High Throughput Automation, Data Analysis to Advance Process Development.

This presentation describes statistical experimental design approaches for high throughput automated experimentation, and the resulting data analysis to guide process development of active pharmaceutical ingredients and intermediates. Laboratory automation drastically decreases the cost and effort associated with routine experimentation during process development. The experimental setup supports development and optimization of multiple unit operations, and includes automated charging, automated sampling, sufficient mixing, and temperature control. To fully realize the value of automated experiments, the experiments must be designed and analyzed in a way that balances speed and rigor, so that experimental findings can be rapidly incorporated into future experimentation. Experiments are designed using various statistically relevant strategies including factorial, screening, and optimal designs. The design helps to strategically allocate the experimental budget for a particular round of automated experiments, which are executed in parallel. Once the data is collected, it is rapidly analyzed by combining data visualization, statistical analysis, and modeling to transform the experimental results into usable process knowledge and guide further rounds of automated experimentation. Once the process design space has been thoroughly investigated, in-depth data analysis and modeling is used to guide experiments and pilot plant batches. A case study where this workflow was applied to an API reaction to guide scale up experimentation and plant condition selection is presented. The case study will discuss the practical implementation of the workflow, the challenges of using small scale models to guide larger scale experiments, and how the results were used to select robust, high yielding operating conditions.