HPLC指纹图谱结合反向传播人工神经网络和判别分析鉴定不同的麻黄药材

Identification of Different Ephedra Plants Using HPLC Fingerprints in Combination with Back-Propagation Artificial Neural Network and Discriminant Analysis

采用HPLC-UV测定36个麻黄药材的指纹图谱,应用化学计量学进行图谱预处理和数据预处理,建立并验证不同种类、不同产地和不同采摘时间的麻黄药材的反向传播人工神经网络(back-propagation artificial neural network,BP-ANN)和判别分析(discriminant analysis,DA)判别模型.研究结果显示,所建BP-ANN模型的预测准确率为83.3%～94.4%、DA模型的性能指标为82.8%～88.5%,可见所建方法能有效判别不同种类、不同产地和不同采摘时间的麻黄药材.该方法基于麻黄药材物质基础的整体性质,判断客观,为其他药材的分析提供了参考.

In this work, the chemical fingerprints of 36 batches of Ephedra plant materials of different species, habitats and picking times were measured using an HPLC system and, after fingerprint processing and data pre-processing with chemometrics, the BP-ANN （back-propagation artificial neural network） models and a DA （discriminant analysis） model were established and validated. The prediction accuracies of the BP-ANN models were in the range of 83.3%-94.4% and the performance indexes of the DA model ranged from 82.8% to 88.5%, indicating that the species, habitats and picking times of Ephedra plants can be identified by the proposed approaches.