The Study Of The Application Of Quantitative Structure-activity/ Property Relationship In Capillary Electrophoresis

There are altogether six parts in this dissertation. In the first part, the history, development and application of QSAR/QSPR are introduced. In the second part, the application of QSPR in capillary electrophoresis is reviewed. The third part is the presentation of the chemometrics methods which are used in this dissertation.In the fourth part, the application of error back propagation artificial neural networks to the prediction of the mobility of a series of benzoic is proposed. The input variables of the neural networks are selected from 15 quantum chemical parameters and molecular connectivity index of these compounds. Two variable selection strategies, stepwise regression and genetic algorithm, are used. The predictive ability of the generated model was evaluated. The results of the evaluation demonstrate the validity of the model.In the fifth part, the artificial neural networks(ANN) was applied to the optimization of the separation conditions in capillary electrophoresis. The effect of the buffer concentration, surfactant concentration, pH value, applied voltage and organic modifier composition on the separation was examined by using orthogonal design. The prediction model based on ANN was built, and the optimum separation conditions were predicted successfully.In the last part, QSPR is applied to calculate the standard phase transfer energy for protonated amines ions. The results show the validity of both the multiple linear regression (MLR) and the ANN models. The results obtained using ANN were compared with the experimental values as well as with those from MLR. Comparison of the results demonstrates the superiority of the ANN models over the regression models.