A context-based ABC model for literature-based discovery.

Area of Science:

• Biomedical Informatics
• Computational Biology
• Literature-Based Discovery

Background:

• The ABC model is a literature-based discovery approach that identifies relationships between entities through intermediate terms.
• The traditional co-occurrence based ABC model lacks biological context, leading to irrelevant paths and requiring expert verification.
• Existing models struggle to filter meaningful connections due to a sole focus on entity co-occurrence.

Purpose of the Study:

• To enhance the ABC model by incorporating biological context for more precise literature-based discovery.
• To develop and evaluate context-based and context-assignment-based ABC models.
• To improve the automatic generation of scientific hypotheses by reducing irrelevant paths.

Main Methods:

• Defined four biological context elements: cell, drug, disease, and organism.
• Proposed two extended ABC models: a context-based ABC model and a context-assignment-based ABC model.
• Evaluated model performance using known protein interactions (APOE-MAPT and FUS-TARDBP) relevant to neurodegenerative diseases.

Main Results:

• The co-occurrence based ABC model achieved low precision (27.1% and 22.1%) for the tested relations.
• The context-based ABC model significantly improved precision (71.4% and 77.9%).
• The context-assignment-based ABC model achieved high precision (89% and 97.5%), outperforming the co-occurrence model.

Conclusions:

• Incorporating biological context substantially increases the precision of the ABC model for literature-based discovery.
• The proposed context-based and context-assignment-based models offer a more efficient and accurate approach to hypothesis generation.
• These enhanced models reduce the need for extensive expert verification in literature-based discovery.