人工神经网络用于异噻唑啉酮的定量结构-活性关系研究  

QSAR of 3-isothiazolinone compounds using artificial neural network

作　　者：何琴[1] 黄保军[2] 李公春[1] 

机构地区：[1]许昌学院化学化工学院,许昌461000 [2]许昌学院表面微纳米材料研究所,许昌461000

出　　处：《安徽农业大学学报》2012年第2期323-326,共4页Journal of Anhui Agricultural University

基　　金：河南省教育厅自然科学研究计划项目(2009B150023)资助

摘　　要：采用误差反传前向人工神经网络(artificial neural network,ANN)建立了21种2-(4-取代-苯基)-3-异噻唑啉酮类化合物的结构与其抗菌活性之间的定量关系模型(ANN模型),以21种3-异噻唑啉酮类化合物的量子化学参数和拓扑指数作为输入、抗菌活性作为输出,所构建网络模型的交叉检验相关系数为0.991 6、标准偏差为0.080 1、残差绝对值≤0.221,应用于外部预测集,预测集相关系数为0.973 1;而多元线性回归(multiple linearregression,MLR)法模型的相关系数为0.841 8、标准偏差为0.303 9、残差绝对值≤0.636。结果表明,ANN模型获得了比MLR模型更好的拟合效果。The study of the quantitative structure-activity relationship（QSAR） on 21 kinds of 2-（4-substi-tuted-phenyl）-3-isothiazolinones was established by the artificial neural network based on the back propagation algorithm.For the artificial neural network method,the quantum chemical parameters about structure and the mo-lecular topological index were used as the inputs of the neural network,and the antibacterial activities as the out-puts of the neural network.As a result,the leave-one-out cross-validation regression coefficient was 0.9916;the standard error was 0.0801;the correlation coefficient of the test set was 0.9731 and the absolute values of resid-ual were less than 0.221.For comparison,the QSAR model was set up by multiple linear regressions（MLR） method.For the model built by MLR,the correlation coefficient was 0.8418;the standard error was 0.3039 and the absolute values of residual were less than 0.636.The results showed that the performance of neural network method is better than that of MLR method.

关 键 词：异噻唑啉酮 定量结构-活性关系 人工神经网络 抗菌活性 大肠杆菌 

分 类 号：O622[理学—有机化学]