Efficient optimization of process strategies with model-assisted design of experiments

Abstract

Conventional design of experiments (DoE) methods require expert knowledge about the investigated factors and their boundary values and mostly lead to multiple rounds of time-consuming and costly experiments. The combination of DoE with mathematical process modeling in model-assisted DoE (mDoE) can be used to increase the mechanistic understanding of the process. Furthermore, it is aimed to optimize the processes with respect to a target (e.g., amount of cells, product titer), which also provides new insights into the process. In this chapter, the workflow of mDoE is explained stepwise including corresponding protocols. Firstly, a mathematical process model is adapted to cultivation data of first experimental data or existing knowledge. Secondly, model-assisted simulations are treated in the same way as experimentally derived data and included as responses in statistical DoEs. The DoEs are then evaluated based on the simulated data, and a constrained-based optimization of the experimental space can be conducted. This loop can be repeated several times and significantly reduces the number of experiments in process development.