Studies Of QSPR Of Chromatographic Retention Values Of Chromatographic Retention Values By Chemometrics Methods

The article mainly research the QSPR of PBDEs Chromatography’s retention time with many methods, which include Multiple Linear Regression(MLR) of Chemometric methods, BP Artificial Neural Network(BP) and Particle Swarm Optimization method of BP Neural Network(PSO-BP). Through the theoretical calculation method to realize the relationship between the physical and chemical properties. Polybrominated Diphenyl Ethers(PBDEs) is a kind of widespread global organic pollutants in the environment, and the concentration in the environment are growing fast, which cause serious harm to human body health and it has become a hot spot in today’s environmental science. Therefore this paper will forecast PBDEs Chromatographic retention by means of QSPR model, the specific research content is as follows:First of all, we should use chemical software HyperChem7.5 to produce molecular images of 126 PBDEs. Molecular configuration is optimized by se mi-empirical method to calculate the molecular surface area and volume. And using another chemical software Chemoffice2010 to get 126 PBDEs compound3 D configuration. The structure optimization was carried out by molecular m echanics; The space coordinates(x, y, z) was carried on through the molecula r mechanics to build a new topological index Ys. Combining with molecular bromine number and replace the molecules within the quality as the indepen dent variable when the chromatographic retention values as dependent variable s. And multiple linear regression method was applied to 7 kinds of chromatog raphic column chromatographic retention from100(PBDEs) compounds to buil d QSPR model. The model is then used to predict chromatographic retention for validation of the 26(PBDEs) compounds.Secondly, QSPR model is set up by using BP artificial neural network algorithm for the molecular structure of the selected parameters and the chromatographic retention. We should select 100 PBDEs compounds as Training set and at the sametime select 26 PBDEs compounds as the test set. After test, we can set sample validation of the model to show that the prediction ability of established QSPR model.What’s more, With the application of BP algorithm, we optimize the structure and parameters optimization BP to avoid the BP algorithm, which need repeated tests to determine the number of hidden layer and to improve learning speed and efficiency of the network. Particle Swarm Optimization accurately solved the over fitting of BP artificial neural network algorithm and problems; effectively make up defects of BP network in the following fields, such as network structure, randomness of weights and threshold selection; and make full use of the global search ability of PSO and the local search ability of BP network. Thereby it will increase network intelligent search ability as well as shorten the network convergence time. If we apply PSO and BP algorithm to the QSPR study of chromatographic retention, the established model will be more stable and stronger in prediction ability.When we combine the three kinds of chemometrics algorithms to set up the QSPR models with high stability and strong predict ability, we can provides new ideas and new methods for chemical analysis, life science, drug synthesis, and environmental science and so on. Its prediction to compounds’ nature and activity has become an important method for the ecological risk assessment of organic pollutants. Therefore it has the very good application prospect with important significance theoretically and practically.