Leveraging Bayesian Approach to Predict Drug Manufacturing Performance

With increase in competition and regulatory requirements, the pharmaceutical and biotechnology industry has focused more resources on ensuring consistency and quality during manufacturing to minimize process deviations and batch failures. Specifically, computational tools are developed to monitor and predict manufacturing performance. The objective of this paper is to identify a process prediction tool, which can leverage available data to predict the performance of future production batches. Hence, if we suspect a process deviation in future batches, strategies can be developed in advance for risk mitigation. Critical process and quality attribute values for an upcoming manufacturing batch were predicted using both a conventional and Bayesian approach, by leveraging historical manufacturing data. Both approaches arrived at a similar prediction when a process performed consistently. However, when there was heterogeneity between the historical and current trends, the Bayesian approach was better at capturing the heterogeneity and thus enabling a more accurate prediction of future batch performance.