除草剂对羊头鲦鱼急性毒性的QSTR研究

QSTR Study for Acute Toxicity of Herbicides to Sheephead Minnows

除草剂会对人类和环境中的非目标生物产生急性或慢性影响,因此对除草剂的使用越来越受到监管机构的关注.使用多元线性回归(MLR)和3种非线性算法(SVM、PPR以及RF)建立定量结构-毒性关系(QSTR)模型,预测除草剂对于羊头鲦鱼(sheepshead minnow)的急性毒性.结果表明,随机森林(RF)模型在所有模型中表现最优异,测试集有较高的R^(2)值(0.901),且RMSE值(0.360)和MAE值(0.199)较低.此外,机理分析表明除草剂对羊头鲦鱼的急性毒性主要与分子的亲脂性、极性和电荷分布等特征有关.

The use of herbicides,which can cause acute or chronic effects on non-target organisms in humans and the environment,is receiving increasing attention from regulatory agencies.In this study,quantitative structure-toxicity relationship(QSTR)models are established using multiple linear regression(MLR)and three nonlinear algorithms(SVM,PPR and RF)to predict the acute toxicity of herbicides to Sheepshead minnow.The results show that RF model gives a higher R^(2)value of 0.901,lower RMSE value of 0.360 and MAE value of 0.199,respectively.In addition,mechanistic analysis shows that the acute toxicity of herbicides to Sheepshead minnow is mainly affected by molecular lipophilicity,polarity and charge distribution characteristics of the molecules.