Multi-Scale Hybrid Modeling to Predict Cell Culture Process with Metabolic Phase Transitions

Keqi Wang, Sarah W. Harcum, Wei Xie · 2024

Read Multi-Scale Hybrid Modeling to Predict Cell Culture Process with Metabolic Phase Transitions by Keqi Wang, Sarah W. Harcum, Wei Xie (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-03883.txt.

Introduction

Keqi Wang, Sarah Harcum, and Wei Xie build a multi-scale model of cell culture growth, in work from 2024. Cultured cells switch metabolic phases as they grow, which makes batches vary. Factories that grow cells for drugs need forecasts they can trust.

Their framework joins three layers in one system. A random model covers single-cell metabolic networks, and a chance model covers phase switches between cells. A population layer then tracks the whole vessel over time. Online signals such as oxygen use and acidity feed the forecast, plus starting values. Offline lab measures tune the system when available. Long-range forecasts of many culture traits are the payoff.

Catalog record

CREATORKeqi Wang, Sarah W. Harcum, Wei Xie
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE88 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-03883

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Multi-Scale Hybrid Modeling to Predict Cell Culture Process with Metabolic Phase Transitions Keqi Wang, Sarah W. Harcum, Wei Xie Abstract To advance understanding of cellular metabolism and reduce batch-to-batch variability in cell culture processes, this study introduces a multi-scale hybrid modeling framework designed to simulate and predict the dynamic behavior of CHO cell cultures undergoing metabolic phase transitions. The model captures dependencies across molecular, cellular, and macro-kinetic levels, accounting for variability in single-cell metabolic phases. It integrates three components: (i) a stochastic mechanistic model of single-cell metabolic networks, (ii) a probabilistic model of phase transitions, and (iii) a macro-kinetic model of heterogeneous population dynamics. This modular architecture enables flexible representation of process trajectories under diverse conditions and incorporates heterogeneous online (e.g., oxygen uptake, pH) and offline measurements (e.g., viable cell density, metabolite concentrations). Leveraging these data and single-cell insights, the framework predicts culture dynamics using only readily available online measurements and initial conditions, delivering accurate long-term forecasts of multivariate culture behavior and uncertainty-aware estimates of batch-to-batch variation. Overall, this work establishes a robust foundation for digital twin platforms and predictive bioprocess analytics, supporting systematic experimental design and process control to improve yield and production stability in biomanufacturing. K eywords Multi-Scale Kinetic Modeling, Stochastic Molecular Reaction Network, Cell Metabolism, Metabolic Phase Shift, Metabolic Flux Dynamics and Analysis, Cell Classification 1 Introduction Over the past several decades, biopharmaceuticals have gained prominence due to significant impact on public health, particularly highlighted by the production of COVID-19 vaccines and therapeutics. The market value of biopharmaceuticals reached $343 billion in 2021, with 67% (107 out of 159) of approved recombinant products produced using mammalian cell systems. Chinese hamster ovary (CHO) cells are used predominantly, accounting for 89% of mammalian cell-derived products [ 73 ] . CHO cells are preferred over other eukaryotic hosts like yeast, mouse myeloma, HEK 293, and insect cell lines due to its ability to produce recombinant proteins that closely mimic human-like glycoforms, an essential quality for many treatments. This capability, combined with genetic adaptability and resilience under harsh conditions typical in late-stage fed-batch cultures, underscores its critical role in biopharmaceutical manufacturing [ 28 ] . However, like all mammalian cell cultures, CHO cells remain sensitive to variations in culture conditions, which can significantly impact yield, increase the heterogeneity of cell populations, and consequently affect the critical quality attributes (CQAs) of protein products [ 16 ] . This inherent heterogeneity also contributes to batch-to-batch variations, as not all cells within a bioreactor undergo metabolic phase shifts simultaneously, even when exposed to the identical microenvironmental conditions. To better understand and accurately replicate cellular behavior under diverse conditions, several modeling approaches—flux balance analysis (FBA) [ 50 , 18 , 72 , 39 ] , metabolic flux analysis (MFA) [ 48 , 49 , 2 , 3 , 59 , 64 , 76 , 57 , 15 , 38 ] , and kinetic modeling [ 37 , 49 , 20 , 21 , 22 ] —have demonstrated strong potential for analyzing metabolic fluxes and other critical cellular activities. These methods deepen the understanding of underlying metabolic mechanisms and support strategic optimization of culture conditions. FBA is an optimization-based approach that predicts feasible