Chiral Field Of Quantitative Structure-activity Relationship Study

Chirality is a common and important phenomenon in nature. Biology activities, material properties and many other important characters are often different between a pair of enantiomers. Scientists have focused more and more on stereochemistry and asymmetric synthesis. However, due to the complexity of system and the illegibility of mechanism, calculations on chiral compounds are not widely performed. In this thesis, QSAR/QSPR calculations on several chiral subjects were introduced.As the most efficient method to produce chiral compounds, catalytic asymmetric reaction is one of the most active fields in stereochemistry researches. Limited by the illegibility of mechanisms, it is always an experimental job to design or screen novel asymmetric catalysts, and to optimize reaction environment. In this thesis, some QSAR models between e.e.% and the structure of catalysts are obtained through multiple linear regression and artificial neural networks. Those models are useful to screen new asymmetric catalysts, to instruct the design of reaction environment, and to testify or improve the existing mechanism. The QSAR studies are carried out in different types of reactions, and are expected to be extended to more of them.Optical rotation is one of the most important properties in chiral compounds. The configuration of enantiomer and the approximate value of e.e.% can be easily evaluated by comparing it with the standard optical rotation value. Preparatory QSPR researches on molar optical rotation values are presented in the following chapters.While doing researches on asymmetric field, we were also performing QSAR calculations between cytotoxic activities and the structure of camptothecin analogues to assist the experiments in our lab. Three types of camptothecin analogues are studied and acceptable QSAR models are established with topological molecular descriptors respectively.A brief introduction to ObjectANN, a generalized neural network analysis software built by ourselves, is mentioned in the last part of this thesis. This software was programmed under object-oriented ideology and suitable design patterns. It can be easily expanded, and is flexible enough to support multiple neural network types and different data formats. With legible user interface and fast calculation speed, it helped us a lot in a variety of QSAR/QSPR studies.