【作者】 王媛；

【导师】 焦龙；

【作者基本信息】 西安石油大学 ， 化学工程（专业学位）， 2020， 硕士

【摘要】 芳香族化合物是自然界普遍存在的有机化合物,也是重要的化工原料。定量构效关系(quantitative structure-activity relationship,QSAR)研究芳香族化合物性质具有重要意义,弥补了试验方法成本高和研究周期长的不足。HQSAR(holographical quantitative structure-activity relationship)相比传统2D-QSAR和3D-QSAR方法,计算更简便、预测能力更高。集成建模可降低样本数量或复杂样本对预测结果的影响,提高模型的预测能力和稳定性。集成建模应用于HQSAR中,既弥补单模型的不足,也为QSAR研究方法提供新思路。因此,本论文围绕HQSAR结合集成建模对多氯联苯、多环芳烃、二噁英等芳香族化合物进行QSAR建模研究。主要内容包括:1.采用HQSAR建立多氯联苯化合物结构与正辛醇/空气分配系数、正辛醇/水分配系数和生物浓缩因子三组活性数据之间的定量关系,该三种活性数据都是衡量多氯联苯在环境行为的重要参数。应用留一交验证和外部测试集验证方法对模型预测能力进行评价,三组HQSAR模型的交叉验证系数Q2分别为0.957、0.954、0.987。研究结果表明,所建立的模型具有较好的预测能力。并通过分子贡献图对多氯联苯分子结构与活性之间的关系进行分析。2.采用HQSAR建立芳香族污染物分子结构与LDPE-水分配系数之间的定量构效关系。应用留一交验证和外部测试集验证方法对模型预测能力进行评价,得到交叉验证系数Q2为0.960和非交叉验证系数R2为0.981。结果表明所建立的模型具有良好的预测能力。并通过分子贡献图对疏水性有机污染物结构与LDPE-水分配系数之间的关系进行分析。3.应用HQSAR方法和集成建模,研究二噁英类化合物结构与正辛醇/水分配系数之间的定量关系。应用留一交验证和外部测试集验证方法对模型预测能力进行评价,得到Q2为0.986和R2为0.991,表明模型具有较好的预测能力和稳健性。并通过分子贡献图分析二噁英化合物分子结构与正辛醇正辛醇/水分配系数内在关系。集成建模选用平均相对误差MRE小于0.8%的成员模型,成员模型总数确定为80,建立集成HQSAR模型,并通过外部测试集验证法对模型进行验证。结果表明,集成HQSAR模型的预测能力和稳健性均提高。4.应用HQSAR方法和集成建模,研究多环芳烃化合物分子结构与色谱保留指数之间的定量关系。应用留一交验证和外部测试集验证方法对模型预测能力进行评价,得到Q2为0.994,R2为0.973。结果表明,建立的HQSAR模型可靠,预测能力强。并通过分子贡献图分析多环芳烃与色谱保留指数之间的内在关系。集成建模选用MRE小于2.0%作为成员模型接纳标准,成员模型总数确定为130,建立集成HQSAR模型,并采用外部测试集验证方法对集成模型预测能力进行评价。结果表明,模型参数表明集成建模稳健、可靠。并与HQSAR进行比较发现,集成HQSAR模型可提高模型预测能力和稳健性。更多还原

【Abstract】 Aromatic compounds are ubiquitous organic compounds in nature and important chemical raw materials.Quantitative structure-activity relationship(QSAR)is of great significance in studying the properties of aromatic compounds,which makes up for the high cost and long research cycle of experimental methods.Compared with traditional 2D-QSAR and 3D-QSAR methods,HQSAR(holographical quantitative structure-activity relationship)has simpler calculation and higher prediction ability.Ensemble modeling can reduce the influence of sample size or complex sample on the prediction results and improve the prediction ability and stability of the model.The application of ensemble modeling in HQSAR method not only makes up for the shortage of single model,but also provides new ideas for QSAR research methods.Therefore,in this study,HQSAR and ensemble modeling were used to study the quantitative relationship between the structure and properties of aromatic compounds.The main contents include:1.HQSAR was used to establish the quantitative structure-activity relationship between the molecular structure of PCBS and the three groups of activity data including n-octanol/air distribution coefficient,n-octanol/water distribution coefficient and biological concentration factor.These three kinds of activity data are all important parameters for measuring the environmental behavior of PCBS.The prediction ability of the model was evaluated by the method of retention cross validation and external test set validation.The cross validation coefficient of the three HQSAR models was 0.957,0.954 and 0.987,respectively.The results show that the model has good predictive power.The relationship between molecular structure and activity of PCBS was analyzed by molecular contribution diagram.2.HQSAR was used to establish the quantitative structure-activity relationship between the molecular structure of hydrophobic organic pollutants and LDPE-water distribution coefficient.The prediction ability of the model was evaluated by using the method of retention cross validation and external test set validation,and the results showed that the coefficient of cross-validation was 0.960 and the coefficient of non-cross-validation was 0.981.The results show that the model has good prediction ability.The relationship between the structure of hydrophobic organic pollutants and LDPE-water distribution coefficient was analyzed by molecular contribution diagram.3.Using HQSAR method and ensemble modeling,the quantitative relationship between the molecular structure of dioxins and the n-octanol/water distribution coefficient was studied.The prediction ability of the model was evaluated by using the method of retention cross validation and external test set validation,and the results showed that the coefficient of cross-validation was 0.986 and the coefficient of non-cross-validation was 0.991,indicating that the model had good prediction ability and robustness.The relationship between the molecular structure of dioxins and the distribution coefficient of n-octanol/water was analyzed by the molecular contribution diagram.For ensemble modeling,the member models with an average relative error of less than 0.8%and a total number of member models of 80 were selected.The ensemble HQSAR model was established,and the model was verified by the external test set validation method.The results showed that the predictive ability and robustness of the ensemble HQSAR model were improved.4.The quantitative relationship between the molecular structure of pahs and the chromatographic retention index was studied by using HQSAR method and integrated modeling.The prediction ability of the model was evaluated by using the method of retention cross validation and external test set validation,and the results showed that he coefficient of cross-validation was 0.994 and the coefficient of non-cross-validation was 0.973.The results show that the established HQSAR model is reliable and has strong prediction ability.The relationship between pahs and chromatographic retention index was analyzed by molecular contribution diagram.average relative error less than 2.0%was selected as the acceptance standard for integrated modeling,and the total number of member models was determined to be 130.Integrated HQSAR model was established,and external test set validation method was used to evaluate the predictive ability of integrated model.The results show that the model parameters are robust and reliable.Compared with HQSAR,it is found that the integrated HQSAR model can improve the prediction ability and robustness of the model.更多还原