A simple AI-driven process intensification protocol for active pharmaceutical ingredients synthesis

Abstract

The synthesis of active pharmaceutical ingredients (APIs) traditionally relies on batch reactors, which often exhibit challenges in terms of both selectivity and heat transfer control. This study investigated the Aza-Michael addition between methylamine and 2-vinylpyridine to synthetize betahistine, an analogue of histamine, converting a traditional batch process into a continuous flow reaction. The aim of the study was to define an intensification protocol capable of identifying optimized operating conditions to maximise betahistine production. A dedicated experimental setup was developed using a custom-built PTFE-based tubular microreactor which allowed for an optimal control of pressure, temperature, residence time, and reactants molar ratio. Analytical characterization was performed using both UHPLC and H-NMR. Process intensification was achieved using two different approaches: a traditional one, based on deterministic mathematical models to simulate the chemical reactions involved, and a modern approach based on Feedforward Neural Networks. The highest selectivity experimentally observed was approximately 82% at a 2:1 methylamine to 2-vinylpyridine ratio and 150°C, with a residence time of 4 minutes. Both optimizing approaches lead to the same results, confirming the advantages of using suitable intensification protocols for shifting to continuous flow batch processes, especially in pharmaceutical synthesis.