Statistični pristopi k doseganju največje doslednosti v postopkih gojenja celične kulture

This thesis investigates the application of statistical methods to optimize consistency in cell culture processes, crucial for biopharmaceutical manufacturing. Emphasizing the production of biosimilar drugs using Chinese hamster ovary cells, the research explores the effectiveness of linear regression, multiple linear regression, and partial least squares (PLS) regression models in predicting viable cell density from capacitance measurements. Key findings include the limitations of simple linear regression due to non-linearity over different growth phases, and the improved accuracy of multiple linear regression when incorporating variables such as temperature and cumulative oxygen flow. PLS regression demonstrated robustness in handling multivariate data, maintaining predictive accuracy throughout the entire process. The study underscores the potential of multivariate models to enhance process consistency, yield, and product quality in biopharmaceutical production.