should test it out. My guess (?) is that if we're ok merging w/ NCIT & UMLS, then MESH will probably be ok as well...

[sierra-moxon]

@gaurav - is this one closeable? :) (of course reopen if not)