Use preferred bioregistry prefixes for normalized entity identifiers

[dhimmel]

Great to see that BERN2 normalizes entities to compact identifiers in resource:identifier format. I noticed that there is an opportunity to standardize the prefixes used with Bioregistry:

• NCBITaxon prefix as per http://bioregistry.io/registry/ncbitaxon
• NCBIGene prefix as per http://bioregistry.io/registry/ncbigene

FYI I didn't check all the entity types BERN2 is capable of tagging for whether they use the preferred prefix.

@cthoyt might also be helpful here.

[cthoyt]

I’d be happy to help. I’d like to try running bern2 myself locally and I’m sure doing this would make it easier to evaluate if the results are useful

[mjeensung]

Hi @dhimmel

Thank you for your suggestions for improving BERN2.

Do you mean that it is more standardized to use NCBIGene:10533 (for gene/protein) and NCBITaxon:10095 (for species) instead of EntrezGene:10533 and NCBI:txid10095?

[dhimmel]

Do you mean that it is more standardized to use NCBIGene:10533 (for gene/protein) and NCBITaxon:10095 (for species) instead of EntrezGene:10533 and NCBI:txid10095?

Exactly. I see a benefit if all entities tagged are represented as Bioregistry supported CURIEs to make integration with other datasets the most straightforward as possible.

Additional notes:

• I was surprised that Bioregistry doesn't list EntrezGene as an alternative prefix for NCBIGene.
• Looking at the NCBI page here, I see "for references in articles please use NCBI:txid10095". @cthoyt how do you reconcile the Bioregistry format being incompatible with the source-recommend format?

[cthoyt]

I just added in EntrezGene to https://bioregistry.io/registry/ncbigene, but I don't really see a standardized way to reconcile things that look like NCBI:txid10095 since it doesn't follow the spirit of CURIEs. You could always do some text-based preprocessing if you get stuff like this.

[mjeensung]

@dhimmel @cthoyt

Thank you for your suggestions! We will consider replacing current prefixes with the prefixes in BioRegistry.

[mjeensung]

Hi @cthoyt,

I was checking BioRegistry and noticed something that I'd like to clarify. While Entrez Gene ID has the preferred prefix NCBIGene, MESH ID does not and instead uses the prefix mesh in its CURIE. My first question is why some CURIEs use preferred prefixes while others use prefixes. My second question is whether it is common in CURIE to use lowercased prefixes such as mesh:C063233 rather than MESH:C063233 as BERN2 does.

[mjeensung]

https://github.com/dmis-lab/BERN2/commit/bbad178247047faf58ff204ea6adb383ae86717f