基于不同建模方法的硝基苯急性毒性QSAR比较研究  

Acute Toxicity of Nitrobenzene Modeling Method Based on Different QSAR Comparative Study

作　　者：金飙[1] 柴欢[1] 金俏[2] 

机构地区：[1]沈阳化工大学环境与安全工程学院,辽宁沈阳110142 [2]北京石油化工学院化学工程学院,北京102600

出　　处：《沈阳化工大学学报》2016年第2期124-129,共6页Journal of Shenyang University of Chemical Technology

摘　　要：采用量子化学密度泛函方法对36种硝基苯类化合物的电子结构参数进行计算,筛选出影响梨形四膜虫的急性毒性显著的5个变量,分别运用多元线性回归法(MLR模型)和人工神经网络(ANN)法建立其结构参数与毒性数据之间的QSAR模型,并进行对比.建立的MLR模型中的相关系数R=0.849,标准误差σs=0.378;ANN模型中的相关系数R=0.989,标准误差σs=0.098.实验结果表明:硝基苯类化合物对梨形四膜虫急性毒性的ANN模型比M LR模型具有更好的拟合能力,预测效果更好.说明影响硝基苯类化合物梨形四膜虫急性毒性的因素是非线性的.By using quantum chemistry density functional method,the electronic structure parameter of36 kinds of nitrobenzene compounds was calculated. The five significantly variables affecting the acute toxicity of pear-shaped four membrane worm were screened out,respectively using the method of multiple linear regression（ MLR） model and artificial neural network（ ANN） method to establish its structure parameters and comparison. The MLR model is set up,and in it,correlation coefficient R = 0. 849,standard error σs= 0. 378; in ANN model,the correlation coefficient R = 0. 989,standard error σs= 0. 098. Experimental results showthat nitrobenzene compounds of pear-shaped four membrane worm acute toxicity of artificial neural network（ ANN model is better than multiple linear regression with fitting ability,prediction effect is better. That affect the nitrobenzene compounds pear-shaped four membrane worm is nonlinear factors of acute toxicity.

关 键 词：密度泛函 硝基苯 多元线性回归 人工神经网络 定量构效关系(QSAR) 

分 类 号：TQ015.9[化学工程] TP391.9[自动化与计算机技术—计算机应用技术]