Some NCBIGene records should have biolink category "GenomicEntity"

[saramsey]

Example: NCBIGene:780896.

Note that under the "General Gene Information" section, it has the property "phenotype only". I think in that case it makes sense to give it a biolink category biolink:GenomicEntity which is our closest Biolink approximation for "locus". I think this would require a fix in ncbigene_to_tsv_to_kg_json.py.

See issue #1213 which first brought this issue to my attention.

[saramsey]

In addition to correctly setting the biolink category, we should prepend "Genetic locus for " to the name and full_name fields for these types of NCBIGene records.

[kvarforl]

looks like this change (for the most part) made it into kg2.5.2.
There are now 1343 nodes in kg2 from NCBIGene that have category genomic entity and contain the string "locus" in their full name

match (n) where n.id starts with "NCBIGene" and n.category_label = "genomic_entity" and lower(n.full_name) contains "locus" return count(*)

---

However,
@saramsey the prepending of "Genetic locus for" made it into full_name but not name. good enough, or additional tweaks required?

match (n{id:"NCBIGene:780896"}) return n.name, n.full_name, n.category_label

n.name	n.full_name	n.category_label
"ACTD"	"Genetic locus associated with Acetabular dysplasia"	"genomic_entity"

[kvarforl]

@dkoslicki @edeutsch can you take a look at the above comment and let me know if you would like additional adjustments?

[edeutsch]

Thanks! I think we wanted the prefix "Genetic locus associated with " on the name field as well. The issue is that the NodeSynonymizer uses the name field to associate things, and was conflating the disease with the locus because they have the same name. The NS could still compensate for this, but fixing it at the KG level seemed to make the most sense?

[kvarforl]

Okay, I've added a fix for this! It's in aacf9c2 ; I mistyped the issue number in my commit message, hence the lack of traditional linkage :)

this should appear in the next kg2 build.

[saramsey]

Great, looks like there is nothing for me to do here, as @edeutsch has communicated what specifically is needed (i.e., on the name field and not just on the full_name field).

[ecwood]

Also, I think some NCBIGene records should have the biolink category "microRNA":

Ex.

{
  "iri": "https://identifiers.org/ncbigene:100126330",
  "synonym": [
    "MIR941-4",
    "MIRN941-4",
    "hsa-mir-941-4"
  ],
  "category_label": "gene",
  "full_name": "microRNA 941-4",
  "deprecated": "False",
  "name": "MIR941-4",
  "description": "Type:ncRNA; Locus:20q13.33; NameStatus:official",
  "provided_by": "identifiers_org_registry:ncbigene",
  "id": "NCBIGene:100126330",
  "category": "biolink:Gene",
  "update_date": "20210302"
}

This is labeled as a gene, but it is a microRNA (it's xref is mirbase:MI0005766 from the ETL of miRBase #1247). Is this a problem?

[ecwood]

The output of the cypher:
match (n {provided_by: 'identifiers_org_registry:ncbigene'}) where lower(n.full_name) starts with "microrna" and lower(n.category_label)="gene" return count(n)

is 1915.

[edeutsch]

I think it hinders our reasoning on KG2, yes, so a fix would be appreciated!

[ecwood]

The microRNA issue is addressed in 3378f83 (forgot to tie that commit to this issue)

[kvarforl]

Okay looks like the naming portion of this is fixed in kg2.6.0:

n.name	n.full_name	n.category_label
"Genetic locus associated with ACTD"	"Genetic locus associated with Acetabular dysplasia"	"genomic_entity"

closing it out, as I think 1379 covers the microRNA stuff :)