采用误差反传前向人工神经网络（artificial neural network,ANN）建立了24种取代芳烃的结构与其对发光菌的急性毒性之间的定量关系模型（ANN模型）。以24种取代芳烃的量子化学参数作为输入,急性毒性作为输出,采用内外双重验证的办法分析和检验所得模型的稳定性和外部预测能力。所构建网络模型的相关系数为0.9834、交叉检验相关系数为0.9780、标准偏差为0.11、残差绝对值≤0.33,应用于外部预测集,外部预测集相关系数为0.9955;而多元线性回归（multiple linear regression,MLR）法模型的相关系数为0.9786、标准偏差为0.12、残差绝对值≤0.36,外部预测集相关系数为0.9904。结果表明,ANN模型获得了比MLR模型更好的拟合效果。%The systematic study of the quantitative structure - activity relationship (QSAR) on 24 substituted aro- matic compounds was performed by the artificial neural network based on the back propagation algorithm. For the artificial neural network method, when using the quantum chemical parameters about structure as the inputs of the neural network and the acute toxicities as the outputs of the neural network, the correlation coefficient was 0. 9834, the leave one out cross - validation regression coefficient was 0. 9780, the standard error was 0. 11, the correlation coefficient of the test set was 0. 9955 and the absolute values of residual were less than 0. 33. In order to make contrast, the QSAR model was set up by multiple linear regressions (MLR) method. For the model built by MLR, the correlation coefficient was O. 9786, the standard error was 0. 12, the absolute values of residual were less than 0.36 and the correlation coefficient of the test set was 0. 9904. The results showed that the performance of neural network method was better than that of MLR method.
展开▼
