Study on the modeling method of general model of Yaobitong capsule intermediates quality analysis based on near infrared spectroscopy
10.16438/j.0513-4870.2024-0955
- VernacularTitle:基于近红外光谱技术的腰痹通胶囊中间体质量分析通用模型建模方法研究
- Author:
Le-ting SI
1
;
Xin ZHANG
2
;
Yong-chao ZHANG
2
;
Jiang-yan ZHANG
2
;
Jun WANG
2
;
Yong CHEN
3
;
Xue-song LIU
3
;
Yong-jiang WU
3
Author Information
1. Department of Pharmacy, Women's Hospital School of Medicine Zhejiang University, Hangzhou 310006, China
2. State Key Laboratory on Technologies for Chinese Medicine Pharmaceutical Process Control and Intelligent Manufacture (Jiangsu Kanion Pharmaceutical Co., Ltd.), Lianyungang 222047, China
3. College of Pharmaceutical Sciences, Zhejiang University, Hangzhou 310058, China
- Publication Type:Research Article
- Keywords:
Yaobitong capsule;
intermediate;
near infrared spectroscopy;
general model;
ant lion optimization least squares support vector machine
- From:
Acta Pharmaceutica Sinica
2025;60(2):471-478
- CountryChina
- Language:Chinese
-
Abstract:
The general models for intermediates quality analysis in the production process of Yaobitong capsule were established by near infrared spectroscopy (NIRS) combined with chemometrics, realizing the rapid determination of notoginsenoside R1, ginsenoside Rg1, ginsenoside Re, ginsenoside Rb1, ginsenoside Rd and moisture. The spray-dried fine powder and total mixed granule were selected as research objects. The contents of five saponins were determined by high performance liquid chromatography and the moisture content was determined by drying method. The measured contents were used as reference values. Meanwhile, NIR spectra were collected. After removing abnormal samples by Monte Carlo cross validation (MCCV), Monte Carlo uninformative variables elimination (MC-UVE) and competitive adaptive reweighted sampling (CARS) were used to select feature variables respectively. Based on the feature variables, quantitative models were established by partial least squares regression (PLSR), extreme learning machine (ELM) and ant lion optimization least squares support vector machine (ALO-LSSVM). The results showed that CARS-ALO-LSSVM model had the optimum effect. The correlation coefficients of the six index components were greater than 0.93, and the relative standard errors were controlled within 6%. ALO-LSSVM was more suitable for a large number of samples with rich information, and the prediction effect and stability of the model were significantly improved. The general models with good predicting effect can be used for the rapid quality determination of Yaobitong capsule intermediates.
