Data science workflows for biopharmaceutical manufacturing process validation stage 1

Abstract

The biopharmaceutical market is innovative, well growing and delivering about 20% of all pharmaceutical product to patients. In order to consistently deliver high product quality the biopharmaceutical manufacturing process needs to be understood, controlled and effectively monitored. Those tasks are commonly addressed in manufacturing process validation, which is also requested from regulatory agencies due its importance in respect to patient risk. Especially the first step of achieving process knowledge by understanding and controlling potential sources of variance and risks is key to ensure successful routine manufacturing. Those activities are usually covered in process characterization studies (PCS) in industry. Within this thesis, an advanced data science workflow for PCS is presented that points towards a holistic risk awareness and control strategy via knowledge obtained from single unit operations. Major novelties described in this thesis ensure on the one hand that information from single unit operations such as fermentation processes are accurately extracted. Moreover, novel statistical power analysis methods are presented to ensure that no critical information or process parameter on product quality has been overlooked. On the other hand an integrated process model has been introduced that facilitates to combine this knowledge from single unit operation by means of Monte Carlo simulation. The integrated process model was successfully applied on a real industrial process to derive holistic risk awareness and a holistic control strategy. By applying this advanced workflow it is anticipated that variance in process output and product quality can be reduced and commensurately producers and patient risk is lowered.