A calibration method based on model updating strategy for the quantitative model of Radix Astragali extract

Abstract

The main purpose of this study was to construct a calibration model based on model updating strategy for the quantitative analysis of three active components in Radix Astragali extract (RAE). To verify the feasibility of quantitative model by a small portable near-infrared (NIR) spectrometer, a comparison with Fourier Transform near-infrared (FT-NIR) spectrometer was also carried out. Partial least square regression (PLSR) models were established for quantitation of three ingredients including astragaloside IV (AST IV), calycosin-7-glucoside (CG) and astragaloside polysaccharide (APS) for quality control and evaluation of RAE. For new samples, a model-based update strategy was proposed to correct the quantitative models for reducing the influence of the manufacturer and extraction method on the NIR model. A novel Cluster Center Distance (CCD) method was also proposed to select the representative candidate samples added to the calibration set for updating the calibration model which was more suitable for new samples, and three sample-selection methods such as Random sampling (RS), Kennard-Stone (KS) and joint x-y distances (SPXY) were used for comparison. The results indicated that the established calibration method based on a model updating strategy might improve the model's adaptability to new samples significantly. The proposed CCD method had certain advantages in model update and could improve the model's accuracy in predicting the new samples.