Insights Into Lactate Metabolism Through Multivariate Analysis of Cell Culture Bioprocess Data

Wednesday, October 19, 2011: 2:35 PM
Symphony III (Hilton Minneapolis)
Huong Le1, Santosh Kabbur2, Ziran Sun3, Luciano Pollastrini3, Kevin Johnson3, Keri Mills3, George Karypis2 and Wei-Shou Hu1, (1)Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, MN, (2)Computer Science and Engineering, University of Minnesota, Minneapolis, MN, (3)Manufacturing Sciences and Technical Services, Genentech, Inc., Vacaville, CA

Multivariate data analysis has recently emerged as a critical tool to gain fundamental understandings of process characteristics, and thus assisting the search for possible intervention means to improve process performance.  The time dynamics of 150 process parameters acquired from 243 production runs from the Genentech’s Vacaville manufacturing facility were investigated in this study.  Two complementary multivariate approaches, kernel-based support vector regression (SVR) and partial least square (PLS) regression, were used to predict process outcome indicated by the final antibody concentration and the final lactate concentration.  Both productivity and cellular metabolic state can be predicted accurately in the early stage of the inoculum train, suggesting that the history of the culture exerts critical impact on the final process outcome.  Furthermore, specific lactate production rate, viable cell density, and viability were identified as key parameters during the inoculum train contributing to the final process performance.  In the production stage, most critical parameters were also related to lactate metabolism and cell growth, suggesting the important role of these two aspects of process characteristics.  Interestingly, lactate consumption appeared to be a prominent factor in determining final process outcome, and was further investigated.  This study represents an important step towards implementing PAT and QbD principles to enhance process understanding and facilitate dynamic control for improved process robustness.

Extended Abstract: File Not Uploaded