多元逐步回归对苯胺类化合物结构与毒性模型研究

采用Chemoffice6．0中MOPAC-AMl量子化学法计算了24种苯胺类化合物的6种量子化学结构参数．其中取17个化合物作为样本集对-lgEC50进行多元逐步回归分析．得到最佳方程．经自由度校正的回归系数R=0．985。应用所建立的QSAR模型验证了苯胺类化合物的EC50值．并通过“Jackknife”中的逐一抽取法进行模型检验，得出该模型具有很好的稳定性．平均残差仅为0．05个对数单位．小于文献值。经过7个预测样本对该模型进行验证．结果表明．该模型具有很好的预测能力。同时分析了苯胺类化合物的毒性机理。

An application of multivariate progressive regression analysis to the quantitative structure-toxicity relationships for aniline compounds

With the ChemOffice6.0 software, 6 quantum chemistry parameters for 24 anilines are calculated. The QSAR of 17 anilines is set up which combined with stepwise regress. The regression which is adjusted by free degree is 0.985, is discussed with the relationships of EC\-\{50\} and their structure parameters. And one\|leave\|one test is also used to examine the relationships equation. It is shown that the QSAR model has a good stability, the calculated values with the QSAR model have a good consistent with the experimental values. and it is tested that the model has a good forecasting ability with other 7 compounds. And we also analyse the toxic mechanism of the aniline compounds.