ARAX synonym returning unexpected Biolink prefixes

[gaurav]

The NodeNorm team received an issue from a user who was receiving unusual identifiers in the ARAX synonym search for Interleukin 2: TranslatorSRI/NodeNormalization#151

Oddly enough, the primary identifier this returns is LOINC:MTHU015779, which is not an identifier prefix for biolink:Protein, SmallMolecule or Gene. By contrast, the SRI NameRes service returns a number of results including PUBCHEM.COMPOUND:130168 and NCBIGene:3558 (https://name-lookup.transltr.io/lookup?string=interleukin%202&offset=0&limit=10), which refer to the small molecule and gene form of interleukin-2 respectively (https://nodenorm.transltr.io/1.3/get_normalized_nodes?curie=PUBCHEM.COMPOUND%3A130168&curie=NCBIGene%3A3558&conflate=false).

Could you please check how this identifier is getting into the ARAX synonym search? If it's coming from an SRI tool, please let us know!

[amykglen]

Hi @gaurav - so in our latest ARAX Node Synonymizer, there are a few clusters for Interleukin 2. The first two correspond to the gene/protein vs. small molecule versions of IL2 and align with the SRI's clusters for those concepts:

https://arax.ncats.io/devLM/?term=NCBIGene:3558
https://arax.ncats.io/devLM/?term=PUBCHEM.COMPOUND:130168

And the third cluster includes that LOINC identifier you reported:

https://arax.ncats.io/devLM/?term=LOINC:MTHU015779

This latter cluster includes only nodes from RTX-KG2 that are not recognized by the SRI NodeNormalizer. They're not tacked onto the first two clusters because their categories are NamedThing (which probably could be improved in RTX-KG2, but nonetheless is the current level of detail provided - I just reported this in this issue: RTXteam/RTX-KG2#277).

Do you have any concerns remaining here?