Annotating the biomedical literature for the human variome

Area of Science:

• Genomics
• Bioinformatics
• Medical Informatics

Background:

• Cataloging human genetic variation and its link to disease is complex.
• Existing methods struggle to systematically capture this information from literature.
• The International Society for Gastrointestinal Hereditary Tumours (InSiGHT) database highlighted a need for a structured approach.

Purpose of the Study:

• To introduce the Variome Annotation Schema for capturing human genetic variation and disease relationships.
• To assess the schema's applicability and reliability in annotating scientific literature.
• To explore the schema's utility for database curation and text mining development.

Main Methods:

• Developed the Variome Annotation Schema to represent concepts and relations in genetic variation literature.
• Applied the schema to a corpus of full-text articles on inherited colorectal cancer.
• Evaluated inter-annotator agreement using F-score for both exact and boundary matching of entities and relations.

Main Results:

• Achieved high inter-annotator agreement (0.78-0.95 F-score) for entity annotation, improving with relaxed boundary matching (≥0.87 F-score).
• Demonstrated reliable agreement for several relation types, with 'cohort-has-size' reaching 0.90 F-score.
• Confirmed the schema's relevance for the InSiGHT database curation process.

Conclusions:

• The Variome Annotation Schema effectively captures key information on human genetic variation and disease from literature.
• The schema and annotated corpus are valuable resources for developing text mining tools in human variome research.
• Text mining holds significant potential for improving the curation of genetic variation data.