Generative PAT Fingerprint Approach for Verification of the Scale-Up of Pharmaceutical Processes

A real-time scale-up support for pharmaceutical reactions, a process analytical techniques (PAT) fingerprint approach, is presented. The approach involves the construction of a fingerprint model using PAT data, here Fourier transform infrared trends, derived from reference experiments in the lab. This generative model describes the reaction profile and the typical variation within the experiment set. During scale-up, the reaction profile of the running batch is compared to the fingerprint model to determine whether it aligns with the expected behavior or deviates from the reference experiments. The analysis can be conducted in real time during batch execution, providing a quantitative method to validate accurate scale-up. The effectiveness of the approach is demonstrated through various examples, including slurry-to-homogeneous reactions, heterogeneous slurry-to-slurry reactions, and an autocatalytic system. In addition to verifying scale-up correctness, the approach also aids in root cause analysis when deviations are observed. By utilization of this fingerprint model, valuable insights regarding process sensitivities and discrepancies between lab and plant settings can be gained.