Multi-omics and machine learning integration identifies TYMP as a peripheral blood biomarker and therapeutic target in Alzheimer's disease

Alzheimer’s disease (AD) is a multifactorial neurodegenerative disorder characterized by dysregulated kinase signaling, neuroinflammation, immune remodeling, and blood-brain barrier dysfunction. In this study, we applied an integrated systems pharmacology approach to investigate the therapeutic potential of resveratrol-derived phytochemicals in AD. Initially, nine structurally related stilbene derivatives were screened for central nervous system (CNS), drug-likeness, and Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) properties. Pterostilbene and desoxyrhapontigenin demonstrated superior lipophilicity, intestinal absorption, and predicted blood-brain barrier permeability, prioritizing them for downstream analyses. Differential gene expression analysis of the GSE85426 dataset identified 818 AD-associated genes, which were further explored using weighted gene co-expression network analysis (WGCNA). Fourteen co-expression modules were constructed, with nine modules significantly correlated with AD phenotype. Integrated overlap analysis combining compound targets, DEGs, curated AD databases, and WGCNA modules identified high-confidence therapeutic targets, highlighting Mitogen-Activated Protein Kinase (MAPK1) and MAPK3 as central hub genes. Functional enrichment analysis revealed dominant involvement of MAPK signaling, neurotrophin signaling, Toll-like receptor pathways, and synaptic vesicle regulation. Immune infiltration profiling demonstrated increased myeloid-associated signatures and subtle adaptive immune alterations in AD samples. Machine learning approaches (Support Vector Machine-Recursive Feature Elimination (SVM-RFE), Least Absolute Shrinkage and Selection Operator (LASSO), and Random Forest) consistently identified Thymidine Phosphorylase (TYMP) as a robust peripheral biomarker candidate. Molecular docking demonstrated stable binding of resveratrol derivatives within a conserved hydrophobic pocket, with pterostilbene exhibiting the strongest binding affinity (-6.80 kcal/mol). Collectively, this multi-layered analysis suggests that resveratrol derivatives may exert therapeutic effects through coordinated modulation of kinase-driven signaling networks and neuroinflammatory pathways, supporting their potential as multi-target candidates for AD intervention.