RSM-GA-BP optimization of Ultrasound-Enzyme-Assisted deep eutectic solvent extraction (UEADESE) for flavonoids from Abelmoschus manihot (L.) leaves and mechanistic insights.

This study aimed to optimize the ultrasonic-enzyme-assisted deep eutectic solvent extraction (UEADESE) Abelmoschus manihot (L.) leaves flavonoids (AMLF) and elucidate the underlying mechanisms. Screening of extraction parameters was first performed via single-factor experiments, followed by optimization using response surface methodology (RSM). A back-propagation (BP) neural network coupled with a genetic algorithm (GA) was constructed to compare its predictive performance with that of RSM. Under the optimal extraction conditions determined as follows: molar ratio of choline chloride to 1,4-butanediol (1:6), 30% water content, 2.4% (w/w) cellulase-xylanase (total enzyme concentration), a mass ratio of cellulase and xylanase 1:3 (w/w), 1:50 (g/mL) solid-liquid ratio, 44 min ultrasonic treatment time, 180 W ultrasonic power, the maximum total flavonoid extraction yield was achieved at 37.54 mg/g. Molecular dynamics simulations revealed that flavonoids were anchored within DES nano-clusters via a persistent hydrogen-bond network, which enhanced solubility while inhibiting thermal degradation of redox-active hydroxyl/glycosidic moieties. Fourier-transform infrared (FT-IR) analysis confirmed that the formation of hydrogen bonds between deep eutectic solvent (DES) and flavonoids contributed to a marked enhancement in extraction efficiency. Antioxidant assays (DPPH, ABTS, hydroxyl radical scavenging, and total reducing power) demonstrated that UEADESE extracts exhibited significantly higher antioxidant activity than conventional solvent extracts. These findings validate UEADESE as a sustainable and effective method for efficiently extracting AMLF while preserving its antioxidant activity, thereby laying a scientific foundation for its potential application in functional food development.