In-situ Raman spectroscopy for real-time monitoring of cell density and metabolites during adenovirus production in HEK293 suspension culture

Adenoviral (AdV) vectors are widely used in human and veterinary vaccines as well as in gene and cancer therapies; however, their manufacturing remains challenging due to the complexity of real‑time process monitoring and control arising from the dual phase nature of production, lytic production cycle, and the cell density effect. In this study, in-situ Raman spectroscopy was evaluated as a process analytical technology (PAT) for real-time monitoring of cell density and metabolites during AdV production in suspension HEK293 cells. Raman spectra were acquired every 30 min in six 3 L bioreactor runs, and partial least squares (PLS) models were developed to monitor viable and total cell densities as well as key metabolites, including glucose, lactate, glutamine, ammonium, and glutamate, across both the growth and infection phases using a global model. The performance was estimated by nested leave-one-batch-out validation, in which each batch was predicted by a model whose spectral preprocessing, spectral range and number of latent variables were selected using the other five batches only. Pooled root mean square errors of prediction over the six held-out batches were 0.63 and 0.44 × 10 6 cells/mL for viable and total cell density (R P 2 = 0.80 and 0.94) and 1.18, 2.11, 0.65 and 0.39 mM for glucose, lactate, glutamine and ammonium (R P 2 = 0.99, 0.98, 0.86 and 0.83), and feeding and dilution events were reproduced in the continuous predictions. Glutamate was not predicted reliably (R P 2 = −0.37), and no usable viral-titre model could be established. Compared with a common full-range model, target-specific selection of the normalisation reference and spectral range lowered the pooled prediction error for the cell densities, glucose and lactate, but not for glutamine, ammonium or glutamate, and the Raman shifts selected consistently across folds helped to interpret the models. Overall, the work demonstrates the feasibility of in-situ Raman spectroscopy for monitoring cell density and major metabolites in both phases of adenovirus production and supports its integration into advanced monitoring and process optimization strategies.