Optimization of ultrasonic extraction and kinetic modeling of jieduquyuziyin prescription polysaccharides via explainable machine learning.

Abstract

Plant polysaccharides, due to their unique pharmacological activities, are highly promising candidates in drug development. However, different extraction methods and processes have significant effects on the structure, yield, and pharmacological activity of plant polysaccharides. In this study, taking Jieduquyuziyin prescription polysaccharides (JPP) as an example, we optimized its ultrasonic-assisted extraction process using response surface methodology and two explainable machine learning (ML) models (random forest and artificial neural networks). In addition, the extraction kinetic equation of JPP was established, and by comparing it with the prediction results of the two ML models, it was ultimately confirmed that the JPP extraction conditions and kinetic model predicted by the RF model were optimal. Structural analysis results showed that JPP had a rough surface and porous internal structure, and contains various monosaccharides such as glucose (65.25 mol%) and galactose (28.59 mol%). Finally, preliminary experiments confirmed that JPP exhibits in vitro antioxidant activity. This provides a certain foundation for the large-scale development and application of JPP.