intracellular flux distributions using stoichiometric models (e.g., genome-scale networks) and assumed cellular objectives such as maximal growth or energy efficiency [ 2 ] . In contrast, MFA uses experimentally measured extracellular rates to infer intracellular fluxes without assuming that the cell operates in an optimal metabolic state. This approach is particularly effective for characterizing metabolic behavior under industrially relevant, non-optimal conditions—such as nutrient limitation or the presence of inhibitory compounds [ 4 ] . MFA therefore provides quantitative insight into how intracellular fluxes are redistributed across key metabolic pathways in response to environmental perturbations and process variations, as demonstrated in studies by Niklas et al. (2012) [ 48 ] and others. Stable isotope techniques, such as 13 C labeling, when integrated with MFA, provide deeper insights into intracellular states and metabolic pathways, enabling precise estimation of metabolic reaction rates. Several modeling tools have been developed for situations involving isotopic and/or metabolic non-steady states [ 38 , 2 , 3 ] . Complementing these approaches, kinetic models typically use Monod and Michaelis–Menten kinetics for metabolic regulatory mechanisms. Specifically tailored for CHO cell cultures, these models encompass key metabolic pathways like glycolysis, the pentose phosphate pathway (PPP), the tricarboxylic acid (TCA) cycle, and the respiratory chain. These models also consider the impact of energy shuttles (ATP/ADP) and cofactors (NADH, NAD + , NADPH, NADP + ), which are crucial for predicting cell responses to hypoxic perturbations. Notable studies, such as Ghorbaniaghdam et al. [ 20 , 21 , 22 ] and Nolan et al. (2011) [ 49 ] , demonstrate how these models characterize cell flux rate responses to environmental perturbations and calculate extracellular metabolite consumption/production rates. For a more comprehensive discussion of metabolic flux analysis and kinetic modeling, refer to Antoniewicz (2015) [ 2 ] and Kyriakopoulos et al. (2018) [ 37 ] . Building on these foundations, recent research has advanced multi-scale frameworks that link intracellular metabolism with bioreactor-level dynamics in CHO culture systems. Pennington et al. (2024) [ 54 , 53 ] developed hybrid dynamic metabolic flux models that couple enzyme-constrained stoichiometric formulations with reactor-level kinetics, enabling improved process optimization under uncertainty. Monteiro et al. (2023) [ 44 ] developed a multi-scale model predictive control framework integrating a reduced mechanistic metabolic model with reactor-level dynamics to optimize CHO cell culture feeding strategies. Similarly, Erklavec Zajec et al. (2021) [ 17 ] developed a detailed deterministic metabolic network model incorporating N-linked glycosylation reactions. Earlier studies by Karra et al. (2010) [ 34 ] and Bayrak et al. (2015) [ 7 ] utilized population-balance and agent-based modeling approaches to capture cell-to-cell heterogeneity within bioreactors. While existing multi-scale models have advanced cross-scale integration and process optimization, most remain deterministic or steady-state in nature and fail to explicitly capture stochastic single-cell kinetics, probabilistic metabolic-phase transitions, or trajectory-level prediction uncertainty. As a result, metabolite heterogeneity within cell populations and the inherent stochasticity of cell-culture processes remain poorly characterized [ 35 , 68 ] . This limitation leads to: (a) reduced reliability and lack of uncertainty quantification in predicting batch-to-batch variations; (b) incomplete understanding of metabolic shifts and environmental responses; (c) inadequate integration of diverse online and offline measurements of critical process parameters (CPPs) and CQAs—including dissolved oxygen, pH, viable cell density, metabolite concentrations, and product titer; and (d) suboptimal control strategies for end-to-end bioprocesses. To decode complex cellular metabolism and achieve well-controlled cultivation dynamics, this study introduces a multi-scale hybrid modeling framework that integrates molecular, cellular, and process-level mechanisms governing CHO cell cultures. The framework is designed to predict culture trajectories using readily available online measurements, along with initial cell density and metabolite concentrations. By leveraging insights from single-cell metabolism, it enhances the prediction and optimization of CHO culture performance. The model captures time-dependent metabolic regulation and phase transitions during culture progression, linking intracellular enzymatic activities to population-level kinetics. It comprises three interconnected components: (1) a stochastic mechanistic model describing single-cell metabolic reaction networks, (2) a probabilistic model representing asynchronous metabolic phase transitions, and (3) a macro-kinetic model characterizing population-level dynamics in heterogeneous cell populations. Together, these

Read the complete text