Optimization of pharmaceutical reactions using the dynamic response surface methodology

Abstract

Design of Dynamic Experiments (DoDE, Georgakis, 2013) and Dynamic Response Surface Methodology (DRSM, Klebanov and Georgakis, 2016; Wang and Georgakis, 2017) are two recent extensions to the Design of Experiments (DoE) Methodology for data-driven modeling. The DRSM models are used here for the optimization of the operating conditions of organic reactions in pharmaceutical process development. In particular, we investigate the problem of maximizing the operating time window, during which the concentrations of different species satisfy their preset constraints. This ensures the existence of a time window that is wide enough for a laboratory confirmation of acceptable compositional ranges in the final reaction mixture. We also consider the robustness of these optimization results with respect to deviations of the operating conditions and the inherent uncertainties in the DRSM model. Two case studies are examined using either simulated data from a complex reaction network or experimental data from an industrial example. In the latter case, we note the possibility of an optimization pitfall that a fractional factorial design can lead us to, when all the factors are significant and additional experimental data to improve the design's resolution are not available.