60种植物类中药提取物的红外光谱分析及其与寒热药性相关性的模式识别评价研究

对60种植物类中药提取物的红外光谱药性特征标记及其模式识别模型进行评价筛选。利用傅里叶变换红外光谱结合(linear discriminant analysis, LDA), (logistic discriminant analysis, Logistic-DA), (principal component analysis-linear discriminant analysis, PCA-LDA), (partial least-squares discriminant analysis, PLS-DA), (random forest, RF), (support vector machine, SVM)六种模式识别技术进行研究。水提取组采用加热回流提取1.5 h，无水乙醇、氯仿、石油醚提取组采用室温超声提取45 min。首先分别建立六种模式识别模型，然后采用四种统计方法综合识别，包括60味中药组内回代、60味中药10次迭代5折交叉验证、48味中药训练集、12味中药测试集。选取组内回代识别正确率、交叉验证识别正确率、组外预测正确率同时很高，且理论图谱反映寒热中药原始图谱分布特征者为适宜模型。LDA和SVM是水提取物红外光谱的适宜模式识别模型，LDA是无水乙醇提取物红外光谱的适宜模式识别模型，SVM是氯仿提取物红外光谱的适宜模式识别模型，石油醚提取识别效果不佳。结论：根据适宜识别模型，通过红外光谱数据可识别表征中药寒热成分和寒热程度的特征参数，寒热成分特征参数为与红外光谱吸收位置波谱相对应的识别模型的识别系数，识别系数大于零为寒性标记，识别系数小于零为热性标记;寒热程度特征参数为识别模型的识别得分，得分越大(正值)则寒性越强，得分越小(负值)则热性越强。

Analysis of Infrared Spectra of 60 Kinds of Plant Extract of Traditional Chinese Medicine and Study on the Identification and Evaluation of Characteristics of the Regional Markers Associated with Cold and Heat Nature

By using the Fourier transform infrared spectroscopy and linear discriminant analysis(LDA), logistic discriminant analysis(Logistic-DA), principal component analysis-linear discriminant analysis(PCA-LDA), partial least-squares discriminant analysis(PLS-DA), random forest(RF), support vector machine(SVM), infrared spectra of 60 kinds of plant extract of Chinese traditional medicine were analyzed and the identification and evaluation of characteristics of the regional markers associated with cold and heat nature were studied. Results indicated that LDA and SVM are suitable for the recognition model of water extract infrared spectral data, LDA is suitable for the identification model of anhydrous ethanol extract infrared spectral data, SVM is suitable for the identification model of chloroform extract infrared spectral data, while petroleum ether extract group recognition effect is not ideal. According to the suitable characteristic parameters identification model, data were analyzed by infrared spectroscopy, and parameters and resistance characteristics of the traditional Chinese drug composition can be obtained. Regional characteristics of these two parameters can be used to identify drug ingredients, and can also be used to indicate different degrees of resistance characteristics of traditional Chinese medicine. Component parameter is model identification coefficient corresponding to the position of spectrum and infrared, with a value greater than zero it is cold nature marker, while with a value less than zero it is heat nature marker; model identification score is a parameter reflecting the degree of cold and heat nature, the greater the score (positive), the more it is cold, while the smaller the score, the more it is hot.a parameter reflecting the degree of cold and heat,the greater the score (positive) is cold more strong, the score is small (negative) heat stronger.