Studies Of QSAR/QSPR For The Risk Assessment Of Chemicals

With the rapid development of chemical industry, a large number of various chemicals are produced and used. Chemicals eventually are distributed in the environment after various ways of using, which strongly influence the environment and human health. Thus, it is very important to assess the security risk of compounds. However, it is a huge project to assay the large number of compounds by the experimental means. So, it is necessary to develop a simple, fast and available approach to measure the security risk of compounds. In this thesis, the quantitative structure-activity/property relationship (QSAR/QSPR) was used to assess the security risk of compounds, to develop the reliable and stable models, to rapidly predict the security risk indice of compounds; on the other hand, from the molecular level, QSAR/QSPR can reveal the structural factors which affect the risk assessment of compounds, further better understand the mechanism of action of the toxicity(ecotoxicity, cytotoxicity) and biodegradation, and while satisfied the needs of people and society, provide the reference information of designing and synthesizing the safer and eco-friendly real green compounds.The chapter one, the basic theory and the procedure of QSAR/QSPR method were mainly introduced. Finally, some applications in the hot area of researches of QSAR/QSPR modeling in recent years were summarized.In the second chapter, by using QSAR approach, the linear genetic function approximation (GFA) and nonlinear least squares support vector machine (LSSVM) models were developed to predict the ecotoxicity of ionic liquids (ILs) towards the marine bacterium Vibrio flscheri.5 descriptors were selected by GFA and used to build the linear model. For the training set, R2= 0.893, RMSEC= 0.402; the test set, R2= 0.903, RMSEP= 0.388. Analyzing the descriptors, the cation structure was the main factor to the toxicity, making the dominant contribution to the ecotoxicity towards Vibrio fischeri, which mainly depended on the size, lipophilic and 3D molecular structure of cations. Moreover, LSSVM model was also built for the prediction of the ecotoxicity of ILs towards Vibrio flscheri. For the training set, R2= 0.910, RMSEC= 0.371; the test set. R2= 0.933. RMSEP= 0.317. The rigorous internal and external validations were conducted for the GFA and LSSVM models, further verifying these models with the excellent robustness and predictive ability. Therefore, both models can be used for the prediction of the ecotoxicity of newly synthesized ILs, and can provide guidance information for designing and synthesizing safer and more eco-friendly ILs.Chapter three, QSAR models were developed that correlate the dataset of 270 ILs with the cytotoxicity to the Leukemia Rat Cell Line (IPC-81).7 descriptors were selected by GFA and used to build the linear model, based on dividing the training set and test set by self-organizing map (SOM) network. For the training set, R2= 0.852, RMSEC-0.402. The prediction R2 and RMSEPwere 0.841 and 0.413, respectively. The LSSVM model was built for predicting more accurately the cytotoxicity of ILs. For the training set, R2= 0.877, RMSEC= 0.390; the prediction R2 and RMSEP were 0.873 and 0.336 for the test set. The rigorous internal and external validations were performed to verify the reliability and predictive ability for GFA and LSSVM models, the result showed that the performance of LSSVM model was a little superior to the GFA model. Discussing the descriptors, the cation structure was the main factor of the toxicity, which mainly depends on the hydrophobicity and space structure of cations.Chapter four, quantitative structure-biodegradation relationship (QSBR) model was built to predict the biodegradation of 291 common organic compounds. The optimal subset of descriptors was selected by GFA approach and applied to develop the linear model. The correlation coefficient of training set, leave one out cross validation, test set were 0.755,0.742,0.697, respectively. The result of the discussion of descriptors indicated that the biodegradation of organic compounds mainly relate to the connectivity, branching and LUMO energy of molecular structures. Through developing the models for the compounds with different function groups, respectively, the more accurate, stable and reliable models were obtained, the analysis result showed the biodegradation of each class compounds were closely related to their molecular structures.