Ultrasound-assisted deep eutectic solvent extraction of polyphenols from Sophora alopecuroides L. seeds: process optimization, antioxidant evaluation and machine learning-assisted structure–activity interpretation

This study first applies ultrasound-assisted deep eutectic solvents (DESs) for green and efficient extraction of polyphenols from Sophora alopecuroides L. seeds. The process was optimised using Plackett–Burman design and response surface methodology, yielding optimal conditions: a choline chloride-to-glucose molar ratio of 1:0.57, a solvent moisture content of 42.5%, a liquid-to-solid ratio of 56.1 mL/g and a temperature of 57 °C, achieving a polyphenol yield of 55.71 mg/g. The extracted polyphenols exhibited concentration-dependent radical scavenging activity against DPPH and ABTS+, with clearance rates exceeding 80% within a defined concentration range. UPLC-IMS-QTOF-MS analysis identified 21 phenolic compounds. Besides, a machine learning model based on DPPH radical scavenging activity was developed, enabling the prioritisation of 10 compounds with high predicted activity. SHAP-based interpretability analysis revealed that aromatic moieties bearing phenolic hydroxyl groups were key structural determinants contributing positively to antioxidant activity. This provides an efficient DES extraction method and mechanistic insights into antioxidant structure–activity relationships, supporting future utilisation.