CultureBoost: an intelligent framework for retrospective analysis of commercial cell culture processes

Commercial-scale monoclonal antibody manufacturing requires stringent productivity control to maximize profitability. However, final productivity is frequently compromised by subtle process deviations beyond standard operating procedures, such as minor pH shifts, gas sparging variations, and nutrient consumption fluctuations. Although modern Industrial Internet of Things systems generate abundant inline and offline data, actionable process insights remain underexploited. To tackle the critical challenge of productivity fluctuations, identification of root causes is urgent. Therefore, this study proposes an intelligent retrospective data analysis framework for commercial-scale cell culture processes, integrating three core modules: inline data processing, offline data processing, and discriminative information extraction. It enables end-to-end automated historical data mining, quantifies factor contributions to batch quality variations, facilitates rapid potential root-cause identification with minimal data science expertise, and guides timely process optimization. Practical efficacy was demonstrated through two commercial-scale case studies manufacturing biosimilars to adalimumab (HS016) and trastuzumab (HS022), with data from over 30 batches collected per product. Critical factors possibly driving productivity declines were identified from early-batch data and successfully validated through targeted operational optimizations in subsequent runs. By converting latent process data into actionable insights, this scalable and operator-friendly framework provides a practical AI/ML-driven solution for continuous batch quality enhancement in GMP-regulated commercial biomanufacturing.