Integrating CADD1 and UHPLC-ESI-HRMS2 to screen natural products for potential multi-target drugs against Alzheimer's disease

Alzheimer's disease (AD) is driven by multiple pathological pathways, making multi-target strategies highly attractive yet challenging to implement efficiently. Here, we present an integrated workflow combining ligand-based AI virtual screening, post-column UHPLC-ESI-HRMS ligand fishing, molecular dynamics simulations, cell-based assays, and network pharmacology to rapidly identify natural product scaffolds with desired activity profiles from large compound libraries. Using Aβ 1–42 and tau as model targets, a PyRMD deep-learning model was trained on curated ChEMBL data and applied to screen a 9.7-million-compound subset of ZINC15. Hits were matched against natural product databases, prominently enriching diarylheptanoids and pointing to turmeric ( Curcuma longa ) as a top candidate plant. Subsequent UHPLC-ESI-HRMS ligand fishing of turmeric extract confirmed non-covalent binding of eight curcuminoids to Aβ 1–42 . 100-ns MD simulations predicted favourable binding (ΔG < –10 kcal/mol) of C1–C4 to both Aβ 1–42 fibril cores and tau PHF (Paired Helical Filaments) cores. In SH-SY5Y cells, C1–C3 dose-dependently protected SH-SY5Y cells from Aβ 1–42 oligomer-induced toxicity and significantly reduced ThT fluorescence in Aβ 1–42 fibrillization and pre-formed fibril systems. Network pharmacology analysis of C1–C3 revealed 107 putative AD-related targets enriched in PI3K-Akt, EGFR, and PD-1/PD-L1 pathways. Although tau binding remains computationally predicted and requires further experimental validation, this study demonstrates a scalable, closed-loop platform capable of rediscovering and experimentally validating known bioactive natural scaffolds with multi-target potential from ultralarge libraries using curcuminoids as a successful proof-of-concept example. • Integrated AI virtual screening and UHPLC-ESI-HRMS rediscovers curcuminoids as multi-target Aβ/tau binders from 9.7 M compounds. • Turmeric curcuminoids bind Aβ₁₋₄₂ non-covalently; MD predicts strong dual binding to Aβ and tau cores (ΔG < −10 kcal/mol). • Curcuminoids protect SH-SY5Y cells from Aβ toxicity and reduce ThT fluorescence in fibril systems. • Network pharmacology identifies 107 CE targets enriched in PI3K-Akt, EGFR, and PD-1/PD-L1 pathways. • Scalable closed-loop platform validates potential natural multi-target scaffolds for Alzheimer's disease drug discovery.