Dual-Branch Contrast Enhancement for Drug Repositioning

Drug repositioning offers a promising strategy to identify novel therapeutic applications for existing drugs. Despite the frequent use of graph neural network in this domain, their efficacy is often hampered by the sparsity of known drug interaction networks. Furthermore, diseases within the same subclass frequently share structural and clinical traits, leading to similar complications and responses to specific treatments among related viruses.To tackle the challenges of sparse drug interaction networks and leverage the shared treatment potential among similar diseases, we propose Dron, a dual-branched contrast-enhanced method for drug repositioning. Dron enhances drug and disease representations using a branched contrastive loss strategy, and improves identification accuracy by aggregating features from related categories. Validation across multiple public datasets shows that Dron significantly enhances drug repositioning performance. Molecular docking experiments on Alzheimer’s disease further confirm Dron’s effectiveness in addressing complex diseases.