Optimizing monoclonal antibody biosimilar production via transfer and active learning for targeted quality profiles

Abstract

Biosimilar development of monoclonal antibodies (mAbs) is gaining significant momentum as numerous blockbuster biologics approach their patent expiry in the current decade. A critical challenge in biosimilar development lies in achieving product quality attributes(PQAs) comparable to the innovator product. PQAs in upstream processing are influenced by multiple factors, including cell line selection, media composition, feeding strategy, supplements, and bioreactor process parameters, with physical parameter optimization playing a pivotal role in enhancing both product titer and modulating PQAs. In this study, we systematically evaluated the impact of physical process parameters—pH and temperature along with initial seeding density (ISD)—on N-glycan profiles and charge variants across four biosimilar development projects (Projects 1–4). Stepwise regression models were developed between process parameters and product quality attributes using JMP software to establish parameter-attribute relationships. Our results demonstrated that lowering culture pH reduced %acidic variants and %galactosylation while increasing %basic variants and %afucosylation (AF). Increased culture temperature resulted in an increase in %acidic variants and a decrease in %AF. This parameter-attribute relationships knowledge base was directly applied in experimental design to expedite the development of a fifth mAb biosimilar development (Project 5), substantially reducing experimental iterations and development timelines, exemplifying the practical implementation of Bioprocessing 4.0 principles.