中国药学(英文版) ›› 2023, Vol. 32 ›› Issue (5): 406-416.DOI: 10.5246/jcps.2023.05.034

• 【研究论文】 • 上一篇    下一篇


苏静华1, 张超1,*(), 孙磊2,*(), 马双成2, 邢以文1   

  1. 1. 苏州市药品检验检测研究中心, 江苏 苏州 215104
    2. 广州大学 化学化工学院 化学基础实验室, 广东 广州 510006
  • 收稿日期:2022-10-27 修回日期:2022-11-13 接受日期:2023-12-24 出版日期:2023-06-02 发布日期:2023-06-02
  • 通讯作者: 张超, 孙磊
  • 作者简介:
    + Tel.: +86-512-68226135, E-mail:
  • 基金资助:
    Pharmaceutical Science Foundation of Jiangsu Medical Products Administration (Grant No. 202128).

Establishment of a variety identification system for Fritillaria Thunbergii Bulbus and Fritillaria Cirrhosae Bulbus based on nucleosides and nucleobases with multivariate analysis and pattern recognition

Jinghua Su1, Chao Zhang1,*(), Lei Sun2,*(), Shuangcheng Ma2, Yiwen Xing1   

  1. 1 Suzhou Institute for Drug Control, Suzhou 215104, Jiangsu, China
    2 Chemical Basic Laboratory, School of Chemistry and Chemical Engineering, Guangzhou University, Guangzhou 510006, China
  • Received:2022-10-27 Revised:2022-11-13 Accepted:2023-12-24 Online:2023-06-02 Published:2023-06-02
  • Contact: Chao Zhang, Lei Sun


浙贝母和川贝母因其具有特色药理活性, 作为中医药已经在临床上使用上千年。本文利用高效液相色谱法, 使用多元统计学分析手段和模式识别技术对浙贝母和川贝母的鉴别进行研究。共收集49份样品, 10种核苷类成分被用于特征组分进行分析测定, 采用聚类分析、主成分分析、最小二乘回归分析等多元分析方法评价品种和特征组分的相关性。通过建模, 利用不同的算法如k-近邻法、偏最小二乘判别、簇类独立软模式法和支持向量机法对未知样本进行品种预测。结果表明, 基于此建立的模式识别模型, 具有较好的预测准确率。结合多元统计分析方法和模式识别技术可以提高在特殊情况下浙贝母和川贝母的品种预测准确率, 有利于质量控制和产品监督。

关键词: 川贝母, 浙贝母, 多元分析, 模式识别, 核苷, 碱基


Fritillaria Thunbergii Bulbus and Fritillaria Cirrhosae Bulbus have been used as traditional Chinese medicine for thousands of years due to their pharmacological activities. In the present study, multivariate analysis and pattern recognition were applied for the identification of Fritillaria Thunbergii Bulbus and Fritillaria Cirrhosae Bulbus by HPLC. A total of 49 samples from two varieties were collected, and 10 nucleosides and nucleobases were chosen for analysis. Multivariate analyses, such as hierarchical cluster analysis, principal component analysis, and partial least square regression analysis, were used to reveal the correlation between its components and varieties. Moreover, different algorithms in pattern recognition, such as k-nearest neighbor, partial least squares discrimination, soft independent modeling of class analogies, and support vector machine, were applied to identify the species of unknown samples. Results by pattern recognition suggested that the prediction accuracy was satisfactory. A combination of multivariate analysis and pattern recognition was more reasonable, improving the analytical accuracy of variety identification.

Key words: Fritillaria Thunbergii Bulbus, Fritillaria Cirrhosae Bulbus, Multivariate analysis, Pattern recognition, Nucleosides, Nucleobases