De Novo Approaches for Bioprocess Parameter Estimation

This white paper discusses the challenges and methodologies associated with Raman spectroscopy in bioprocess monitoring. It outlines the empirical calibration methods commonly used, such as partial least-squares (PLS) regression, and identifies significant challenges including nonstationarity, extreme covariates, and the limitations of cross-validation in bioprocess data. The paper emphasizes the need for robust Raman method development and presents the MAVERICK System's de novo modeling approach, which integrates fundamental biological mechanisms with empirical data. This hybrid model aims to enhance predictive accuracy and reduce overfitting risks. The paper also highlights the historical context of Raman spectroscopy's application in pharmaceuticals and bioprocesses, detailing its evolution and the growing interest in its capabilities for monitoring cellular metabolites and product quality attributes. The findings suggest that while empirical methods have been widely adopted, there is a pressing need for innovative approaches to address the complexities of bioprocess data.