以精准化监管为导向的牛黄与其临床急重病症用药中替代品的判别研究
Study on the Discrimination of Bovis Calculus and Its Substitutes in Clinical Medication for Acute and Severe Illnesses Guided by Precise Supervision
石岩 1荆文光 1程显隆 1魏锋 1熊婧1
作者信息
- 1. 中国食品药品检定研究院,北京 102629
- 折叠
摘要
目的:研究实现牛黄、培植牛黄和体外培育牛黄的准确判别,为实现此类型品种的精准化监管提供技术支持.方法:首先使用快速蒸发离子化质谱(REIMS)技术对牛黄、培植牛黄和体外培育牛黄样品进行测定,然后通过机器学习中的逻辑回归模型对测得的样品数据进行特征降维,得到仅有22个特征离子通道的数据.使用特征降维后的数据训练经过超参数优化后的逻辑回归模型.结果:该逻辑回归模型对测试集中3种牛黄类药材样品的判别准确率为0.94,精确度为0.90,F1得分为0.93,AUC值为0.99.结论:REIMS技术和逻辑回归模型相结合的分析方法可快速而准确地实现牛黄、培植牛黄和体外培育牛黄的品种判别,对此类型品种精准化监管提供技术支撑.
Abstract
Objective:To establish a method for the accurate discrimination of Bovis Calculus,Bovis Calculus Cultured and Bovis Calculus Sativus,and provide technical support for the precise supervision of this type of variety.Methods:Firstly,Samples of the Bovis Calculus,Bovis Calculus Cultured and Bovis Calculus Sativus were determined by rapid evaporative ionization mass spectrometry(REIMS).Then,a logistic regression model in machine learning was used to perform feature dimensionality reduction on the measured sample data,resulting in data with only 22 characteristic ion channels.Train a logistic regression model with hyperparameter optimization using data obtained from feature dimensionality reduction.Results:The results of test set validation indicated that the logistic regression model could accurately identify the samples of Bovis Calculus,Bovis Calculus Cultured and Bovis Calculus Sativus with accuracy of 0.94,precision of 0.90,Fl score of 0.93 and AUC of 0.99.Conclusion:To sum up,the method based on REIMS technique and logistic regression model can quickly and accurately achieve simultaneous discrimination and identification of Bovis Calculus,Bovis Calculus Cultured and Bovis Calculus Sativus.This study can provide technical support for precise supervision of this type of variety.
关键词
牛黄/快速蒸发离子化质谱/机器学习/逻辑回归/特征选择Key words
Bovis Calculus/REIMS/machine learning/logistic regression/feature selection引用本文复制引用
基金项目
国家重点研发计划项目(2019YFC1711500)
国家"重大新药创制"科技重大专项(2018ZX09735-006)
中国食品药品检定研究院2022年度关键技术研究基金项目(GJJS-2022-10-2)
中国食品药品检定研究院学科带头人培养基金(2023X10)
出版年
2024
NETL
NSTL
万方数据