平性药有效成分与药性相关性研究

本文精选了50味平性药及50味对照非平性药作为研究对象;从各味药的有效成分入手,利用其所含有的特定的骨架及官能团信息直观地进行了数字化表征;采用模式识别中的支持向量机方法进行了分类研究,挑选最优特征时,总正确率最高可以达到73%,表明了中药的有效成分与其药性之间确实存在一定的相关关系。并且通过对识别率的比较发现,平性药药性的不显著本质上可能是由于其本身所含有效成分更高的分子多样性。本文对探索平性药药性的物质基础具有一定的指导作用。

Study on the correlation between the active ingredients of the mild traditional chinese medicine and their properties

In this paper,50 mild medicines and 50 non-mild medicines as control groups were chosen for modeling study.Starting from a variety of active ingredients of each medicine,digital characterization was carried out using their containing specific backbone types and functional groups.Then,Support Vector Machine,one of the popular pattern recognition methods,was used to do the classification work and gave the best total accuracy of 73% when choosing appropriate features,which indicated the strong relationship between the active ingredients and the property of Chinese medicine.And it was found out that the mild medicines' not obvious property may be due to the higher molecular diversity of their active ingredients than those of non-mild cold and heat ones by comparing their respective recognition ratio.This work may guide the study of the basic factors that underlie the special property of mild medicines.