Network-Medicine-Guided Drug Repurposing for Alzheimer's Disease: A Multi-Dimensional Systems Pharmacology Approach

Alzheimer's disease (AD) drug development faces persistent challenges from blood-brain barrier limitations and inadequate integration of medicinal chemistry considerations with computational predictions. We developed a comprehensive Central Nervous System (CNS)-focused network medicine framework integrating machine-learning-validated BBB penetration prediction (95.7% accuracy, 0.992 AUC-ROC), modality-specific tractability assessment, and transparent evidence classification to identify viable drug repurposing candidates. CNS-specific pre-filtering refined 24,474 DGIdb compounds to 8247 CNS-relevant drugs, analyzed through multi-dimensional network scoring and systematic pharmaceutical property assessment. Modality stratification generated separate rankings for small molecules (3667 candidates), peptides (73 candidates), and biologics (3 candidates), acknowledging distinct BBB penetration mechanisms. Analysis revealed 64.8% of small molecules achieving Class I (Highly Tractable) status, with 83.6% demonstrating favorable BBB penetration. Plerixafor emerged as the top-ranked small molecule (score: 1.170), while trofinetide achieved the highest peptide ranking (score: 1.387), though classified as speculative, pending AD-specific validation. Successful identification of the FDA-approved AD therapeutics memantine and donepezil among the top candidates validated the computational performance, while the predominance of mechanistic evidence classifications (86.7%) highlighted that network predictions represent hypothesis-generating tools requiring systematic experimental validation rather than definitive therapeutic recommendations. The framework bridges computational predictions with pharmaceutical development requirements, providing actionable prioritization for systematic preclinical investigation addressing AD intervention.