Bug: Disease-Gene: missing HGNC links

[joeflack4]

Overview

Recently when running the disease-gene pipeline in mondo, we noticed that a lot of disease-gene source annotations were getting unexpectedly removed.

Explanation by example

id: MONDO:0013576
-relationship: has_material_basis_in_germline_mutation_in http://identifiers.org/hgnc/5716 {source="MONDO:mim2gene_medgen", source="OMIM:614102"} ! IGKC
+relationship: has_material_basis_in_germline_mutation_in http://identifiers.org/hgnc/5716 {source="MONDO:mim2gene_medgen"} ! IGKC

The source OMIM:614102 was removed because the OMIM gene-disease pipeline did not find evidence for an association between OMIM:614102 and HGNC:5716 (IGKC). However, we were surprised, because the label for it is IGKC, and we observed that IGKC is visible in the "Gene/Locus" field of the "Phenotype-Gene Relationships" table on https://omim.org/entry/614102.

You can also see it visible in morbidmap.txt, which is the data file that represents all of these "Phenotype-Gene Relationships" tables:

Phenotype	Gene/Locus And Other Related Symbols	MIM Number	Cyto Location
Kappa light chain deficiency, 614102 (3)	IGKC, IGKCD	147200	2p11.2

So why the removal? Because pipeline is not looking for HGNC symbols or IDs in morbidmap.txt. It's looking for them in mim2gene.txt. And there is a discrepancy where even if the HGNC symbol shows up for the association morbidmap.txt, it does not always appear in the same association in mim2gene.txt:

MIM Number	MIM Entry Type (see FAQ 1.3 at https://omim.org/help/faq)	Entrez Gene ID (NCBI)	Approved Gene Symbol (HGNC)	Ensembl Gene ID (Ensembl)
147200	gene	3514		ENSG00000211592
614102	phenotype

Possible solutions

We will contact OMIM to see if there is some reason why morbidmap.txt and mim2gene.txt appear to be out of sync in this way. The best solution may involve a fix on their end.

Otherwise, we can fix this on our end in a number of ways:
a. At the end of the pipeline, run an additional SPARQL query to see if there are any associations which are missing HGNC evidence, and if so, we can add it.
b. We can do the same thing, but rather than SPARQL, do a check in Python right before adding the association.
c. Earlier in the Python pipeline, we can combine all of the HGNC associations from both morbidmap.txt and mim2gene.txt.

For any of these a-c solutions, we can utilize hgnc_complete_set.txt, which we are already downloading and using elsewhere in the pipeline. We can use the first two columns, hgnc_id (e.g. HGNC:5716), and symbol (e.g. IGKC), to map any symbols we see in morbidmap.txt to their HGNC IDs.

[joeflack4]

@twhetzel fyi

[sabrinatoro]

(discussed with Trish on February 3rd, 2025)

Since there are a handful of genes that are not (or not yet) in HGNC, we will ignore these gene-disease associations.

Decision: if a gene is not in HGNC, ignore it (ie do not add this gene in Mondo).