Conturction Of QSAR Models For Predicting Biological Half-lives Of Xenobiotics In Fish

With the development of society,more and more xenobiotics produced by human activities are discharged into environment.The aquatic ecosystem plays an important role in the living environment of human beings and fish is a kind of important organism of the aquatic ecosystem.Therefore the assessing of the bioaccumulation of xenobiotics in fish is significant for the evaluation of ecological risks.Biological half-lives(t_1/2)of xenobiotics is one of key parameters for assessing the bioaccumulation.Therefore,the evaluation of the accumulation effect of chemicals in fish is significance to the evaluation of ecological risks.The number of existing registered chemicals has exceeded 140,000,however,the experimental method is slow,costly and difficult to meet the demand for the evaluation of chemical ecological risks assessment.It is necessary to develop model prediction methods for replacing experiments.Based on Quantitative Structure-activity Relationship?QSAR?approach,six QSAR models were developed to predict t_1/2 of chemicals in fish according to the principle of development and validation of QSAR models.The main work is shown as follows:?1?Biological half-lives and 3D structure of 653 xenobiotics were collected.The molecular structure was optimized by Gaussian 09,and Dragon6.0 software was used to calculate the molecular structure descriptors.All chemicals were divided into training set and validation set randomly as the ratio of 2:1.Multiple Linear Regression?MLR?and Support Vector Machine?SVM?were used for developing QSAR models to predict t_1/2/2 of chemicals in fish.The application domains of these models were characterized by the William plot and the meaning of descriptors were interpreted.The universal models had been demonstrated to have higher goodness of fit,robustness and predictive ability.They can be used to predict t_1/2 of chemicals such as polycyclic aromatic hydrocarbons?PAHs?,polychlorinated biphenyls?PCBs?,poly brominated diphenyl ethers?PBDEs?,exogenous pesticides,pharmaceuticals,alkanes,naphthenic hydrocarbons,alkenes,alcohols,ethers,acids,esters,ketones,halogenated compounds,aromatic,sulfur,nitrogen and phosphorus compounds.?2?Polyhalogenated aromatic hydrocarbons contain PCBs,PBDEs,dioxins and other compounds with similar structure.Organophosphorus compounds contain Organophosphorus pesticide and organophosphate.These harmful compounds that accumulated in organisms and ecosystems were not predicted exactly compared with other compounds in universal models.Therefore,it is necessary to develop local QSAR models to predicted t_1/2/2 of these chemicals.119 polyhalogenated aromatic hydrocarbons and 51 Organophosphorus compounds were collected and used to develop the models.The optimization method,calculation method and application domain representation were the same as the universal models.The two local models contained 4 and 8 molecular descriptors respectively,and the goodness of fit,robustness,and predictive ability of which were satisfying.Models could be used to predict t_1/2 of chemicals included in the application domain.