Continuous flow process optimization aided by machine learning for a pharmaceutical intermediate

In this paper, we demonstrate the use of machine learning to optimize the continuous flow process of a crucial intermediate in the production of Nemonoxacin. Our focus is to achieve the good yield and enantioselectivity in the construction of chiral methyl group utilize the initial 29 experimental datasets and consider six important variables. Employing Single-Objective Bayesian optimization (SOBO), we achieved an impressive predicted yield of up to 89.7%, which is consistent with the experimental results, with a yield of 89.5%. Additionally, A Multi-Objective Bayesian Optimization (MOBO) algorithm, namely qNEHVI, to strike a balance between yield and enantioselectivity in the continuous flow system is applied. The algorithm’s prediction, with a yield of 81.8% and enantioselectivity of 97.85%, was experimentally validated, yielding 83.8% and 97.2%, respectively. This study effectively demonstrates that Bayesian optimization is a powerful tool for optimizing the continuous process in the production of active pharmaceutical ingredients (APIs).