基于模式识别的唇形科中药寒热药性物质基础研究  

Study on the material basis of cold and hot properties of traditional Chinese medicines in Lamiaceae based on pattern recognition

作　　者：张传耀[1] 周晴 王校[1] 卜菲菲[1] 王鹏 ZHANG Chuan-yao;ZHOU Qing;WANG Xiao;BU Fei-fei;WANG Peng(School of Traditional Chinese Medicine,Anhui University of Chinese Medicine,Hefei 230012,China;School of Medical Information Engineering,Anhui University of Chinese Medicine,Hefei 230012,China;Graduate School,Anhui University of Chinese Medicine,Hefei 230012,China)

机构地区：[1]安徽中医药大学中医学院,安徽合肥230012 [2]安徽中医药大学医药信息工程学院,安徽合肥230012 [3]安徽中医药大学研究生院,安徽合肥230012

出　　处：《药学学报》2023年第2期429-438,共10页Acta Pharmaceutica Sinica

基　　金：国家自然科学基金项目(81673622);安徽省自然科学基金课题(1508085MH202);安徽省高校自然科学研究重点课题(KJ2019A0464).

摘　　要：基于模式识别算法对唇形科中药寒热药性的物质基础进行研究,并建立寒热药性识别模型。首先,建立唇形科中药物质成分数据库。采用物质成分三级分类体系,利用频数分析、关联规则分析、逻辑回归、特征选择等数据挖掘方法得到唇形科中药寒热药性的物质基础,并建立识别模型。卡方检验结果显示,在一级、二级、三级分类水平上,寒热药性的物质成分构成比均有显著差异(P<0.05),差异随物质分类层次而变化,各级分类下平均变异系数分别为42.30%、79.07%、91.51%,占比差异一级分类<二级分类<三级分类。关联规则分析结果显示,在三级分类条件下,有效的关联规则较多。进一步分析得到27个核心类群及34个特异类群,特征选择结果得到15个决定性类群。采用多元逻辑回归分析成功建立了寒热药性识别模型,总体准确度可达到89%。唇形科中药寒热药性的物质基础差异明显,且互有交叉;双环二萜等27种物质为寒热药性的核心类群,其中,15种物质为决定性类群;唇形科中药寒热药性往往由多种类别的物质成分交互作用而体现,单一类别的物质常不能用来表征寒热药性,多种物质类群的有机组合是寒热药性的物质基础。To study the material basis of cold and hot properties of traditional Chinese medicines(TCMs)in Lamiaceae and to establish a cold and hot properties identification model,a database of material components of TCMs in Lamiaceae was established.A three-level classification system of material components was used to obtain the material basis of cold and hot properties of the Lamiaceae family by using data mining methods such as frequency analysis,association rule analysis,logistic regression,and feature selection.Several identification models were established to recognize the cold and hot properties.The chi-square test results showed that the material composition ratios of cold and hot properties were significantly different at the first-level,second-level,and third-level classification(P<0.05),and the differences varied as the levels of substance classification changed.The average coefficients of variation were 42.30%,79.07%,and 91.51%at the first-level,second-level,and thirdlevel classification levels,respectively.In other words,in terms of the percentage differences in material composition ratio,the first-level was smaller than the second-level,and the second-level was smaller than the third-level.The results of the association rule analysis showed that under the third-level classification,there were many effective association rules,and 27 core groups and 34 specific groups of chemical components were obtained based on these rules.15 decisive groups were obtained from the feature selection results.Multinomial logistic regression analysis was used to successfully establish a cold and hot properties identification model with an overall accuracy of 89%.The material basis of cold and hot properties of TCMs in Lamiaceae is different and intersect with each other.Twenty-seven groups of chemical components,such as bicyclic diterpenes,are the core groups of cold and hot properties,of which 15 groups are the decisive groups.The cold and hot properties are often characterized by the interaction of multiple classes of substanc

关 键 词：模式识别 唇形科 中药 寒热药性 物质基础 识别模型 

分 类 号：R917[医药卫生—药物分析学]