Bayesian optimization of crystallization processes to guarantee end-use product properties

Abstract

For pharmaceutical solid products, the issue of reproducibly obtaining their desired end-use properties depending on crystal size and form is the main problem to be addressed and solved in process development. Lacking a reliable first-principles model of a crystallization process, a Bayesian optimization algorithm is proposed. On this basis, a short sequence of experimental runs for pinpointing operating conditions that maximize the probability of successfully complying with end-use product properties is defined. Bayesian optimization can take advantage of the full information provided by the sequence of experiments made using a probabilistic model of the probability of success based on a one-class classification method. The proposed algorithm's performance is tested in silico using the crystallization and formulation of an API product where success is about fulfilling a dissolution profile as required by the FDA. Results obtained demonstrate that the sequence of generated experiments allows pinpointing operating conditions for reproducible quality.