Quantitative Structure Property Relationship And Quantum Chemistry Application

As a main branch of chemometrics research, quantitative structure-activity relationship (QSAR), or called quantitative structure-property relationship (QSPR), is one of the hotspots and plays an important role nowadays. It has been widely used in chemistry, life sciences, environment sciences, and so on. The present work is a systematic research of quantitative structure-retention relationship (QSRR). In addition, in order to improve the predictive results of models and describe molecular structure better, some valuable information of quantum chemistry is used in QSAR/QSPR. The contents are mainly the following four parts:The first part is about QSRR. Two predictive models are built for the molecules having the same and different structural characters. For polychlorinated biphenyl (PCBs), QSRR model was built by principal component regression (PCR). A regression model with correlation coefficient of 0.9999 and standard deviation of 2.4636 was obtained. The standard deviation of prediction obtained by cross-validation was 3.0656. For the data of liquid fuels, correlation coefficient was 0.9998, standard deviation was 9.987, and standard deviation of prediction obtained by cross-validation was 11.17. The predictive error was controlled about 1%, and which was satisfactory.The second part is still about temperature-programmed retention index. Because its values vary with chromatographic conditions, systematic researches on the conversion were carried out. And three different models of the conversion between different initial temperatures, heating rates and columns were proposed as shown. The performance of the conversion makes temperature-programmed retention index data published be shared by different authors and inter-laboratories possible.The third part is abased on quantum chemistry. The main aim is to compare topological index with quantum descriptors in the process of prediction of PCBs properties. At the same time, generalized correlation coefficient was attained to describe the diversity characters of two kinds of descriptors. From the results, it can be known that two kinds of descriptors have respective advantages and it is a good choice to utilize the information of both the decriptors in the modelling.The fourth part is to develop a new quantum topological descriptor based on the theory of atoms in molecules. Molecular structure, such as atoms and bond, can be described very well by the properties of critical points. At the same time, molecular similarity measures have utilized the properties of critical points. In addition, a good QSAR model can be builted by this new descriptors attained from the properties of critical points.