Prediction of Enantiomeric Excess in Asymmetric Reaction

In order to assist enantioselective catalyst screening for asymmetric hydrogen transfer to acetophenone, the model of quantitative structure-activity relationship between the enantiomeric excesses(e. e.) of the products and the structures of the catalysts was constructed. The chiral catalysts can be synthesized by chiral amino alcohol and metal complex precursor, thus, the chiral catalysts were represented by the combination of conformation-independent chirality codes or conformation-dependent chirality codes which were derived from the structures of chiral amino alcohols and indictor variables which were used to distinguish metal complex precursor. The variables selection was performed by genetic algorithm, and then the mathematics model was built by Random Forest to make prediction of e.e. A correlation coefficient of R-2 = 0.769 was obtained for the test set, and the result of out-of-bag(OOB) cross-validation of the whole dataset was R-2 = 0.785.