Ingest UniProt database registry and prefixes

[ialarmedalien]

UniProt

Description

UniProt provides a list of external databases that it links to, along with information on how to create the appropriate URLs for entries.

Prefixes

185 freshly-baked prefixes and attendant database information

I'm interested in getting the prefixes used by UniProt integrated into the Bioregistry. As mentioned above, UniProt provides a list of external databases that it links to, with info including

• prefix
• db name, description
• URL
• format of link for an individual entity

Some prefixes are already aligned, either with the Bioregistry preferred prefix or with a synonym, and some represent a subclass or superclass (e.g. UniProt uses merops to refer to merops.entry). Other entries are missing from Bioregistry -- e.g. PATRIC.

It would be great to be able to incorporate this prefix data into Bioregistry programmatically. A coworker and I have already done some work on mapping the UniProt and Bioregistry prefixes where there is not an exact match.

What would be the best way to proceed?

[bgyori]

Thanks @ialarmedalien! Indeed, this UniProt registry is already integrated into the Bioregistry (https://bioregistry.io/metaregistry/uniprot) but as you noted, the alignment is not complete.

You can see a summary of the alignment here: https://bioregistry.io/api/metaregistry/uniprot/mappings.json, this covers 95 mappings currently.

To add more to this, we could do some combination of the following:

• The code for UniProt registry processing and alignment is implemented here: https://github.com/biopragmatics/bioregistry/blob/main/src/bioregistry/external/uniprot/__init__.py so if there is any general/programmatic improvement to the alignment, it could be implemented here.
• For prefixes that already exist in the Bioregistry but the automated alignment didn't find the correct mapping, manual mappings can be curated here: https://github.com/biopragmatics/bioregistry/blob/main/src/bioregistry/data/curated_mappings.sssom.tsv. You can see several examples of mappings to the UniProt registry curated here already.
• For prefixes that don't yet exist in the Bioregistry at all and are relevant, we might want to create new prefixes and then add the mapping (or let the auto-alignment create the mapping). Not sure how many of these there are, depending on the number, we could do this manually or find some more scalable solution.

@cthoyt might also have more sugesstions!

[cthoyt]

Note that many of the entries in UniProt's cross-reference database aren't actually new prefixes, but are providers for UniProt, NCBI Taxonomy, Entrez, or other gene identifiers.

As Ben mentioned above, I've previously taken a pass at finding some (if not all) of the novel prefixes and adding new Bioregistry records for them.

The remaining ones that are just providers don't get automatically incorporated by the alignment workflow. These could be manually added as needed to the providers key in the uniprot or appropriate prefix. Are there specific ones you want to be able to use?

[ialarmedalien]

I am pulling in a load of data from UniProt and want to remap prefixes from what they publish to what is registered in the Bioregistry (as far as is possible). I have various comments and questions.

One issue is that UniProt refers to databases using unique IDs of the form DB-nnnn in their ext database list, but the data itself uses the abbreviation field from that list, so the DB-nnnn IDs are a bit of a red herring. It would be more useful if Bioregistry stored the abbreviation, since that is what appears in external dbxrefs in their datasets, and would make it easier to find the preferred prefix for that DB.

A related question: is it possible to specify that a prefix is from a certain metaregistry and then do a prefix normalisation? If not, how hard would it be to implement that kind of functionality? It seems as though there is already a priority list with fallbacks for normalising prefixes, but if you know all your data is from the same source, it would be very handy to be able to just specify that.

I know there are various dbs that UniProt links out to that just use an ID from an existing DB (e.g. STRING uses UniProt IDs; the Alliance of Genome Resources, AGR, uses the IDs from member DBs, e.g. AGR:MGI:12345). For those that don't appear in the Bioregistry, I don't know if it's that useful to add them to Bioregistry unless the prefix is used elsewhere.