Offline Raman monitoring of SARS-CoV-2 VLP upstream stage using two culture media

This work aimed to develop global chemometric models for offline Raman spectroscopy monitoring SARS-CoV-2 virus-like particles (VLP) production in two culture media. The biochemical parameters monitored were viable cell density (Xv), cell viability, viral titer, glucose, glutamine, glutamate, lactate, and ammonium. Partial Least Squares (PLS) and Artificial Neural Networks (ANN) were the regression techniques utilized and compared. The effects of different spectral preprocessing approaches were also assessed. ANN demonstrated greater predictive capacity for most analyzed parameters, as indicated by the lowest mean absolute errors (MAE) according to a one-tailed Student’s t-test (α = 0.05). However, the fitting performances of the best models created by PLS and ANN for Xv, glutamine, and ammonium were statistically similar. ANN provided better models for the other parameters. Therefore, it was demonstrated that the ANN modeling is the most suitable approach to create monitoring chemometric models for VLP production from Raman spectral data, which is flexible and enables their adaptation to different culture media and recombinant baculovirus infections using the <i>Sf9</i> insect cell line as host.