Multivariate data analysis as a PAT tool for early bioprocess development data

Early development datasets are typically unstructured, incomplete and truncated, yet they are readily available and contain relevant process information which is not extracted using classical data analysis techniques. In this paper, we illustrate the power of multivariate data analysis (MVDA) as a Process Analytical Technology tool to analyze early development data of a PER.C6 (R) cell cultivation process. MVDA increased our understanding of the process studied. Principal component analysis enabled a thorough exploration of the dataset, identifying causes for batch deviations and revealing sensitivity of the process to scale. These findings were previously undetected using traditional univariate analysis. The lack of structure and gaps in the early development datasets made it impossible to fit them to more advanced partial least square regression models. This paper clearly shows that MVDA should be routinely used to analyze early development data to reveal relevant information for later development and scale-up. The value of these early development runs can be greatly enhanced if the experiments are well-structured and accompanied with full process analytics. This up-front investment will result in shorter and more efficient process development paths, resulting in lower overall development costs for new biopharmaceutical products. (C) 2013 Elsevier B.V. All rights reserved.