Statistical evaluation of data in downstream processing

Biopharmaceutical process development is commonly very empirically driven and needs a large number of experiments delivering corresponding data from diverse analytical tools and monitoring processes. Statistical evaluation of such data to determine precision, accuracy, reproducibility, and robustness of measurements are daily business for all scientists. Within the last two decades new trends emerged changing and challenging classical statistical tools and led to new developments. All fields of omics studies such as genomics, lipidomics, proteomics, metabolomics and more were established. High throughput methods for screening, process development and analytics were set up and are current routine in many companies. Both lead to an exponential increase of included variables and parameters, and consequently, to huge expansion of the amounts of data. Furthermore, the release of the PAT guidance by the FDA in 2004 as the enabling aspect of Quality-by-Design in biopharmaceutical production changed the perspective on data interpretation including the need for modelling strategies, for predictability and evaluation of data in real time or close to. Nowadays, a number of different software is available for these calculations of “classical statistics” and “new approaches” including modelling and simulation are on the market. For an international company such as Boehringer-Ingelheim it is highly important to keep up-to-date with the current developments, ease internal communication through harmonization of wording and tools where possible and steady optimization of the output as well as gain of knowledge from generated data. This can be especially challenging in the biotechnology field due to the different backgrounds of the scientists involved, e.g., engineering, biotechnology, biology, etc., often corresponding different approaches and high disparity in know-how and understanding for statistics.