search

DiseaseOntology / HumanDiseaseOntology

Repository for the Human Disease Ontology.
Creative Commons Zero v1.0 Universal
336 stars 109 forks source link

Unexpected removal of UMLS refs #1344

Open drobbarfield opened 3 months ago

drobbarfield commented 3 months ago

Is your request related to a specific disease? Please describe.

We analyzed differences in the disease ontology version we had in our database for the last couple of years and the latest and found some unexpected removals of UMLS refs.

For example, Lynch syndrome (https://uts.nlm.nih.gov/uts/umls/concept/C4552100) was removed from Lynch syndrome (DOID:3883).

umls-ref-removal.txt

Describe the proposed change(s) Please review the attached spreadsheet and add the UMLS refs back as appropriate.

References NA

csbjohnson commented 3 months ago

Hi Dylan, @drobbarfield, thank you for your request on updating UMLS refs. We will review it and get back to you soon.

Best, Claudia Marie Sánchez-Beato Johnson

allenbaron commented 3 months ago

Thanks @drobbarfield for sharing these with us. DO's mappings to UMLS are primarily automated based on mappings from other vocabularies, such as MeSH and NCI. These don't always agree with one another. As I review various mappings, I'm going to add supporting notes for future reference.

The Lynch syndrome example probably lost the UMLS xref to Lynch syndrome because of complex and confusing mappings of UMLS to MeSH, NCI, and Orphanet... and lack of mappings to OMIM phenotypic series (or the DO). NCI and Orphanet have two distinct terms with a parent-child relationship that are using disease names which, in clinical practice, have mostly become synonymous (Lynch syndrome and HNPCC), while the MeSH term correctly has both names, but is linked to 4 separate UMLS entities. I'm adding back the lost UMLS mapping you shared and updating the current UMLS mappings to indicate they are actually broad matches and should not be included as xrefs.