Active Learning-Assisted Exploration of [PO₄₀Mo₁₂]^3- for Alzheimer's Therapy Insights

Alzheimer's disease (AD), involving amyloid-β (Aβ) aggregation, has potential therapeutic modulators in polyoxometalates (POMs) like [PMo₁₂O₄₀]^3-. To clarify their inhibitory mechanisms, a multiscale computational strategy integrating active-learning Bayesian Optimization (BO) and density functional theory (DFT) is employed to explore low-energy configurations of isolated amino acids, [PMo₁₂O₄₀]^{3 -}-amino acid complexes, and [PMo₁₂O₄₀]^{3 -}-peptide systems. Hydrogen bonding and Coulombic repulsion dominate adsorption stability. Crucially, oxygen atoms in the [PMo₁₂O₄₀]^{3 -} cluster form multiple weak interactions (e.g., van der Waals, hydrophobic) with alkyl side-chain hydrogens in Aβ peptides. The synergistic effect of these weak interactions induces robust binding between the POM and peptide chains, stabilizing a tightly bound complex that sterically hinders Aβ self-assembly. Notably, simulations predict that the cluster preferentially targets hydrophobic amino acids with alkyl chains (valine, lysine, leucine, isoleucine) located in Aβ regions critical for aggregation-specifically, namely Aβ12, Aβ16-18, Aβ24, Aβ28, Aβ31-32, Aβ34-36, and Aβ39-41. These insights highlight the role of multivalent weak interactions in POM-mediated inhibition and identify key interfacial residues for therapeutic targeting.