红外指纹图谱和聚类分析法在赤芍产域分类鉴别中的应用  

The Application of Fingerprint Infrared Spectra and Clustering Analysis in the Discrimination of Geographical Origin of Paeonia lactiflora Pall

作　　者：徐永群 黄昊[1] 周群[1] 周红涛[2] 胡世林[2] 孙素琴[1] 

机构地区：[1]清华大学化学系 [2]中国中医研究院中药研究所,北京100700

出　　处：《分析化学》2003年第1期5-9,共5页Chinese Journal of Analytical Chemistry

基　　金：国家中医药管理局科技重大项目 (国中医药科 2 0 0 1ZDZX0 1)

摘　　要：以赤芍的红外指纹图谱为依据 ,采用主成分分析法对来自 18个产地的赤芍进行了聚类分析。可将18个产地大致分为 6类 ,这一分类与地理位置有较明显的对应关系 ,同一区域内赤芍的性能较为相似 ,可作为传统中医界对赤芍药材质量评价的依据。用径向基函数人工神经网络法预测了 4 5个赤芍样本的产区 ,结果表明 ,径向基函数人工神经网络法具有较强的预测能力 ,用它可鉴别赤芍的产区。可为药材的质量控制提供一个快捷、准确。Based on the fingerprint infrared spectrum of paeonia lactiflora Pall, paeonia lactiflora Pall samples from 18 areas were clustered into 6 classes with principal component analysis. This. result agrees with their geographical origins. The spectra of samples were similar with each other in the same class, which can he considered as the criterion of evaluating the pharmic quality of Paeonia lactiflora Pall. 45 geographical origins were predicted with radial basis function, and it was. demonstrated to be a powerful artiticial neural net ( ANN) method in discrimination. So a rapid, accurate, and operable method in the medicinal materials quality control was developed.

关 键 词：红外指纹图谱 赤芍 产域 分类 鉴别 红外光谱 聚类分析 主成分分析 径向基函数人工神经网络 植物药 

分 类 号：R284[医药卫生—中药学]