Computer-aided solvent design for suppressing HCN generation in amino acid activation

A highly toxic compound, hydrogen cyanide (HCN), was discovered to result from the reaction between Ethyl cyano (hydroxyimino) acetate (Oxyma) and diisopropylcarbodiimide (DIC), a popular reagent combination for amino acid activation. The reaction solvent has been found to influence the amount of HCN produced so that judicious solvent choice offers a route to suppressing HCN formation. Given the safety implications and the time-demanding nature of experimental solvent selection, we employ a methodology of quantum mechanical computer-aided molecular design (QM-CAMD) to design a new reaction solvent in order to minimize the amount of HCN formed. In this work, we improve on the original QM-CAMD approach with an enhanced surrogate model to predict the reaction rate constant from several solvent properties. A set of solvents is selected for model regression using model-based design of experiments (MBDoE), where the determinant of the information matrix of the design, known as D-criterion, is maximized. The use of a model-based approach is especially beneficial here as it links the large discrete space of solvent molecules to the reduced space of solvent properties. The resulting surrogate model exhibits an improved adjusted coefficient of determination and leads to more accurate predicted rate constants than the model generated without using MBDoE. The proposed DoE-QM-CAMD algorithm reaches convergence in one iteration. In the future, the main reaction of amino acid activation will be considered to design a solvent that maintains the rate of the main reaction while minimizing HCN generation.