Evaluation of flow cytometry and chemometric models for monitoring and predicting antigen production at full-scale

Abstract

Reducing the batch to batch variability in the productivity of target antigens is a key goal in vaccine manufacturing processes. Flow cytometry (FC) was investigated to quantify population heterogeneity with respect to oxidative stress levels and its effect on growth and productivity in the manufacture of whooping cough vaccine by Bordetella pertussis. Samples from the entire production train of fermenters were analyzed by means of flow cytometry. Principal Component Analysis (PCA) and Partial Least Squares (PLS) were applied to analyze the highly correlated flow cytometry data. A PCA model showed clear correlations between the distributions of reactive oxidative species (ROS) and scattering from samples collected at different times to the pertactin (PRN) yield measured at the end of the purification process following fermentation. Regression models based on PLS regression were used for predicting the final concentration of PRN from distributions of ROS and scattering at different times during the last fermentation step. The PLS regression model based on the entire flow cytometry distribution was found to be significantly more accurate as compared to a model based on averages of the distribution.