Correlation and Predictive Modeling of Extracellular Cell Culture Metabolites to Drug Product Titer and Cellular Viability Using Untargeted High Resolution Mass Spectrometry

Abstract

The goal of the thesis project was to investigate the outperformance of one media to another using untargeted mass spectrometry metabolite analysis. Media 1 had previously consistently outperformed Media 2 throughout growth analysis. The correlation between extracellular cell culture metabolites and both cell viability and protein titer was assessed. Also, predictive modeling using both partial least squares discriminant analysis and random forests was used to determine the significant distinguishing characteristics contributing to the variance in metabolite quantitation between both Media 1 and Media 2. Many metabolites were found to be correlated to both cell viability and protein titer, however those correlated as well as being the distinguishing characteristics via modeling aided in the determination as to the nature of Media 1 outperforming Media 2. It was found that both folic acid and pantothenic acid were both significant in the predictive modeling and had a strong positive correlation to Media 1 viability.