Spectroscopy-Based Cell Culture Predictive Monitoring.

Abstract

Spectral monitoring combined with chemometrics models resulting from machine learning approaches allows cell culture to be monitored almost in real time. This process analytical technology offers to drastically reduce the amount of hands-on time and laboratory testing needed to monitor this crucial biomanufacturing step. In this article, we propose a method to anticipate future spectra. The method is based on extrapolation of the spectra in a reduced-dimensionality space, followed by retroprojection in the original space. Passed to regular chemometrics models already fitted, these anticipated spectra enable predictive cell culture monitoring up to several dozen hours with satisfactory quality. This anticipation paves the way for course-correction and enhanced operations such as reduced need for night shifts.