中药当归和独活的数字化鉴定
投稿时间:2024-01-16  修订日期:2024-02-26   点此下载全文
引用本文:
DOI:
摘要点击次数: 76
全文下载次数: 0
                                      
作者中文名作者英文名单位中文名单位英文名E-Mail
付娆 furao 中国食品药品检定研究院 National institutes for food and drug control 1035379479@qq.com 
张佳婷 zhangjiating 中国食品药品检定研究院 National institutes for food and drug control 1837433515@qq.com 
贺方良 hefangliang 中国食品药品检定研究院 National institutes for food and drug control hefangliang0602@126.com 
王献瑞 wangxianrui 中国食品药品检定研究院 National institutes for food and drug control niuyun006097@163.com 
郭晓晗 guoxiaohan 中国食品药品检定研究院 National institutes for food and drug control guoxiaohan@nifdc.org.cn 
荆文光 jingwenguang 中国食品药品检定研究院 National institutes for food and drug control jingwenguang@nifdc.org.cn 
李明华 liminghua 中国食品药品检定研究院 National institutes for food and drug control liminghua@nifdc.org.cn 
余坤子 yukunzi 中国食品药品检定研究院 National institutes for food and drug control ykz@nifdc.org.cn 
杨建波 yangjianbo 中国食品药品检定研究院 National institutes for food and drug control yjb@nifdc.org.cn 
程显隆 chengxianlong 中国食品药品检定研究院 National institutes for food and drug control cxl@nifdc.org.cn 
魏锋* weifeng 中国食品药品检定研究院 National institutes for food and drug control weifeng@nifdc.org.cn 
张明童 zhangmingtong 甘肃省食品检验研究院 Gansu Food Inspection and Research Institute 519815751@qq.com 
马潇 maxiao 甘肃省食品检验研究院 Gansu Food Inspection and Research Institute 2484649834@qq.com 
基金项目:国家重点研发计划“中医药现代化”重大专项(2023YFC3504105);中药材及饮片质量控制重点实验室项目(2022GSMPA-KL03)
中文摘要:目的:基于UPLC-QTOF-MS分析并经量化处理,实现中药当归和独活的数字化鉴定分析。方法:首先利用UPLC-QTOF-MS技术对中药当归和独活进行分析,然后利用Progenesis QI软件进行峰位校正和提取,并将中药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,同时采用交叉验证和对所建模型进行分析,优选最佳模型用于中药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,准确率和精确率为100%,ROC曲线下面积为1.00,且经外部验证能够准确鉴定中药当归和独活。结论:基于UPLC-QTOF-MS量化数据,结合ANNs算法能够高效准确的实现当归和独活的数字化鉴定,该方法可为实现中药数字化鉴定分析提供参考和帮助。
中文关键词:当归  独活  机器学习  特征筛选  数字化  UPLC-QTOF-MS  
 
Study on Digital Identification of Angelicae Sinensis Radix and Angelicae Pubescentis Radix based on UPLC-QTOF-MS analysis
Abstract:Objective: in order to realize digital identification of Angelicae Sinensis Radix and Angelicae Pubescentis Radix based on UPLC-QTOF-MS analysis and quantized processing. Methods: Firstly, UPLC-QTOF-MS technique was utilized to analyze Angelicae Sinensis Radix (ASR) and Angelicae Pubescentis Radix (APR). Then Progenesis QI software was used to perform peak correction and extraction and the mass spectra of ASR and APR were convert into data matrices (retention time_mass-to-charge ratio-ionic intensity). Further feature screening was performed based on the information gain and the rate of information gain, and combined with artificial neural networks (ANNs), support vector machine (SVM), logistic regression (LR), and K Nearest Neighbors (KNN) machine learning algorithms to establish data identification models. At the same time, cross validation of the constructed models were used to screen the best model for external verification and digital identification of ASR and APR. Results: A total of 603 feature data variables were obtained through feature screening. Compared with SVM, LR and KNN algorithm models, the identification model based on screened feature data and ANNs algorithm had the best identification effect with an accuracy and precision rate of 100% and an area under ROC curve of 1.00, and it can accurately identify Angelicae Sinensis Radix (ASR) and Angelicae Pubescentis Radix (APR) by external verification. Conclusion: Based on UPLC-QTOF-MS quantized data and ANNs algorithm, the digital identification of Angelica sinensis and Radix Angelicae Pubescentis can be realized efficiently and accurately, which can provide reference and assistance for the digital identification and analysis of traditional Chinese medicine.
keywords:Angelicae Sinensis Radix  Angelicae Pubescentis Radix  machine learning  feature screening  digitization  UPLC-QTOF-MS
查看全文   查看/发表评论  下载PDF阅读器