论文部分内容阅读
以分子电性距离矢量(MEDV-13)表征新型均三氮苯类衍生物的分子结构,通过最佳变量子集回归建立了34种化合物除草活性的QSAR模型,模型的相关系数为0.888。模型通过R_(CV)~2、FIT、VIF等指标检验具有良好的稳健性和预测能力。根据进入模型的3个电性距离矢量m_(15)、m_(56)、m_(91)来看,影响除草剂除草活性的主要因素是分子的-CH_2-、>CH-、-N-和-X等结构片段。以m_(15)、m_(56)、m_(91)为人工神经网络的输入层,设定3:6:1的网络结构,所建BP模型的相关系数为0.976,相关性明显高于多元线性回归模型。结果表明,用电性距离矢量表征均三氮苯类衍生物的除草活性是合理而有效的。
The molecular structure of novel triazine derivatives was characterized by the molecular electrical distance vector (MEDV-13). The QSAR model of 34 herbicidal activities was established by the best subset regression. The correlation coefficient of the model was 0.888. The model has good robustness and predictive ability through R_ (CV) ~ 2, FIT, VIF and other indicators test. According to the three electrical distance vectors m_ (15), m_ (56) and m_ (91) entering into the model, the main factors that affect the herbicidal activity of herbicides are the -CH_2 -, -CH_ -X and other structural fragments. Taking m_ (15), m_ (56) and m_ (91) as the input layer of artificial neural network, the network structure of 3: 6: 1 is set. The correlation coefficient of BP model built is 0.976, which is obviously higher than multivariate Linear regression model. The results show that it is reasonable and effective to characterize the herbicidal activity of all-triazine derivatives by using electrical distance vector.