Supporting biomedical ontology evolution by identifying outdated concepts and the required type of change

Area of Science:

• Bioinformatics
• Ontology Engineering
• Knowledge Representation

Background:

• Ontology evolution poses challenges for semantically enriched data systems, particularly in dynamic fields like biomedicine.
• Automatic identification and revision of outdated ontology concepts remain open research problems.

Purpose of the Study:

• To develop and evaluate an approach for identifying ontology concepts needing revision and suggesting appropriate update actions.
• To leverage structural, external knowledge, and temporal data for improved ontology maintenance.

Main Methods:

• Utilized supervised learning classifiers to analyze ontology structure, external knowledge (PubMed, UMLS), and historical data.
• Applied the approach to four major biomedical ontologies: ICD-9-CM, MeSH, NCIt, and SNOMED CT.
• Evaluated classifier performance based on accuracy in identifying concepts requiring revision and the type of revision needed.

Main Results:

• Achieved high accuracy in identifying outdated concepts, with performance ranging from 68% for SNOMED CT to 91% for MeSH.
• Demonstrated the effectiveness of the approach across diverse biomedical ontologies.
• Obtained an average accuracy of 71% across all tested datasets.

Conclusions:

• The proposed method effectively automates the identification and revision suggestion for outdated ontology concepts.
• This approach offers a valuable solution for managing large, dynamic biomedical ontologies.
• Improved ontology maintenance can enhance data annotation, indexing, and reasoning in biomedical systems.