The role of model-based optimisation in designing formulations

Claire Adjiman
Department of Chemical Engineering, Imperial College London

The need for extensive experimental campaigns to design new formulations can lead to high development costs and, in cases where new materials are involved (e.g., a new active pharmaceutical ingredient) and there is therefore limited supply. Planning an experimental campaign, using a method such as factorial design for instance, is especially difficult when some of the factors include discrete options, such as a choice of polymers, surfactants or solvents.

In this talk, we explore ways in which model-based optimisation can facilitate the design of formulations. We focus in particular on approaches that enable different ingredients to be selected for the final formulation and that lead to a list of promising designs to be tested experimentally.

We focus on the design of experiments to improve performance. We discuss the use of computer-aided molecular design to guide the search for better performing molecules and formulations. This requires the integration of structure-property models and property-performance models within an optimisation framework and leads to a prioritised list of candidate designs that can be tested experimentally. When the necessary models are not available, we show how physical or virtual experiments can be used to build surrogate models. These concepts are illustrated through examples from the design of reaction solvents and drug delivery systems.