近红外漫反射光谱法结合CP-ANN和PLS高通量分析草麻黄药材

目的: 建立草麻黄药材的近红外漫反射光谱高通量分析方法。 方法: 测量草麻黄样品的近红外漫反射光谱(near infrared diffuse reflectance spectra,NIRDRS),应用化学计量学技术进行光谱处理和数据预处理,分别建立并验证草麻黄药材的产地和采摘时间判别对向传播人工神经网络(counter-propagation artificial neural network,CP-ANN)模型及麻黄碱和伪麻黄碱含量预测偏最小二乘(partial least square,PLS)模型。 结果: 草麻黄药材的产地和采摘时间判别CP-ANN模型的验证样品预测准确率分别为100.0%和80.0%;麻黄碱和伪麻黄碱含量预测PLS模型的验证样品预测均方根误差(root mean square errors of prediction,RMSEPs)小,分别为1.12和0.236,预测值与参考值的相关系数(correlation coefficients)大,分别为0.9721和0.9309。 结论: 采用所建方法能同时对草麻黄药材的产地和采摘时间进行准确判别,对其麻黄碱和伪麻黄碱的含量进行准确预测。该方法准确、快速,无需特殊的样品处理,也不使用化学试剂。

Objective: To establish a high-throughput method for the analysis of Ephedra sinica plants by near infrared diffuse reflectance spectroscopy combined with chemometric techniques. Methods: The near infrared diffuse reflectance spectra (NIRDRS) for Ephedra sinica plants were determined.Two types of models were built and validated after spectra processing and data pre-processing,including the counter-propagation artificial neural network (CP-ANN) models for the discriminations of habitats and harvest times of Ephedra sinica plants and the partial least square (PLS) models for the determinations of ephedrine and pseudoephedrine. Results: The prediction accuracies of the CP-ANN models for the validation samples were 100.0% for the habitats and 80.0% for the harvest times.For the validation samples,the root mean square errors of prediction (RMSEPs) of the PLS models were 1.12 and 0.236,the correlation coefficients of the prediction and reference values 0.9721 and 0.9309,separately for ephedrine and pseudoephedrine. Conclusions: The proposed approach could simultaneously identify habitats and harvest times of Ephedra sinica plants and determine ephedrine and pseudoephedrine in the plants.