BioMNEDR: mechanism-guided network embedding for drug repurposing.

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug repurposing accelerates therapeutic discovery but current methods lack multi-scale mechanism integration, limiting interpretability.
• Existing computational approaches often fail to capture complex drug-disease associations.

Purpose of the Study:

• To introduce BioMNEDR (mechanism-guided network embedding for drug repurposing), a novel framework for enhanced drug repurposing.
• To improve the accuracy and interpretability of computational drug repurposing by integrating multi-scale biomedical mechanisms.

Main Methods:

• BioMNEDR integrates heterogeneous biomedical networks using biologically curated meta-paths.
• It generates low-dimensional embeddings that preserve protein-protein interactions and functional hierarchies.
• Multi-path predictions are integrated using an XGBoost classifier.

Main Results:

• BioMNEDR achieves state-of-the-art performance, outperforming strong baselines in AUROC, AUPR, recall, and F1-score.
• The framework maintains a balanced trade-off between precision and recall.
• Case studies demonstrate successful rediscovery of approved drugs and prioritization of new candidates.

Conclusions:

• BioMNEDR offers a robust computational framework for systematic drug repurposing by explicitly modeling multi-scale mechanisms.
• The approach enhances both predictive accuracy and biomedical interpretability in drug discovery.
• BioMNEDR facilitates the identification of promising drug candidates, such as cromoglicic acid for Alzheimer's disease.