• A novel GNN-based framework integrating heterogeneous knowledge graphs improves drug repurposing prediction accuracy by up to 15% over baselines.
• The model identifies promising drug candidates for Alzheimer's disease and COVID-19, several of which are corroborated by recent clinical trials.
• The framework demonstrates robustness across different datasets and can be adapted to other biomedical prediction tasks.
• The integration of multi-relational data enhances the interpretability of predictions, aiding researchers in understanding underlying mechanisms.