The Representation Of Chiral Compound And It’s Appilication To Prediction Of Catalytic Resolution Of Chiral Alcohols

Chirality is one of the fundamental attributes of the nature. As the need for enantiopure compoundsgrows, the ability to prepare compounds enantioselectively is becoming increasingly important.Theenantioselective experiments is high cost, therefore, in this thesis chemoinformatic techniques was used totake advantage of available experimental data to establish models of structure-enantioselectivityrelationships, and assist the experimental design of preparation of single enantiomers. Studies ofstructure-enantioselectivity relationships rely on chiral codes to distinguish between enantiomers, thus, wesuggested chiral substituent code and used it to construct the models to make prediction of the catalyticresolution of chiral alcohols.EAID derived from the extended adjacency matrix is a highly discriminating molecular topologicalindex. In this thesis, we improved the molecular descriptor, and demonstrated its discrimination ability forthe molecules of several large databases and more than5million alkanes whose formula is C23H48. Further,we suggested atomic EAID, and atomic EAID have been successfully applied to a dataset for identificationof chiral center.The chiral substituent code was specially designed for the datasets of the compounds which include aspecific chiral center bonded directly with two common substituents. Based on predefining positionalinformation of four substituents attached to stereocenter, the code was generated from the topological andphysicochemical properties of two variable substituents.In order to assess the prediction ability of chiral substituent code, we applied the code to three datasetsas follows:1） Chiral secondary alcohol products, which composed of50pairs of enantiomers of chiral alcohols, ofenantioselective reduction of ketones catalysted by DIP-chloride. The relationships between absoluteconfiguration of main product and chiral substitute of the product were constructed by random forest,and the constructed models were used to predict enantioselectivity in the asymmetric reaction.Compared with chiral molecular connectivity index and chiral topological charge Index, the resultesobtained by chiral substituent code were better. 2）86enantiomeric pairs of primary alcohols involved in racemic resolutions by transesterifications, orhydrolyses catalyzed by Pseudomonas cepacia lipase. The absolute configuration of major productwas predicted by constructing chemoinformatics models. The subset of the chiral ligand code selectedby genetic algorithm was fed to random forest to construct prediction model, and theenantioselectivity of all15pairs of enantiomers in test set was predicted correctly. The percentage ofcorrect prediction of leave-one-pair out cross-validation of the whole data set reached90%.3）67pairs of enantiomers of secondary alcohols as the enantioselective products of asymmetricreactions that were catalyzed by a biocatalyst, lipase from Candida rugosa. The models ofstructure-enantioselectivity relationship were constructed using random forests. The resulting modelswere assessed both in terms of single enantiomers and pairs of enantiomers, and sactifactory resultswere obtained.