中药化学成分对大鼠心脏毒性的QSAR研究

QSAR Study on Rat Cardiotoxicity of Chemical Component of Chinese Herbs

目的:建立中药化学成分对大鼠心脏毒性的预测方法,为常规的动物实验提供参考,为中药安全性评价提供了新的途径和方法。方法:本文使用Mold 2软件(Version 2.0.0)对收集到的1 034个化学成分进行了分子描述符计算,采用随机森林算法(Random Forest,RF)和支持向量机(Support Vector Machine,SVM)方法对描述符进行筛选。然后使用推进式决策树算法、支持向量机算法、正则化判别分析算法和随机森林算法构建模型,分别计算了模型的准确率和Kappa值,筛选出最优预测模型,并对中药化学成分进行预测。结果:通过比较预测模型的准确度和Kappa值,得出随机森林模型为最优算法模型,准确率为86.3%,Kappa值为0.725。最后,对《中华人民共和国药典》(2010版)记载的有毒中药,例如吴茱萸、北豆根、九里香等包含的化学成分进行了预测,得到一些有意义的结果。结论:定量构效关系模型(Quantitative Structure-Activity Relationship,QSAR)对中药化学成分进行预测研究可以为进一步的实验、临床研究提供重要参考。

In order to provide a new way and method for safety evaluation of Chinese materia medica（CMM） and also to provide a reference for conventional animal experiments, computer toxicity prediction technique and method were established to predict the cardiotoxicity of CMM. Mold2 software（version 2.0.0） was used to calculate molecular descriptors of 1034 chemical components. Then, the random forest（RF） method and the support vector machine（SVM） method were used to screen the descriptors. After that, boosting trees method, SVM, regularized discriminant analysis method and RF method were used to build up prediction model, respectively. Finally, the cardiotoxicity of chemical components was predicted by the quantitative structure-activity relationship（QSAR） model with the best accuracy and Kappa value. The results showed that by comparing the accuracy and Kappa value of prediction model, it was found that the RF model was the optimal algorithm model with 86.3% accuracy and the Kappa value of 0. 725. Through the prediction research on chemical components of Chinese herbs with toxicity recorded in the Pharmacopoeia of People＇s Republic of China（version 2010）, such as Evodia rutaecarpa, North bean root, Murraya incense, some meaningful results had been received. It was concluded that QSAR model on prediction research of chemical components of Chinese herbs provided important references for further experimental studies and clinical researches.