当归和独活的数字化鉴定方法研究
Digital Identification of Angelicae Sinensis Radix and Angelicae Pubescentis Radix
付娆 1张佳婷 1贺方良 2王献瑞 1郭晓晗 1荆文光 1李明华 1余坤子 1杨建波 1程显隆 1魏锋 1张明童 3马潇3
作者信息
- 1. 中国食品药品检定研究院 中药民族药检定所,北京 102629;国家药品监督管理局 药品监管科学全国重点实验室,北京 102629
- 2. 中国食品药品检定研究院 中药民族药检定所,北京 102629;中国药科大学 中药学院,江苏 南京 211198
- 3. 甘肃省食品检验研究院,甘肃 兰州 730000
- 折叠
摘要
目的:基于超高效液相色谱-四级杆飞行时间质谱法(UPLC-QTOF-MS)分析并经量化处理,探索当归和独活的数字化鉴定分析方法.方法:利用UPLC-QTOF-MS对当归和独活进行分析,利用Progenesis QI软件进行峰位校正和提取,并将当归和独活质谱图量化为数据矩阵(保留时间-质荷比-离子强度),进一步基于信息增益和信息增益率进行特征筛选,再结合人工神经网络(ANNs)、支持向量机(SVM)、逻辑回归(LR)、K邻近(KNN)机器学习算法建立数据鉴定模型,同时采用交叉验证对所建模型进行分析,优选最佳模型用于当归和独活的数字化鉴定分析.结果:通过特征筛选得到 603 个特征数据变量,与SVM、LR、KNN算法模型相比,以筛选的特征数据和ANNs算法构建的鉴定模型具有最佳的辨识效果,准确率和精确率均为 100%,工作特征(ROC)曲线下面积为1.000,且经外部验证能够准确鉴定当归和独活.结论:基于UPLC-QTOF-MS量化数据,结合ANNs算法能够高效准确地实现当归和独活的数字化鉴定,该方法可为中药数字化鉴定分析提供参考.
Abstract
Objective:To explore a method for the digital identification and analysis of Angelicae Sinensis Radix(ASR)and Angelicae Pubescentis Radix(APR)based on ultra-performance liquid chromatography-quadrupole time-of-flight mass spectrometry(UPLC-QTOF-MS)analysis and quantization processing.Methods:UPLC-QTOF-MS was utilized to analyze ASR and APR.Progenesis QI software was used for peak correction and extraction,converting the mass spectra of ASR and APR into data matrices(retention time-mass to charge ratio-ionic intensity).Feature screening was further conducted based on information gain and information gain ratio.Data identification models were then established using artificial neural networks(ANNs),support vector machines(SVM),logistic regression(LR),and K-nearest neighbors(KNN)machine learning algorithms.Cross-validation and model analysis were employed to select the best model for the digital identification and analysis of ASR and APR.Results:A total of 603 feature data variables were obtained through feature screening.Compared with the SVM,LR,and KNN algorithm models,the identification model constructed with the screened feature data and ANNs algorithm demonstrated the best recognition effect,with both accuracy and precision rates of 100%and an area under the ROC curve of 1.000.External validation confirmed that the model could accurately identify ASR and APR.Conclusion:The digital identification of ASR and APR can be efficiently and accurately achieved based on UPLC-QTOF-MS quantized data combined with the ANNs algorithm.This method can provide a reference for the digital identification and analysis of Chinese medicine.
关键词
当归/独活/机器学习/特征筛选/数字化/超高效液相色谱-四级杆飞行时间质谱法Key words
Angelicae Sinensis Radix/Angelicae Pubescentis Radix/machine learning/feature screening/digitization/UPLC-QTOF-MS引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据