Supervised Machine-Learning Algorithm using Low Data Sets: Flow Chemistry Optimization of the Key Urea Moiety Construction in Larotrectinib

Abstract

Unsymmetrical urea is a ubiquitous moiety in pharmaceuticals, providing a linkage between pharmacophores. The assembly of an unsymmetrical urea bridge in various therapeutic agents can be accomplished through several approaches. Conventional methods involving hazardous compounds such as phosgene and isolated isocyanates pose safety concerns; a safe surrogate of phosgene, namely CDI, is popularly employed for the construction of both symmetrical and unsymmetrical urea. While the use of CDI for the small-scale synthesis of NCEs is a popular strategy, translation of the same chemistry to the large-scale manufacture of unsymmetrical urea containing APIs often encounters certain challenges such as symmetrical urea formation, solubility, and purification issues. Consequently, alternate approaches involving the intermediacy of a stable alkyl/aryl carbamate are typically adopted in manufacturing scenarios. Herein, we describe an effective supervised ML approach involving minimal data sets of flow chemistry parameters to accelerate the process optimization of CDI-based unsymmetrical urea construction for the anticancer drug Larotrectinib. A series of multi-output regression and ensemble models were evaluated to identify the best one that can be employed for rapid and effective reaction optimization. Using this approach, we were able to arrive at the optimal experimental conditions that can be potentially applied for Larotrectinib scale-up with good product purity and yield.