Identifying Missing IS-A Relations in Orphanet Rare Disease Ontology

Area of Science:

• Ontology development and maintenance
• Bioinformatics
• Medical informatics

Background:

• The Orphanet Rare Disease Ontology (ORDO) is crucial for rare disease research and applications.
• Ensuring the accuracy of ORDO, particularly its hierarchical 'is-a' relations, is vital for reliable downstream use.
• Existing methods for ontology quality assurance may not fully capture all types of missing hierarchical information.

Purpose of the Study:

• To implement an automated quality assurance pipeline to detect missing 'is-a' relations within the Orphanet Rare Disease Ontology (ORDO).
• To propose a novel computational approach for identifying potential omissions in ontological hierarchies.
• To assess the feasibility and effectiveness of automated auditing for ontology completeness.

Main Methods:

• Extraction of lexical features from concept names within ORDO.
• Generation of related and unrelated feature-sharing concept pairs.
• Development of a method to identify potential missing 'is-a' relations based on shared derived term-pairs from concept pairs.
• Validation of identified missing relations using the Unified Medical Language System (UMLS).

Main Results:

• The pipeline identified 705 potential missing 'is-a' relations in the ORDO 2022-06-27 release.
• External validation using UMLS confirmed 164 of these potential relations.
• The approach demonstrates promise in identifying gaps in ontological hierarchies.

Conclusions:

• The developed automated pipeline is an effective tool for auditing 'is-a' relations in ORDO.
• The findings highlight the ongoing need for ontology curation and quality control.
• Further domain expert review is recommended to validate the remaining identified potential missing relations.