论文部分内容阅读
为研究拟除虫菊酯农药对大鼠急性毒性的定量构效关系,以电拓扑状态指数(en)和分子电性距离矢量(mk)表征了20个拟除虫菊酯农药的分子结构,利用最佳变量子集回归方法建立了四元QSAR模型,该模型的相关系数R为0.954。模型经Jackknife法和LOO交互检验证明有良好的稳定性和预测能力。以模型中的四个参数(e7、m28、m33、m42)为人工神经网络输入层,设定4∶3∶1的网络结构,构建人工神经网络的BP算法模型,相关系数和预测能力大有提高,R=0.9999,平均误差仅为0.0105。结果表明,该方法揭示了影响拟除虫菊酯农药对大鼠急性毒性的本质因素,为高效低毒拟除虫菊酯类农药的分子设计提供理论依据。
In order to study the quantitative structure-activity relationship of pyrethroid pesticides against acute toxicity in rats, the molecular structures of 20 pyrethroid pesticides were characterized by the electrical topological index (en) and the molecular electrical distance vector (mk) Subset regression method is used to establish quaternary QSAR model. The correlation coefficient R of this model is 0.954. The model was verified by Jackknife method and LOO cross-validation to have good stability and predictive ability. Taking the four parameters (e7, m28, m33, m42) of the model as the input layer of artificial neural network, the network structure of 4: 3: 1 is set, and the BP algorithm model of artificial neural network is constructed with a large correlation coefficient and predictive ability Increase, R = 0.9999, the average error is only 0.0105. The results showed that this method revealed the essential factors affecting the acute toxicity of pyrethroid pesticides in rats and provided a theoretical basis for the molecular design of pyrethroid insecticides with high efficiency and low toxicity.