Map to UMLS semantic types

[cmungall]

See also NCATS-Tangerine/beacon-aggregator#15 and NCATS-Tangerine/translator-knowledge-beacon#15

[cmungall]

@RichardBruskiewich has already done a lot of this, e.g. 9951bee

[RichardBruskiewich]

The UMLS semantic "Groups" (e.g. GENE, CHEM, DISO, etc.) are posted to the Biolink Model, but I've not yet put in the finer grained codes (e.g. 'aapp' etc. see https://metamap.nlm.nih.gov/Docs/SemanticTypes_2013AA.txt). I guess I could tackle these soon when I'm bored with straight computer software development :-)

[cmungall]

the goal isn't to include all of these. We're calling this done when existing entity types in BLM have a mapping

[cmungall]

Is this completed?

[RichardBruskiewich]

Ancient issue, long neglected... I note that the "semantic types" were note added. They are more fine grained than the UMLS semantic groups. I vaguely seem to recall that MediKanren may have been using them, but don't know if that is still the case.

I can strive to add them into Biolink today.

I'll use the CURIE xmlns prefix UMLSST (ST == Semantic Type). I'll obviously be selective on which ones to include since the scope of types is a superset of what Biolink covers.

For good measure(?), I can add the "codes" as UMLSSC:

e.g.

T083	geoa	Geographic Area	GEOG	Geographic Areas

will have mappings:

UMLSSG:GEOG
UMLSST:geoa
UMLSSC:T083

Also, I assume that Biolink categories can be "one-to-many" UMLS semantic types, where the Biolink model does not make fine grained distinction between the UMLS semantic types of a given UMLS Semantic Group,

For example, the UMLS Semantic Group "DEVI" doesn't seem to have any is_a: device categories in the model, so I'll aggregate all the DEVI semantic types as mappings to the one 'device' category in Biolink.

[RichardBruskiewich]

@cmungall are biolink category mappings supposed to be one-to-one? The reason I ask is that we have two UMLSSG:GENE mappings: one to the Biolink category "genomic entity" and the other (obviously) to "gene".

I wonder if this presents any practical difficulties with software using the Biolink model? I suspect perhaps...

The UMLS semantic types under the Semantic Group "GENE" are:

T028	gngm	Gene or Genome	GENE	Genes & Molecular Sequences
T085	mosq	Molecular Sequence	GENE	Genes & Molecular Sequences
T086	nusq	Nucleotide Sequence	GENE	Genes & Molecular Sequences
T087	amas	Amino Acid Sequence	GENE	Genes & Molecular Sequences
T088	crbs	Carbohydrate Sequence	GENE	Genes & Molecular Sequences

Which are (carelessly) broad in Biolink scope.

[RichardBruskiewich]

Completed the above task (I think...) and issued a pull request: #209