The machine learning assisted optimization of ultrasound-synergistic enzymatic extraction and comprehensive evaluation of polysaccharides from Saposhnikovia divaricata

Abstract

Saposhnikovia divaricata (S. divaricata) is a common medicinal material with remarkable medicinal value, but there is limited research on its polysaccharides. This research employed ultrasound-synergistic enzymatic extraction to obtain polysaccharides, while leveraging response surface methodology (RSM) and support vector regression (SVR) models to optimize the extraction procedure. According to the SVR model, the optimal extraction conditions that led to a superior polysaccharide yield of 11.82 % included maintaining a solid-solvent ratio of 1:13 g/mL, conducting the process at an extraction temperature of 45 °C with 420 W of power, adjusting the pH to 4.5, extending the ultrasonic treatment to 35 min, and incorporating a 3 % enzyme dosage. Additionally, a comparative analysis of polysaccharides extracted through different methods assessed their physicochemical characteristics, structure, and antioxidant capacity. The entropy weight method was employed for a comprehensive evaluation of the different extraction techniques. In the final assessment, the ultrasound- synergistic enzymatic extraction emerged as the top-performing technique, achieving the highest composite score. Notably, the antioxidant efficacy was driven primarily by the levels of uronic acid, molecular weight and mannose. This research establishes the theoretical foundation and scientific rationale for the efficient extraction and bioactive application of S. divaricata polysaccharides (SDP).