多种分子光谱快速分析黄芩中的黄芩苷含量

摘要

In order to study the adaptability of using different kinds of spectral information for quick analysis of the quality of radix scutellariaes(RS),the contents of baicalin in 73 batches of RS determined by high performance liquid chromatography(HPLC) were used as dependent variables(y),the near infrared spectra(NIRs),ultraviolet-visible spectra(UV-Vis) and multi-source complex spectral(MSCs) information including UV-visible and near infrared spectra were used as independent variables(x).The partial least square regression(PLSR) and a novel method called as keeping same relationship between X and Y space on K nearest neighbors (KNN-KSR for short) were applied to predict the contents of baicalin in the RS samples based on above three kinds of spectral information.By comparing the root mean square error of prediction (RMSEP),the average relative error (MRE) and correlation coefficient(R) between the measured and predicted values in validation set were applied to evaluate prediction precision.Regardless of the types of spectral information,the prediction precision of KNN-KSR method was always better than the PLSR method.The analysis results of baicalin based on NIRs were the best,and those based on the UV-visible spectra were the second.Although the prediction error of baicalin contents was the biggest based on the multi-source complex spectral information,it was still lower than 6％.The error satisfied the requirement of industrial analysis.The multi-source complex spectrometer has the advantages of small volume,light weight,low cost and portability.It is expected to improve the analytical precision of the instrument by optimizing its wavelengths and modeling method,so that it could be adapt to rapid spot acquisition of more herbs,and to analyze and monitor the quality of follow-up products.%为考察根据不同类型光谱信息进行黄芩质量快速分析的适应性,采用高效液相色谱(HPLC)法测定了73批黄芩样品中的黄芩苷含量并作为y值,以各样品的近红外、紫外-可见光谱及包含紫外、可见及近红外的多源复合光谱信息作为x值;根据各类光谱信息分别采用偏最小二乘回归(PLSR)与K最近邻样本保形映射(KNN-KSR)方法进行样品中黄芩苷的预测,根据验证集样本真实值与预测值的均方根偏差(RM-SEP)、平均相对误差(MRE)与相关系数(R)评价预测精度.结果表明,采用KNN-KSR方法根据各类光谱信息预测黄芩苷时,各项指标均优于PLSR方法的结果;其中基于近红外光谱对黄芩苷的分析结果最好,紫外-可见光谱次之,基于多源复合光谱信息对黄芩苷的预测误差最大,但其MRE仍在6％以下,可满足工业分析的精度要求.由于多源复合光谱仪具有体积小、重量轻、成本低及便携等优点,通过优化仪器波长范围及建模方法,有望改进该仪器的分析精度,使之适应更多药材现场采购的快速检测及后续产品的质量分析与监控需求.