Leveraging the UMLS As a Data Standard for Rare Disease Data Normalization and Harmonization.

Area of Science:

• Medical Informatics
• Bioinformatics
• Rare Disease Research

Background:

• Rare disease datasets are fragmented across various resources.
• Lack of standardized data hinders effective analysis and clinical application.
• The Unified Medical Language System (UMLS) is a potential solution for data normalization.

Purpose of the Study:

• To evaluate the Unified Medical Language System (UMLS) for normalizing and harmonizing rare disease data.
• To propose extensions to the UMLS for global adoption as a rare disease data standard.

Main Methods:

• Analyzed data mappings between the UMLS and four rare disease resources (GARD, Orphanet, OMIM, MONDO).
• Assessed both curated and knowledge graph-generated disease mappings.
• Evaluated the accuracy of UMLS normalization and categorization of rare disease concepts.

Main Results:

• UMLS normalized over 50% of concepts from GARD, MONDO, and Orphanet, and 100% from OMIM.
• 3,876 UMLS concepts were identified across 58,636 mappings.
• Manual review showed 99% accuracy in UMLS mappings, including synonyms, subtypes, and siblings.

Conclusions:

• The UMLS accurately represents rare disease concepts, genes, and phenotypes, supporting data harmonization.
• Recommends adopting the UMLS as a data standard for rare diseases.
• Facilitates future clinical and community applications for rare disease data.