Auditing the multiply-related concepts within the UMLS.

Area of Science:

• Medical Informatics
• Computational Linguistics
• Knowledge Representation

Background:

• The Unified Medical Language System (UMLS) integrates diverse biomedical vocabularies.
• Understanding relations between concepts is crucial for knowledge discovery.
• Multiply-related concepts present challenges in data integration and interpretation.

Purpose of the Study:

• To audit relations in multiply-related UMLS concepts.
• To determine if relations originate from source vocabularies or UMLS integration.
• To explain the co-occurrence of multiple relations between concepts.

Main Methods:

• Analysis of relation compatibility at concept and term levels.
• Manual review of 288 concepts with contradictory relations.
• Examination of relations within source vocabularies and their contribution to UMLS integration.

Main Results:

• Contradictory relation combinations are rare (0.7%) in UMLS; homogeneous combinations occur in one-third of cases.
• Source vocabularies contribute to multiply-related concepts in only a quarter of instances.
• Manual analysis revealed conceptualization differences across source vocabularies.

Conclusions:

• Manual analysis effectively explains cross-vocabulary conceptualization differences in relations.
• Source vocabulary relation exploitation aids understanding of multiply-related concepts.
• The study clarifies the origin and nature of multiple relations in the UMLS.