Asset 503: bad answer ID for thrombin/F2 gene

NCATSTranslator / Tests

3 stars 2 forks source link

Asset 503: bad answer ID for thrombin/F2 gene #83

Open colleenXu opened 2 months ago

colleenXu commented 2 months ago

Asset 503 says the expected output name is UNIPROTKB:P00734, which is the gene F2 + its proteins (Dev NodeNorm).

However, the output ID CHEMBL.TARGET:CHEMBL204 is likely something no ARA/KP recognizes. NodeNorm doesn't resolve this ID..

I imagine the expected output ID needs to be changed to either NCBIGene:2147 (the gene ID which is primary since we're doing gene-protein conflation) or UniProtKB:P00734 (note the corrected CURIE format). I imagine the name needs changing to "F2/prothrombin/thrombin" or something like that.

colleenXu commented 2 months ago

I think all tools have been consistently failing this test due to the answer ID. This asset is specifically described in https://github.com/NCATSTranslator/Tests/issues/69

maximusunc commented 1 month ago

Putting UNIPROTKB:P00734 into Name Resolver gives PR:000051041 which seems to be a subclass of thrombin. And using conflation in NN still gives the same ID. Are you using a different conflation?

colleenXu commented 1 month ago

I don't agree with using prothrombin N-glycosylated 4 (human) / PR:000051041 as the CURIE. That looks like a very specific variation of prothrombin (N-glycosylated at Asn121) with very little literature on it.

It may be worth asking the original writer of the test what ID to use?

This is what I know:

F2 gene encodes the protein prothrombin, which is cleaved to create thrombin.
UniProtKB:P00734 is a ID for the F2 gene/prothrombin.
I was using NodeNorm Dev w/ conflation and found that UniProtKB:P00734 resolves to a F2/prothrombin entity.
I think UniProtKB:P00734 is an acceptable answer ID, but I'm not sure what the original writer of this test wanted.

sandrine-muller-research commented 1 week ago

please have a look at my response to #84 this is the way I worked with the tests assets: since I am testing the UI I report what the UI states. I think One asset can be a pass or a fail for several reasons (I used to express them but I am not sure the metadata was kept in the long run). Getting what the UI states is helpful to use those test assets for other suites (e.g. a normalization suite).

sandrine-muller-research commented 1 week ago

Please assign this email address/GitHub account in the future, I do not receive any notification with the other one. Thanks!