See https://github.com/biopragmatics/obo-db-ingest
It would be quite easy to add these as builds, and distribute the sqlite on s3.
Advantages:
- easy to query in OAK (though some methods and commands e.g
tree wouldn't make sense as these wouldn't follow expected structural shapes for ontologies)
- fast to query via SQL
Note there is ongoing discussion about URIs for these, but semantic-sql doesn't care, we store things natively as CURIEs, and the prefix table can be swapped to anything.
Ideally the products would be built and distributed (obo/owl/json) upstream, to avoid running the build step, as this introduces an additional source of potential pipeline failure, we also have to determine memory/disk requirements
cc @cthoyt