A multi-dimensional modeling framework integrating metabolomics analysis and process modeling for bioprocess...

Upstream bioprocess development relies heavily on design of experiment (DoE) and response surface methodology, which often focus on endpoint outcomes while overlooking the dynamic trajectories of cell culture processes. To overcome this limitation, this study established a multi-dimensional modeling framework integrating deterministic screening design (DSD), time-resolved metabolomics, and process modeling to elucidate how critical process parameters (CPPs) regulate antibody titer and quality attributes during the cell culture process. First, a DSD-based response surface model identified key regulators of antibody titer and quality. Subsequently, metabolomics analysis revealed the underlying metabolic mechanisms: higher temperature and inoculation density promoted amino acid and energy metabolism, whereas pH modulated central carbon flux. Building on these insights, Gaussian process regression (GPR) dynamic models were developed that integrate time-series metabolite profiles and CPPs to predict viable cell density and antibody titer with high accuracy (test-set R² > 0.90). This model significantly outperformed conventional linear (PLS) and complex machine-learning approaches (boosted trees, neural networks). This work provides a coherent methodological pipeline for mechanistic understanding and real-time prediction of bioprocess, offering a valuable tool for robust bioprocess optimization and scale-up.