芳香族化合物的快速生物降解性QSBR研究

QSBR on ready biodegradability of aromatic compounds

分别采用线性基团贡献、支持向量机与人工神经网络法对芳香族化合物的快速生物降解性进行定量结构-生物降解关系(QSBR)研究,得到不同基团对芳香族化合物快速生物降解性的贡献值。线性基团贡献法对于训练组和测试组的预测正确率是80.3%和79.2%,总的预测正确率达80.1%;支持向量机的预测正确率分别是84.8%、85.4%和84.9%,而人工神经网络法的预测正确率分别是97.7%、81.2%和95.2%。结果表明,这3种方法的预测效果均较好。

The quantitative structure-biodegradability relationship（QSBR） studies were performed with the ready biodegradability of aromatic compounds by the linear group contribution method,artificial neural network and support vector machine approach based on principal component analysis.The contributions of various groups to ready biodegradation were obtained.The prediction accuracy of the linear group contribution method was 80.3% for the training set,79.2% for the test set,and 80.1% for all compounds.With the support vector machine approach,it was 84.8%,85.4% and 84.9%,respectively;with the neural network approach,it was 97.7%,81.2% and 95.2%,respectively.Results showed that three methods had well prediction results for the training set and the test set.