Temperature-responsive biphasic deep eutectic solvent system enables integrated extraction and self-separation of polyphenols from Moringa oleifera Lam. leaves: a machine learning-optimized biorefinery approach.

The sustainable recovery of bioactive phytochemicals is frequently hindered by the toxicity and separation bottlenecks of conventional volatile organic solvents. Herein, a novel amine-free temperature-responsive biphasic deep eutectic solvent (TRDES) system, composed of betaine, levulinic acid, and octanoic acid (BLevA-OA), was constructed for the integrated extraction and separation of polyphenols from Moringa oleifera Lam. (M. oleifera) leaves. This system exhibits reversible thermomorphic behavior, transitioning from a homogeneous single phase at 60 °C to maximize mass transfer, to a biphasic state at 25 °C for spontaneous phase-selective separation. To address the nonlinear complexity of the extraction process, a hybrid Gaussian Whale Optimization Algorithm-Support Vector Regression (GSWOA-SVR) model was developed, which precisely optimized the parameters to achieve a maximum total polyphenol content of 39.8 mg GAE/g DW, which is 1.7 times that of conventional ethanol extraction. Furthermore, multiscale computational analyses combining COSMO-RS and Molecular Dynamics (MD) simulations elucidated the microscopic mechanism: the superior extraction efficiency stems from a cooperative network of specific hydrogen bonding and hydrophobic interactions, while the temperature-triggered phase transition is driven by the kinetic modulation of component miscibility and diffusion. This work establishes a sustainable, closed loop strategy for the high efficiency recovery of polyphenols from M. oleifera leaves, offering a theoretical and practical foundation for the industrial application of smart green solvents in the circular bio-economy.