Solvent design assisted by mechanistic insights: methods and application to peptide synthesis

Abstract

Peptides are an emerging novel class of pharmaceutical agents with many potential therapeutic applications. However, during their manufacture, hazardous solvents, e.g., N,N-dimethylformamide (DMF), are frequently used, posing a major concern for process sustainability and safety. This work seeks to develop an efficient computational method of selecting and designing green alternative solvents that can achieve both high reaction efficiency and minimisation of side reactions in peptide synthesis. This method, named DoE-QM-CAMD, embeds three key components into the conventional framework of computer-aided molecular design (CAMD): a) mechanistic knowledge of a chemical reaction, b) accurate quantum mechanical (QM) modelling of reaction kinetics and c) model-based design of experiments (MBDoE). In the DoE-QM-CAMD framework, the kinetic modelling of a multi-step reaction is carried out by focusing on the key reaction species and step(s) that contribute to the underlying kinetic barrier of the whole reaction pathway, and this is guided by the mechanistic insights obtained from a computational (QM) study. Furthermore, the predictive accuracy of the original QM model is ensured through a benchmarking study. Central to DoE-QM-CAMD is the construction of an accurate surrogate model for computationally expensive QM calculations, which is trained using a minimal amount of information-rich solvent data generated from a set of judiciously designed computer experiments (i.e., QM calculations) by MBDoE. The use of the surrogate model in CAMD allows for quick identification of promising alternative solvents to DMF that experimental tests should focus on. Reassuringly, the alternatives found include two solvents already used in peptide synthesis, i.e., dichloromethane and acetonitrile, in addition to several novel solvents that have not been tested. In summary, this work provides a framework of computer-aided solvent design for optimal reaction kinetics, which also shows promise for the design of other molecules in a wider range of potential applications.