(223c) Intensification of CHO Cell Bioprocesses Using Hybrid Unstructured Kinetic Models

In 2023, the market size of biologics was estimated at a value of USD 349.6 billion and is projected to grow to USD 699.5 billion by 2032 (1). The utilization of Chinese Hamster Ovary (CHO) cell culture for the production of biotherapeutics is a cornerstone of the biopharmaceutical industry (2). This is essential for the production of various biologics, including monoclonal antibodies, multispecifics, including bispecifics and trispecifics (3), and antibody-drug conjugates (ADCs)(4). The primary advantage of using CHO cells lines in their ability to produce glycosylation profiles on proteins that are analogous to the profiles found on native proteins produced by humans (5), are less immunogenic and have better safety and pharmacokinetic profiles (6). Yet, despite their widespread use, developing and scaling up these processes remains technically challenging due to the sensitivity of CHO cells to a wide range of parameters, including nutrient profiles, feed strategies, and environmental variables such as pH, temperature, shear, and mixing gradients (7).

Over the past few decades, the productivity of CHO cell culture processes has improved significantly. However, the forecasted demand for therapeutics produced by these processes necessitates even higher productivity in the future. To meet this demand, multiple attempts have been made to intensify these processes using various techniques. One such technique is N-1 perfusion, which involves utilizing a perfusion process in the seed train while maintaining an existing fed-batch-based production train. This approach aims to enhance cell density and productivity before the production phase, thereby increasing overall yields.

Identifying a high-titer process for an intensified process that can increase titer while maintaining product quality profiles, often requires a significant number of experiments, which can be time-consuming and resource-intensive. This is attributed to the extensive number of process parameters that can impact the output of the cell culture across the seed and production stage. Furthermore, the impact of a multitude of these parameters on the final product quality attributes is also not clearly mapped, leading to multiple exploratory experiments that are necessary to understand the parameter space further. To alleviate this experimental burden, in-silico modeling has been suggested as a valuable tool. In-silico models can simulate various process conditions and predict outcomes, reducing the need for extensive wet lab experiments. However, despite the availability of numerous kinetic models for CHO cells in the literature, including both structured (8) and unstructured kinetic models (9), fitting these models to specific processes remains challenging. This difficulty is primarily due to the lack of comprehensive knowledge about major metabolic pathways that affect nutrient uptake and byproduct formation, especially in cases where the cell line has been significantly modified, as is often the case.

In such situations, data-driven techniques for process simulation can provide valuable insights that are otherwise difficult to obtain. Hybrid models, which combine mechanistic and data-driven approaches, offer a powerful solution for optimizing CHO cell culture processes. These models leverage the strengths of both approaches, using mechanistic models to describe the fundamental biology of the cells and data-driven models, such as neural networks, to capture complex interactions and predict outcomes.

In this talk, we will present multiple case studies where mechanistic and hybrid models have been used for the optimization and development of mammalian cell culture processes across different molecule modalities and for different unit operations. We will discuss the advantages and pitfalls of mechanistic models, followed by the hybridization of these models using neural networks. The case studies will highlight how in-silico models can describe CHO cell culture growth dynamics and optimize cell culture attributes to enhance productivity while maintaining optimum product quality attributes.

Validation approaches for these models, including both retrospective and prospective methods, will also be discussed. Retrospective validation involves comparing model predictions with historical data, while prospective validation entails testing model predictions in new experiments. Additionally, we will address the regulatory considerations when designing and implementing hybrid models in biopharmaceutical manufacturing.

By leveraging hybrid models, the biopharmaceutical industry can achieve higher titers and better product quality, ultimately improving patient access to essential medicines. These models enhance cell culture productivity and maintain product quality by suggesting optimal growth conditions throughout the biomanufacturing process. The integration of mechanistic and data-driven approaches represents a significant advancement in the field of bioprocessing, offering a pathway to more efficient and cost-effective production of biotherapeutics.

References

1. Biologics Market Size to Reach USD 699.5 Billion by 2032, Impelled by Emergence of Advanced Drug Delivery Systems [press release]. BioSpace2024.
2. Kim JY, Kim Y-G, Lee GM. CHO cells in biotechnology for production of recombinant proteins: current state and further potential. Applied Microbiology and Biotechnology. 2012;93(3):917-30.
3. Sebastião MJ, Hoffman M, Escandell JA-O, Tousi F, Zhang J, Figueroa B, et al. Identification of Mispairing Omic Signatures in Chinese Hamster Ovary (CHO) Cells Producing a Tri-Specific Antibody. LID - 10.3390/biomedicines11112890 [doi] LID - 2890. (2227-9059 (Print)).
4. Tian F, Lu Y, Manibusan A, Sellers A, Tran H, Sun Y, et al. A general approach to site-specific antibody drug conjugates. Proceedings of the National Academy of Sciences. 2014;111(5):1766-71.
5. Haryadi R, Chan KF, Lin PC, Tan YL, Wan C, Shahreel W, et al. Generating and characterizing a comprehensive panel of CHO cells glycosylation mutants for advancing glycobiology and biotechnology research. Scientific Reports. 2024;14(1):23068.
6. Donini R, Haslam SM, Kontoravdi C. Glycoengineering Chinese hamster ovary cells: a short history. Biochemical Society Transactions. 2021;49(2):915-31.
7. Neuss A, Steimann T, Tomas Borges JS, Dinger R, Magnus JB. Scale-up of CHO cell cultures: from 96-well-microtiter plates to stirred tank reactors across three orders of magnitude. Journal of Biological Engineering. 2025;19(1):5.
8. Jiménez del Val I, Kyriakopoulos S, Albrecht S, Stockmann H, Rudd PM, Polizzi KM, et al. CHOmpact: A reduced metabolic model of Chinese hamster ovary cells with enhanced interpretability. Biotechnology and Bioengineering. 2023;120(9):2479-93.
9. Xing Z, Bishop N, Leister K, Li ZJ. Modeling kinetics of a large-scale fed-batch CHO cell culture by Markov chain Monte Carlo method. Biotechnology Progress. 2010;26(1):208-19.