Enhancing Drug Repositioning Through Local Interactive Learning With Bilinear Attention Networks

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug repositioning accelerates therapeutic development by identifying new uses for existing drugs.
• Integrating heterogeneous data is crucial for accurate drug-disease association prediction.
• Existing computational methods often struggle to capture complex drug-disease interactions.

Purpose of the Study:

• To develop DRGBCN, a novel deep learning method for drug repositioning.
• To enhance drug-disease inference by integrating multiple similarity networks.
• To improve the accuracy and reliability of predicting potential drug-disease relationships.

Main Methods:

• Constructed a comprehensive drug-disease network using multiple drug and disease similarity networks.
• Employed a layer attention mechanism to learn graph convolutional layer embeddings.
• Utilized a bilinear attention network to capture pairwise drug-disease interactions.
• Incorporated a multi-layer perceptron for final drug evaluation.

Main Results:

• DRGBCN achieved an average AUROC of 0.9399 in 10-fold cross-validation, outperforming baseline methods.
• Case studies on bladder cancer and acute lymphoblastic leukemia demonstrated practical applicability.
• Network analysis revealed successful clustering of similar drugs, offering insights into drug-disease interactions.

Conclusions:

• DRGBCN is a promising computational tool for effective drug repositioning.
• The method enhances the discovery of novel therapeutic applications for existing medications.
• DRGBCN contributes to advancing precision medicine through improved drug-disease association predictions.