Development of Statistical Models to Identify Structural Features and Noncovalent Interactions Influencing Asymmetric Catalysis

项目总结

Project Summary This proposal describes the development of new statistical methods to evaluate the weak, noncovalent interactions involved in asymmetric catalysis. Noncovalent interactions are essential to biological recognition as well as to the function of biological catalysts—enzymes. The project will be conducted in the context of two enantioconvergent substitution reactions of a-chloroglycine esters. The proposed reactions would provide facile access to valuable aryl and allylic unnatural a-amino acids, whereas current synthetic methods can result in low enantioselectivities or yields. Preliminary results show the proposed reactions to be capable of high enantioselectivities when catalyzed by arylpyrrolidino squaramide derivatives, but testing for the optimal catalyst has revealed nonintuitive trends in enantioselectivities and low yields. To improve current capabilities for reaction optimization, new statistical methods will be developed to allow simultaneous optimization of yield and enantioselectivity. A set of good-, modest-, and poor-performing catalysts will be selected, and reactions will be performed using each catalyst. Enantioselectivities as well as reaction rates will be measured to build the data set. Steric and electronic features of the catalysts will be modeled computationally to generate parameters for the statistical analysis. A multivariate linear regression will then be conducted to generate predictive models for both enantioselectivity and yield. After optimizing these reactions, enantioselectivity measurements will be made with each catalyst over a range of temperatures. From this data, ∆∆H‡ and ∆∆S‡ can be calculated for individual catalysts in each reaction. Predictive statistical models will then be created for ∆∆H‡ and ∆∆S‡, drawing from the computational parameter library developed for the optimization models. Based on the parameters that appear in these models, structural features relevant to the enantioselectivities of the catalysts can be identified. The enthalpic and entropic contributions of each relevant structural feature could also be quantitatively assessed. This outcome would contribute to a deeper understanding of how noncovalent interactions operate in the enantiodetermining step of these reactions. As a result of this knowledge, improved catalysts and substrates could be designed. Application of the proposed methodology to any catalytic transformation would result in more efficient reaction optimization and catalyst design for that reaction. Eventually, this work could bring about de novo catalyst design following the creation of a comprehensive library of computational parameters and statistical models encompassing ∆∆G‡, rate, ∆∆H‡, and ∆∆S‡ for representative groups of reactions. The impact of the proposed work would not only contribute to the field of asymmetric catalysis, but it would also provide improved methods of accessing unnatural a-amino acids. These compounds are essential to biological studies of living systems, and this methodology could advance the synthesis of health-improving pharmaceuticals.