Hybrid Modeling of an Active Pharmaceutical Ingredient Flow Synthesis in a Ring-Opening Reaction of an Epoxide with a Grignard Reagent

Abstract

This work presents hybrid modeling of an active pharmaceutical ingredient flow synthesis in a ring-opening reaction of an epoxide with a Grignard reagent. Conducting this reaction in a batch reactor is challenging due to its fast and highly exothermic nature, which makes it difficult to collect kinetic data. To address this, a flow experiment was conducted with varying parameters, such as the inner diameter and temperature, to collect kinetic information. A onedimensional flow mechanistic model was then developed, and two reaction mechanisms were assessed, with the Swain and Boyles mechanism found to show better results than the Meisenheimer and Casper mechanism. In this model, the impact of changes in reactor inner diameters was considered by adjusting the preexponential factor as a function of the varying Reynolds number, which simplifies the effort needed to consider the inner diameter effects relative to the development of two-dimensional models. The reaction presents a further difficulty due to the generation of impurities, and the underlying mechanism is still unclear. To tackle this challenge, a data-driven approach, a random forest regression model, was adopted to quantify them. The adoption of the data-driven part enabled the quantification of the impurity with high accuracy, demonstrating the potential strength of this approach when not all reaction mechanisms are fully understood. Overall, the hybrid model approach allowed for a deeper understanding of the reaction and improved the ability to predict its behavior.

Kinetic Study and Model-Based Design Space Determination for a Drug Substance Flow Synthesis Using an Amination Reaction via Nucleophilic Aromatic Substitution

“…In the previous work, the pre-exponential factor was adjusted using the Reynolds number, and the model successfully captured the variations in concentration profiles at different inner diameter conditions. 18 Theoretically, the pre-exponential factor is thought to mirror molecular collisions. However, on a large scale, such as in a reactor, the Reynolds number can provide a good approximation of the effect of mixing on the reaction rate.…”

“…However, on a large scale, such as in a reactor, the Reynolds number can provide a good approximation of the effect of mixing on the reaction rate. The same approach was applied in this study, with the adjusted preexponential factor being estimated using a linear equation that incorporates the Reynolds number as a variable, as shown in eq 12: 18 A a Re b…”

“…To address this issue, we proposed an alternative approach which incorporates mixing effects into more streamlined one-dimensional models. 18 The modeling and simulation approach is not limited to process understanding and mathematical representation but can be extended to design space. Quality by Design (QbD) is a systematic approach used in the pharmaceutical industry to ensure that the products meet the required quality standards (ICH Q8).…”

“…19 The approach involves identifying and controlling critical process parameters (CPPs) and critical material attributes (CMAs) that are important in achieving the desired critical quality attributes (CQAs). 18 Within the QbD framework, the design space is defined as a multidimensional space of input variables and process parameters. It is the region in which a manufacturing process can produce a product with the desired quality attributes consistently.…”

“…Experimental studies have shown that the mixing can have a significant impact in flow chemistry. − Mixing processes typically require the use of complex 2D models that are time-consuming and computationally expensive. To address this issue, we proposed an alternative approach which incorporates mixing effects into more streamlined one-dimensional models …”