Using a Large Margin Context-Aware Convolutional Neural Network to Automatically Extract Disease-Disease Association...

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Machine Learning

Background:

• Disease-disease association (DDA) research offers critical insights for disease treatment and drug discovery.
• Current search tools struggle to efficiently retrieve the latest DDA findings from literature.
• Lack of specialized datasets and tools hinders disease-disease association extraction (DDAE).

Purpose of the Study:

• To develop the first publicly available DDAE dataset.
• To create a neural network model for extracting DDA from biomedical literature.

Main Methods:

• Formulated DDAE as a supervised machine learning classification task.
• Built a DDAE dataset comprising 521 PubMed abstracts with disease mentions and relation annotations.
• Developed and evaluated machine learning models, including Support Vector Machine (SVM) and Convolutional Neural Network (CNN), culminating in a large margin context-aware CNN architecture.

Main Results:

• The DDAE dataset included disease mentions, Medical Subject Heading IDs, and relation annotations.
• The SVM-based approach achieved an F1 score of 80.32%, while the CNN-based approach achieved 73.32%.
• The proposed large margin context-aware CNN model achieved the highest F1 score of 84.18%, demonstrating superior performance.

Conclusions:

• The study presents the first publicly available DDAE dataset, advancing text-mining research in this area.
• A novel large margin context-aware CNN model was developed, significantly improving DDAE performance.
• The developed dataset and model provide valuable resources for researchers investigating disease-disease associations.