Add custom SSSOM sources from RGD

[cthoyt]

https://download.rgd.mcw.edu/ontology/mappings/

[nichtich]

These are invalid as well:

#mapping_provider: RGD

mapping_provider must be an URI.

In addition they contain undefined CURIE prefixes (tested with sssom-js):

$ for txt in *.txt; do echo $txt>&2; sssom $txt; done > /dev/null
efo_hp_rgd.sssom.tsv.txt
Error: Unknown or invalid CURIE 'MONDO:0000088' in line 822
efo_mp_rgd.sssom.tsv.txt
Error: Unknown or invalid CURIE 'HP:0011123' in line 221
efo_rdo_rgd.sssom.tsv.txt
Error: Unknown or invalid CURIE 'DOID:0080630' in line 17
efo_vt_rgd.sssom.tsv.txt
rdo_mim_rgd.sssom.tsv.txt
Error: Unknown or invalid CURIE 'DOID:0040007' in line 17

[cthoyt]

@nichtich that's a good point about the mapping provider.

When it comes to unknown CURIE prefixes, my software stack tends to diverge from any standards and always normalize based on the Bioregistry. This eliminates a lot of headaches, since the differences between the way people write CURIEs/URIs for the same entities in the same semantic space is typically not important for downstream usage

As such, we can write a custom wrapper in SeMRA to load these files, that handles our prior knowledge. (from a philosophical side, this is a bit sad since SSSOM is supposed to reduce the need for this, but as people are just learning the standard, and we're improving its documentation, these kinds of inconsistencies are probably going to be common)

In the sssom-py implementation, we also made it possible to auto-load in the Bioregistry's prefix map for standardization. We also have preliminary javascript support for similar operations implemented in https://github.com/biopragmatics/curies.rs, which compiles to WASM and gets published on NPM. This might be an area that can help reduce re-implementation of certain functionalities in sssom-js :)

[nichtich]

Sure each SSSOM parser can have a set of default prefixes but this must be documented very well.

https://github.com/biopragmatics/curies.rs, which compiles to WASM and gets published on NPM.

I don't understand the purpose of this package. Is it just a wrapper around https://w3id.org/biopragmatics/bioregistry.epm.json for fast lookup?

[nichtich]

I've created a small script to convert bioregistry prefixes but the RGD files are invalid still because namespace prefix Orphanet is not included in Bioregistry (only lowercase orphanet). Could quickly be fixed by change in bioregistry but silently converting all CURIE prefixes to lowercase seems not right, unless it's explicitly recommended by SSSOM specification.

[cthoyt]

If I recall correctly, the SSSOM spec doesn't prescribe any specifics on how (extended) prefix maps should be handled. The SSSOM-py implementation uses the data structures from the curies package to take in prefix maps / extended prefix maps, but doesn't process them any further. The creation of the builtin prefix map for SSSOM-py is based off of the Bioregistry in combination with the Bioregistry's "contexts" feature, which allows for defining some overrides (see configuration here: https://bioregistry.io/context/obo)

To address this specifically, I added an uppercase Orphanet synonym to the Bioregistry in biopragmatics/bioregistry#1557