【摘要】 生物半减期（t1/2）是评价外源化合物在鱼体内蓄积效应的重要参数。实验测定t_1/2的速度慢、成本高,难以满足化学品生态风险评价的需求,需要发展替代实验的模型预测方法。本研究搜集了653种化合物t1/2实测值,采用多元线性回归（MLR）和支持向量机（SVM） 2种方法,建立了鱼体logt1/2的定量构效关系（QSAR）预测模型。MLR模型的校正决定系数（R（adj）²）为0.751,均方根误差(RMSE_train)为0.587,去一法交叉验证系数(Q_LOO²)为0.735,外部验证系数(Q_ext²)为0.682,这表明模型具有较好的拟合度、稳健性和预测能力。SVM模型具有更好的拟合和预测能力(R_adj²=0.839,RMSE_train=0.457,Q_ext²=0.708)。采用Williams法对模型的应用域进行表征。所建模型可用于预测多环芳烃、多氯联苯、多溴联苯醚、有机磷农药、药物等典型化合物,以及其他烷烃、环烷烃、烯烃、醇、醚、酸、酯、酮、含卤素化合物、芳香族化合物、含硫、氮、磷化合物的在鱼体内的logt1/2值。更多还原

【Abstract】 Biological half-life(t_1/2) of chemicals is a key parameter for assessing bioaccumulation of xenobiotics.As the number of synthetic chemicals is huge,it is not practical to experimentally measure their t_1/2 values one by one. It is important to develop alternative methods to predict the t_1/2 values of xenobiotics. In this study,t_1/2 values of 653 chemicals in fish were collected,and multiple linear regression（MLR） and support vector machine（ SVM）methods were adopted to develop quantitative structure-activity relationship（QSAR） models for predicting t_1/2. For the MLR model,the adjusted determination coefficient(R_adj²) is 0.751,the root-mean-square error（ RMSE） is 0.587,leave-one-out cross validated coefficient(Q_LOO²) is 0.735 and external explained variance(Q_ext²) is 0.682,which indicate that the MLR model has high goodness of fit,robustness,and predictive ability. The results of the SVM model also show high goodness of fit and good predictive ability( R_adj.train²= 0.839,RMSEtrain= 0.457,Q_ext²=0.708). The model application domains were characterized by the Williams plot. The obtained models can be used to predict logt_1/2 of chemicals including polycyclic aromatic hydrocarbons,polychlorinated biphenyls,poly brominated diphenyl ethers,pesticides,pharmaceuticals,alkanes,naphthenic hydrocarbons,alkenes,alcohols,ethers,acids,esters,ketones,halogenated compounds,aromatics,organosulfur compounds,organonitrogen compounds and organophosphorus compounds.更多还原