Ontologizing health systems data at scale: making translational discovery a reality

Area of Science:

• Biomedical Informatics
• Computational Biology
• Clinical Data Science

Background:

• Common data models standardize electronic health record (EHR) data but lack semantic integration for deep phenotyping.
• Open Biological and Biomedical Ontology (OBO) Foundry ontologies offer computable biological knowledge for data integration.
• Mapping EHR data to OBO ontologies is challenging, requiring manual curation and domain expertise.

Purpose of the Study:

• To introduce OMOP2OBO, an algorithm for mapping Observational Medical Outcomes Partnership (OMOP) vocabularies to OBO ontologies.
• To facilitate semantic integration of EHR data for enhanced deep phenotyping.
• To improve the identification of patients, particularly those with rare diseases, for potential genetic testing.

Main Methods:

• Developed OMOP2OBO, an algorithm to automate the mapping of OMOP vocabularies to OBO ontologies.
• Applied the algorithm to map conditions, drug ingredients, and measurement results.
• Evaluated the coverage of mapped concepts across clinical practice data from 24 hospitals.

Main Results:

• Generated mappings for 92,367 conditions, 8,611 drug ingredients, and 10,673 measurement results.
• Achieved 68-99% coverage of clinical concepts, demonstrating broad applicability.
• Successfully identified undiagnosed rare disease patients who could benefit from genetic testing.

Conclusions:

• OMOP2OBO effectively aligns OMOP vocabularies with OBO ontologies, overcoming semantic integration challenges.
• The algorithm significantly advances EHR-based deep phenotyping capabilities.
• This approach opens new avenues for discovering and characterizing patient cohorts for clinical research and genetic testing.