机器学习在中药寒热药性研究中的应用进展  

Application progress of machine learning in study on cold and hot properties of Chinese materia medica

作　　者：王佳柔 周璐 袁慧敏 生玉涵 张雅绮 汤阳 孙燕 郑丰杰 李宇航 Wang Jiarou;Zhou Lu;Yuan Huimin;Sheng Yuhan;Zhang Yaqi;Tang Yang;Sun Yan;Zheng Fengjie;Li Yuhang(Beijing University of Chinese Medicine,Beijing 100029,China;Traditional Chinese Medicine(Zhong Jing)School,Henan University of Chinese Medicine,Zhengzhou 450046,China;School of Traditional Chinese Medicine,Beijing University of Chinese Medicine,Beijing 100029,China)

机构地区：[1]北京中医药大学,北京100029 [2]河南中医药大学中医学院(仲景学院),郑州450046 [3]北京中医药大学中医学院,北京100029

出　　处：《国际中医中药杂志》2025年第3期423-428,共6页International Journal of Traditional Chinese Medicine

基　　金：北京中医药大学纵向发展基金(90020172120017)。

摘　　要：中药药性理论的科学诠释是中药现代化研究热点。明确单味中药寒热药性及寒热程度,对指导临证精准用药有重要价值。近年来,中药寒热药性研究多在动物、细胞及分子水平展开;立足药性客观物质基础,从热力学、多组学等生物学效应角度出发;借助红外热成像等技术进行分析;形成“性-构关系”等多种研究模式。相关研究从单一物质成分或指标,发展到趋于整合多源信息、多维数据的新模式。但如何处理数据样本量巨大、冗余性强、异质性高等问题,以及如何整合多维信息等问题仍为研究难点。机器学习可凭借其强大的计算和学习能力,在中药寒热药性研究中展现出良好的判别、预测能力,在中药寒热药性研究中发挥重要作用。目前应用较多的算法有线性判别分析、Logistic判别分析、支持向量机、决策树、随机森林等。现有研究的数据维度有待丰富,算法有继续优化空间,并有待建立更为细致的中药寒热药性判别模型。The scientific interpretation of the theory of medicinal properties of TCM is a research hotspot in the modernization of TCM.It is of great value to clarify the property and degree of cold and heat in Chinese materia medica for guiding clinical precise medication.In recent years,the research on the cold and heat properties of Chinese materia medica has been carried out at the animal,cell and molecular levels.Based on the objective material basis of medicinal properties,from the perspective of biological effects such as thermodynamics and multiomics;with the help of infrared thermal imaging and other technologies for analysis;forming a variety of research models such as"property-structure relationship".Related research has developed from a single material component or index to a new model that tends to integrate multi-source information and multi-dimensional data.However,how to deal with the problems of large sample size,strong redundancy,high heterogeneity,and how to integrate multi-dimensional information are still research difficulties.With its powerful computing and learning ability,machine learning can show good discrimination and prediction ability in the study of cold and hot properties of Chinese materia medica,and play an important role in the study of cold and hot properties of Chinese materia medica.At present,the most widely used algorithms are linear discriminant analysis,Logistic discriminant analysis,support vector machine,decision tree,random forest and so on.The data dimension of the existing research needs to be enriched,the algorithm has room for further optimization,and a more detailed discriminant model of cold and hot properties of Chinese materia medica needs to be established.

关 键 词：四气(中药) 寒热药性 机器学习 综述 

分 类 号：R28[医药卫生—中药学]