基于GC-IMS及机器学习的不同炒制程度山楂饮片的快速鉴别
Identification of Crataegi Fructus decoction pieces under different stir-frying degrees with GC-IMS and machine learning
刘纹纹 1董红敬 2张敏敏 3刘双 3马鑫慧 1王珍强 1王晓2
作者信息
- 1. 山东中医药大学药学院,山东济南 250300;齐鲁工业大学(山东省科学院)山东省分析测试中心山东省大型精密分析仪器应用技术重点实验室,山东济南 250014
- 2. 山东中医药大学药学院,山东济南 250300;齐鲁工业大学(山东省科学院)山东省分析测试中心山东省大型精密分析仪器应用技术重点实验室,山东济南 250014;齐鲁工业大学(山东省科学院)药学院山东省高等学校天然药物活性成分研究重点实验室,山东济南 250014
- 3. 齐鲁工业大学(山东省科学院)山东省分析测试中心山东省大型精密分析仪器应用技术重点实验室,山东济南 250014;齐鲁工业大学(山东省科学院)药学院山东省高等学校天然药物活性成分研究重点实验室,山东济南 250014
- 折叠
摘要
目的:构建不同炒制程度山楂饮片的快速鉴别方法.方法:采用气相色谱-离子迁移谱(gas chromatography-ion mobility spectrometry,GC-IMS)分析不同炒制程度山楂饮片中的挥发性成分及其含量,采用偏最小二乘判别分析(partial least squares discriminant analysis,PLS-DA)、岭回归和弹性网络3种数据分析方法进一步筛选具有差异的特征性成分;基于特征性成分,采用7种机器学习算法开展不同炒制程度山楂饮片的鉴别与构建区分模型.结果:通过GC-IMS从不同炒制程度的山楂饮片中共检测出47种挥发性成分,包含10个醇类、9个醛类、8个酯类、6个杂环类、5个酮类、4个有机酸类、2个烃类、2个不饱和烃类和1个酚类;结合PLS-DA、岭回归及弹性网络3种数据分析方法共筛选出6个特征性成分;7种机器学习算法的预测结果显示,支持向量机径向核函数(support vector machine radial kernel,SVM-R)和朴素贝叶斯(naive Bayes,NB)具有较好的预测能力,可用于不同炒制程度山楂饮片的快速鉴别与区分.结论:本研究为不同炒制程度山楂饮片的快速鉴别与区分提供了一种简便、快速的方法,同时可为山楂及其炮制品质量评价体系的建立提供参考.
Abstract
OBJECTIVE To develop a method for rapid identification of Crataegi Fructus decoction pieces under different stir-frying degrees.METHODS Gas chromatography-ion mobility spectrometry(GC-IMS)was employed for identifying the contents of volatile compounds in C.Fructus decoction pieces under different stir-frying degrees.Three data analytic methods of partial least squares discriminant analysis(PLS-DA),ridge regression and elastic network were employed for further screening for featured differential compounds.Based upon the featured compounds,machine learning algorithms were utilized for constructing models for identifying and discriminating C.Fructus decoction pieces under different stir-frying degrees.RESULTS A total of 47 volatile compounds were detected from C.Fructus decoction pieces under different stir-frying degrees by GC-IMS,including 10 alcohols,9 aldehydes,8 esters,6 heterocycles,5 ketones,4 organic acids,2 hydrocarbons,2 unsaturated hydrocarbons and 1 phenolic.Six featured compounds were selected by combining the data analytic methods of PLS-DA,ridge regression and elastic network.Finally,among 7 machine learning models,SVM-R and NB demonstrated the best prediction capability.It could be used to quickly identify and discriminate C.Fructus decoction pieces under different stir-frying degrees.CONCLUSION This study provides a simple and quick method of quickly identifying and discriminating C.Fructus decoction pieces under different stir-frying degrees.Also it offers references for establishing their quality evaluation methods.
关键词
山楂/炒制程度/机器学习/挥发性成分/气相色谱-离子迁移谱Key words
crataegi fructus/stir-frying degree/machine learning/volatile compounds/GC-IMS引用本文复制引用
基金项目
山东省重点研发计划项目(2021CXGC010508)
山东省泰山学者项目(tstp20221138)
出版年
2024
NETL
NSTL
万方数据