Online Simultaneous Determination of Astragalus Polysaccharides and Calycosin-7-O-β-D-Glucoside in Astragali Radix Percolate Based on Near-Infrared Spectroscopy Technology

Abstract

As a crucial extraction process in traditional Chinese medicine, quality control of percolation still faces challenges in real-time monitoring methods. To address this challenge, this study focused on the Astragalus percolation process and established an NIRS-based method for synchronous online monitoring of two bioactive markers in Astragalus percolates: Astragalus polysaccharides (APSs) and calycosin-7-O-β-D-glucoside (CG), achieving rapid and nondestructive analysis. In this study, near-infrared (NIR) spectra were collected online at different time points during percolation to determine APS and CG concentrations by means of NIRS technology, with high-performance liquid chromatography (HPLC) and ultraviolet–visible spectrophotometry (UV–Vis) used as reference methods. Two modeling approaches—partial least squares regression (PLSR) and support vector regression (SVR)—were employed to establish quantitative analytical models for these bioactive components, with model performance optimized through spectral preprocessing and feature variable selection. Results demonstrated that SVR-based models achieved superior predictive accuracy compared with PLSR. The optimal APS model showed calibration and validation set R² values of 0.9995 and 0.9874, respectively, while the CG model yielded 0.9811 (calibration) and 0.9632 (validation). Both components exhibited residual prediction deviation (RPD) values exceeding the threshold (RPD > 3), with 6.5349 for APS and 3.8357 for CG, confirming excellent predictive capability. Paired t-test analysis of external test sets (p > 0.05) revealed no statistically significant difference between measured and predicted values, further validating the model's robustness for unknown sample prediction. The concentrations of APS and CG in the Astragalus percolation solution can be simultaneously determined by this method within 30 s, significantly improving analytical efficiency compared with the conventional method (60–80 min per sample), while featuring simple operation, solvent-free consumption, low cost, and pollution-free advantages. This study demonstrates that the combination of NIRS and chemometrics enables real-time monitoring of multiple key substance concentrations during the percolation process. As a green analytical technology, NIRS shows significant potential for improving production efficiency and ensuring product quality consistency.