基于R语言的近红外光谱对甘草中指标成分定量分析

Quantitative Analysis of Index Components in Glycyrrhizae Radix et Khizoma by Near Infrared Spectroscopy Based on R Software

目的:建立较优的甘草质控成分(水分、总灰分、甘草苷、甘草酸)的近红外定量模型,实现快速检测。方法:基于2015年版《中国药典》方法测定97批甘草中水分、总灰分、甘草苷及甘草酸的含量。采用近红外光谱仪扫描其近红外光谱。采用R语言筛选最佳光谱预处理方法,建立近红外定量模型。结果:水分和甘草苷近红外定量模型的最佳预处理方法为一阶导数,其中水分测试集和验证集的相关系数分别为0.930 0和0.929 9,均方根误差分别为0.243 2和0.203 8,甘草苷测试集和验证集的相关系数分别为0.930 3和0.907 6,均方根误差分别为0.093 9和0.128 9;总灰分近红外定量模型的最佳预处理方法为MSC,测试集和验证集的相关系数分别为0.926 5和0.917 7,预测均方根误差分别为0.109 6和0.103 7;甘草酸近红外定量模型的最佳预处理方法为SNV,测试集和验证集的相关系数分别为0.918 1和0.915 7,预测均方根误差分别为0.274 8和0.236 0。结论:该研究建立了较优的甘草质控成分的近红外定量模型,其模型的准确度均较高,为实现快速检测奠定了基础。

Objective: To establish a better near infrared quantitative model for quality control of Glycyrrhizae Radix et Rhizoma of components( moisture,total ash,liquiritin and glycyrrhizic acid) in liquorice,in order to realize rapid detection. Method: The contents of moisture,total ash,liquiritin and glycyrrhizic acid were determined in 97 samples based on the methods set forth in Chinese Pharmacopoeia. Meanwhile,the near infrared spectrum was scanned using near infrared spectroscope. R software was used to screen out the spectral pretreatment and build the quantitative models. Result: The optimum spectral pretreatment method for establishing the near infrared quantitative model of moisture and liquiritin was the first order derivative. For moisture, the correlation coefficients of test and validation were 0. 930 0 and 0. 929 9,and the root mean square errors were 0. 243 2 and 0. 203 8,respectively. For liquiritin,the correlation coefficients of test and validation were 0. 930 3 and 0. 907 6,and the root mean square errors were 0. 093 9 and 0. 128 9,respectively. The optimum spectral pretreatment method for establishing the near infrared quantitative model of total ash was MSC. The correlation coefficients of test and validation were 0. 926 5 and 0. 917 7,and the root mean square errors were 0. 109 6 and 0. 103 7,respectively. The optimum spectral pretreatment method for establishing the near infrared quantitative model of glycyrrhizic acid was SNV. The correlation coefficients of test and validation were 0. 918 1 and 0. 915 7,and the root mean square errors were 0. 274 8 and 0. 236 0,respectively. Conclusion: In this study,a better near infrared quantitative models for quality control of components of Glycyrrhizae Radix et Rhizoma were established,with a high accuracy,which laid a foundation for rapid detection of the components in Glycyrrhizae Radix et Rhizoma.