Data-Driven Modeling for the Enhanced Understanding for the Crystallization of an Active Pharmaceutical Ingredient

Extensive research efforts are dedicated to studying and understanding the dynamics of the crystallization of an active pharmaceutical ingredient (API), aiming to optimize product quality, yield, and robustness. In this study, we developed, tuned, and demonstrated a data-rich experimentation workflow that can be used to characterize and model the dynamic behavior of an antisolvent crystallization of an API. First, automated, parallel experiment technology and empirical models are used to describe the API solubility as a function of temperature and solvent composition. Next, an efficient protocol that leverages laboratory automation and process analytical technologies was used to generate an accurate chemometric model to quantify API supernatant concentration with in situ Fourier-transformed infrared spectroscopy. Using a 2² full-factorial design and data-rich experimentation, supernatant concentration profiles were obtained to characterize the impact of the crystallization temperature and antisolvent charge rate on the isolation procedure. These multivariate, time-series data profiles were subsequently modeled using dynamic response surface methodology (DRSM). This approach provided a comprehensive understanding of the sensitivity of the operating parameters on the crystallization process to enable rapid process development. The DRSM model successfully predicted the concentration dynamics based on antisolvent fraction, addition rate, and temperature for the training data set from the initial design of the experiment as well as a separate validation experiment. This study highlights how the combination of data-rich experimentation and process modeling can lead to valuable process knowledge for optimization and control strategies for crystallization in pharmaceutical manufacturing.