采用傅里叶变换红外光谱法（Fourier transform infrared spectroscopy, FTIR）、 偏最小二乘判别分析（partial least square discriminant analysis, PLS-DA）和系统聚类分析（hierarchical cluster analysis, HCA）快速鉴别不同产地川东獐牙菜（Swertia davidi Franch）。 采集4个不同地区70株样品不同部位的红外光谱数据, 原始光谱数据经预处理（自动基线校正, 自动平滑, 一阶求导, 二阶求导）后导入OMNIC 8.2, 比较吸收峰的差异; 用SIMCA-Pa+10.0进行偏最小二乘判别分析（PLS-DA）, 以前三个主成分三维得分图比较产地鉴别效果; 红外光谱数据导入SPSS 19.0, 进行系统聚类分析（HCA）, 通过树状图比较不同部位分类效果。 结果显示, （1）不同产地根的光谱图在1 739, 1 647, 1 614, 1 503, 1 271, 1 243, 1 072 cm-1附近的吸收峰有差异, 不同产地茎的光谱图在1 503, 1 270, 1 246 cm-1吸收峰附近有差异; （2）相同产地不同部位的光谱特征峰有差异; （3）PLS-DA分析结果显示自动基线校正+自动平滑+二阶求导这种预处理方式分类效果最好, 根的红外光谱数据产地鉴别效果最佳; （4）HCA的树状图, 显示根的聚类分析结果正确率83%, 茎的聚类分析结果正确率49%, 叶的聚类分析结果正确率70%。 FTIR技术结合PLS-DA与HCA方法能够快速准确地鉴别不同产地川东獐牙菜, 不同部位产地鉴别效果有差异, 根的光谱数据产地鉴别效果最佳, 二阶求导处理增强了样品的特异性, 使PLS-DA的三维主成分得分图分类效果更明显。

Fourier transform infrared spectroscopy combined with partial least squares discriminate analysis (PLS-DA) and hierarchical cluster analysis (HCA) were used to rapidly discriminate the Swertia davidi Franch which collected from different origins. The original infrared spectra data of different parts of all the 70 samples which collected from four different regions were preprocessed by automatic calibration, automatic smoothing, the first derivative and the second derivative. Then the processed data were imported into OMNIC 8.2 and the absorption peaks were compared; PLS-DA was performed by SIMCA-P+ 10.0 and the effect of discrimination of different origins was compared by 3D score plot of the first three principal components; the infrared spectral data were imported into SPSS 19.0 for HCA to compare classification results of different parts by the dendrogram. The results showed that: (1) There were differences among the spectra of the roots of different origins in the spectral peaks in 1　739, 1　647, 1　614, 1　503, 1　271, 1　243, 1　072 cm-1. The spectra of the stems of different origins showed differentiation in the wavelength in 1　503, 1　270, 1　246 cm-1; (2) The characteristic peaks of different parts of the same origin were different; (3) PLS-DA indicated that the data which were processed by automatic correction, automatic smoothing and second derivative have showed the best classification. In addition, the discrimination of roots which collected from different origins could be the best; (4) Tree diagram of HCA showed that the accuracy rate of cluster in roots, stems and leaves were 83%, 56%, and 70%, respectively. In conclusion: FTIR combined with PLS-DA and HCA can rapidly and accurately differentiate S. davidi that collected from different origins, the origin discrimination effect of different parts was clearly different that the classification of roots is the best, the second derivative could enhance the specificity of the samples, the classification in 3D score plot could be visualized and obvious.