硫代氨基甲酸酯类衍生物抗HIV-1逆转录酶抑制活性的QSAR研究  

QSAR study of thiocarbamates as non-nucleoside HIV-1 reverse transcriptase in hibitors

作　　者：周昌鸿[1] 张运陶[1] 

机构地区：[1]西华师范大学应用化学研究所,四川南充637002

出　　处：《计算机与应用化学》2011年第2期203-208,共6页Computers and Applied Chemistry

摘　　要：采用以MLR为基学习器的Boosting算法模型,对79种硫代氨基甲酸酯类衍生物做抗HIV-1逆转录酶抑制活性的QSAR研究。以E-Dragon软件计算的7组描述符分别为自变量,以化合物的半数效应浓度EC_(50)值为因变量构成7个原始数据集,用PSO算法筛选变量并建立MLR模型。各描述符建立的MLR模型中仅有RDF描述符模型同时通过外部预测和内部验证,故确定以其建立关于硫代氨基甲酸酯类衍生物抗HIV-1逆转录酶抑制活性的Boosting-MLR预测模型。Boosting-MLR模型与MLR模型相比,训练结果的决定系数R^2分别为0.728和0.741,预测结果R^2则分别为0.718和0.667,表明其泛化能力明显增强。对Boosting-MLR模型进一步进行稳定性验证,证明其预测稳定性较高。The QSAR study of 79 thiocarbamates as non-nucleoside HIV-1 reverse transcriptase inhibitors based on Boosting regression combined with MLR.Seven descriptors have been calculated by E-Dragon software as independent variables.The EC_（50） values of compounds are used as dependent variables.Seven models have been constructed by MLR and their variables are selected by particle swarm optimization.By testing the generalization of each model,the values of the internal and external validation of the RDF descriptor model are more than 0.5.So the Boosting-MLR model has been constructed by RDF descriptor.Boosting-MLR model has improved the correlation coefficients（R^2） of prediction to 0.718,up from 0.667 when only MLR was applied.By the model validation,Boosting-MLR has shown the stability generalization.

关 键 词：Boosting MLR HIV-1逆转录酶抑制剂 QSAR 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]