(417d) An Integrated High-Throughput Experimentation and Semi-Empirical Modeling Approach to Building Process Knowledge in Pharmaceutical Process Development

We demonstrate the power of collaboratively integrating predictive modeling and high-throughput experimentation to deliver data needed for research and scale up in a pharmaceutical process research and development setting. Since semi-empirical models need empirical data as input and high-throughput experiments can be very effectively designed when they are informed by a model prediction, these approaches lead naturally to the following integrated workflow. A high-throughput experiment is done as the first step, a model is developed based on this data as a second step, the space of process options is explored using the model as a third step, and the most promising model results are examined in targeted high-throughput experiments as a final step. This workflow can be used to speed up development timelines and improve process understanding by reducing the number of experiments performed while at the same time providing a wealth of data about the process. A combination of high-throughput experiments and semi-empirical models provides the best option for this task as the possibilities for process options can be too many to fully study with even high-throughput experimentation, ab initio models can be too computationally intensive and not sufficiently accurate, and although the processing options can be fully studied rapidly with semi-empirical models, they require experimental measurements as input and should be experimentally verified. As an example, we will focus on high-throughput solubility measurements and solubility modeling with the NRTL-SAC (Non Random Two Liquid Segment Activity Coefficient) solubility model. We have extensively and routinely applied this workflow in building knowledge around solubility of new Active Pharmaceutical Ingredients (API's) and intermediates in the synthesis of API's. The impact of this collaboration in terms of time savings will be presented as well as an example of a typical result.