化学学报 ›› 2002, Vol. 60 ›› Issue (2): 328-333. 上一篇    下一篇

研究论文

色谱数据可视化及天然植物药指纹特征发现方法

程翼宇;余杰;吴永江   

  1. 浙江大学化学工程与生物工程学系
  • 发布日期:2002-02-15

A visualization method of chromatographic data for discovering fingerprint feature sof natural herbal medicines

Cheng Yiyu;Yu Jie;Wu Yongjiang   

  • Published:2002-02-15

提出一类色谱分析数据可视化方法,并用于发现天然植物药的化学指纹特征。选取中药材川芎作为典型研究对象,采用核主成分分析和空间投影变换法对色谱分析数据进行预处理,提取特征信息,再利用二维灰度映像对变换后的数据进行可视化表达,发现其化学指纹特征,从而直接反映出药材质量类别间的化学模式差异。将该方法用于辨识34个不同产地及等级的川芎样品,结果令人满意,证明其具有视觉模式分辨优点,是表达隐含特征指纹和辨识复杂化学物质体系的有力工具。

关键词: 色谱数据处理, 指纹分析, 谱图, 川芎, 药用植物, 模式识别, 科学计算可视化, 化学信息学

A novel visualization method of chromatographic data is proposed and applied to discover the chemical fingerprint features of herbal medicines. As an example, Chuan-Xiong was selected for research in this paper. Using Kemel Principal Component Analysis and spatial projection transformation, the chromatographic analysis data were processed and the hidden fingerprint features of the herbal medicine cna be effectively discovered and extracted. Then the transformed data were visualized with two-dimensional grayscale images, which can visually reflect the discrepancy of chemical pattern between different quality classifications of the herbal medicine. This method was used to identify 34 Chuan-Xiong samples from different areas and quality grades, and the results showed that different patterns can be satisfactorily classified with visual sense. It was proved that the method is a powerful tool for discovering and representing the hidden fingerprint features in complicated chemical substance system.

Key words: PROFILE ANALYSIS, SPECTRA, LIGUSTICUM WALLICHII, MEDICINAL PLANT, PATTERN RECOGNITION

中图分类号: