Morphology-driven non-invasive soft sensors for Chinese hamster ovary cell cultivation

Abstract

Traditional offline monitoring of Chinese hamster ovary cell fermentation processes suffers from severe systemic time delays. To achieve real-time monitoring of key physiological and biochemical parameters, this study proposes a non-invasive soft sensing framework based on cellular micro-morphology. Phase-contrast microscopic images were acquired throughout the fed-batch cultivation cycle, systematically extracting multidimensional features at the single-cell level, including geometric dimensions, shape, and internal optical texture. Given the highly non-linear mapping relationship between microscopic phenotypes and macroscopic metabolism revealed by univariate analysis $ \left( {\left| r \right| < 0.5} \right) $, a multivariate predictive model was constructed using the Random Forest (RF) algorithm, which was evaluated in parallel with a Partial Least Squares Regression model. Blind testing using an independent test batch subjected to an abnormal temperature disturbance demonstrated that the RF model effectively overcame the limitations of linear algorithms, achieving high-precision, cross-batch predictions for viable cell density (R² = 0.86), glucose (R² = 0.77), and lactate (R² = 0.90). Furthermore, from the perspective of model interpretability, feature analysis indicated that standard deviation features, representing the morphological heterogeneity of the cell population, showed a higher predictive contribution in characterizing the evolution of metabolic states than mean features. This study suggests the tremendous potential of microscopic population morphological heterogeneity as a digital biomarker for bioprocess monitoring, providing a reliable data-driven strategy for the at-line evaluation of industrial bioreactors.

Keywords

Chinese hamster ovary cells / Soft sensor / Microscopic morphology / Machine learning / Process analytical technology

References

[1] Liang G, Sha S, Wang Z, et al.. Soft-sensor model development for CHO growth/production, intracellular metabolite, and glycan predictions. Front Mol Biosci, 2024, 11: 1441885

[2] Bauer N, Oberist C, Poth M, et al.. Genomic barcoding for clonal diversity monitoring and control in cell-based complex antibody production. Sci Rep, 2024, 14: 14587

[3] Gomez N, Wieczorek A, Lu F, et al.. Culture temperature modulates half antibody and aggregate formation in a Chinese hamster ovary cell line expressing a bispecific antibody. Biotechnol Bioeng, 2018, 115(12): 2930-40

[4] Kaur R, Jain R, Budholiya N, et al.. Long term culturing of CHO cells: phenotypic drift and quality attributes of the expressed monoclonal antibody. Biotechnol Lett, 2023, 45(3): 357-70

[5] Sharma D, Singh K. AI-enhanced bioprocess technologies: machine learning implementations from upstream to downstream operations. World J Microbiol Biotechnol, 2025, 41: 278

[6] Suman S, Murr M, Crowe J, et al.. In-line prediction of viability and viable cell density through machine learning-based soft sensor modeling and an integrated systems approach: an industrially relevant PAT case study. Biotechnol Prog, 2025, 41(3): e3520

[7] Dabros M, Genasci G, Blanchard L, et al.. On-line monitoring and control of fed-batch fermentations in winemaking. Chimia, 2016, 70(12): 900

[8] Rodrigues KC, Sonego JL, Bernardo A, et al.. Real-time monitoring of bioethanol fermentation with industrial musts using mid-infrared spectroscopy. Ind Eng Chem Res, 2018, 57(32): 10823-31

[9] Radel S, Schnoller J, Groschl M, et al.. On chemical and ultrasonic strategies to improve a portable FT-IR ATR process analyzer for online fermentation monitoring. IEEE Sens J, 2010, 10(10): 1615-22

[10] Cavaglia J, Giussani B, Mestres M, et al.. Early detection of undesirable deviations in must fermentation using a portable FTIR-ATR instrument and multivariate analysis. J Chemom, 2019, 33(8): e3162

[11] Nettesheim P, Burggräf P, Steinberg F. Applications of machine learning in the brewing process: a systematic review. Discov Artif Intell, 2024, 4: 80

[12] de Juan A, de Oliveira RR. Hyperspectral image and chemometrics. A step beyond classical spectroscopic PAT tools. Anal Bioanal Chem, 2026, 418: 23-34

[13] Rathinavelu S, Pavan SS, Sivaprakasam S. Hybrid model-based framework for soft sensing and forecasting key process variables in the production of hyaluronic acid by Streptococcus zooepidemicus. Biotechnol Bioproc E, 2023, 28: 203-14

[14] Shi J, Ho A, Snyder CE, et al.. Accelerating biopharmaceutical cell line selection with label-free multimodal nonlinear optical microscopy and machine learning. Commun Biol, 2025, 8: 157

[15] Jeong J, Kim S. Application of machine learning for quantitative analysis of industrial fermentation using image processing. Food Sci Biotechnol, 2025, 34: 373-81

[16] Möller J, Rosenberg M, Riecken K, et al.. Quantification of the dynamics of population heterogeneities in CHO cultures with stably integrated fluorescent markers. Anal Bioanal Chem, 2020, 412: 2065-80

[17] Aleixandre-Tudo JL, Nieuwoudt H, Aleixandre JL, et al.. Chemometric compositional analysis of phenolic compounds in fermenting samples and wines using different infrared spectroscopy techniques. Talanta, 2018, 176: 526-36

[18] Borràs E, Ferré J, Boqué R, et al.. Data fusion methodologies for food and beverage authentication and quality assessment–a review. Anal Chim Acta, 2015, 891: 1-14

[19] Munawar AA, Meilina H, Pawelzik E. Near infrared spectroscopy as a fast and non-destructive technique for total acidity prediction of intact mango: comparison among regression approaches. Comput Electron Agric, 2022, 193: 106657

[20] Charabi I, Abidine MB, Fergani B, et al.. DeepF-SVM: a new hybrid deep learning model for enhanced sensor-based human activity recognition. Cluster Comput, 2025, 28: 910

[21] Barzegar R, Sattarpour M, Deo R, et al.. An ensemble tree-based machine learning model for predicting the uniaxial compressive strength of travertine rocks. Neural Comput Appl, 2020, 32: 9065-80

[22] Qin Y, Song K, Zhang N, et al.. Robust NIR quantitative model using MIC-SPA variable selection and GA-ELM. Infrared Phys Technol, 2023, 128: 104534

[23] Zhou X, Yang J, Su Y, et al.. Aggregation and assessment of grape quality parameters with visible-near-infrared spectroscopy: introducing a novel quantitative index. Postharvest Biol Technol, 2024, 218: 113131

[24] Kothari V, et al.. Segregated models of cell populations. Curr Opin Biotechnol, 2021, 69: 126-34

[25] Fredrickson AG, Mantzaris NV. A new set of population balance equations for microbial and cell cultures. Chem Eng Sci, 2002, 57(12): 2265-78

[26] Stamatakis M, Zygourakis K. A mathematical and computational approach for integrating the major sources of cell population heterogeneity. J Theor Biol, 2010, 266(1): 41-61

[27] Emmanuel I, Sun Y, Wang Z. A machine learning-based credit risk prediction engine system using a stacked classifier and a filter-based feature selection method. J Big Data, 2024, 11(1): 23

[28] Xiang Q, Xia Y, Song J, et al.. Characterization of the key nonvolatile metabolites in Cheddar cheese by partial least squares regression (PLSR). Food Chem, 2023, 403: 134034

[29] Wetzel SJ. Unsupervised learning of phase transitions: from principal component analysis to variational autoencoders. Phys Rev E, 2017, 96(2): 022140

[30] Chen T, Men J, Zhao M, et al.. The spectral fusion of LIBS and MIR coupled with random forest (RF) for the quantitative analysis of soil pH. J Anal Spectrom, 2021, 36(5): 1084-92

[31] Kapetanakis DS, Mangina E, Finn DP. Input variable selection for thermal load predictive models of commercial buildings. Energy Build, 2017, 137: 13-26

[32] Zhu Z, Chen X, Li W, et al.. Understanding the effect of temperature downshift on CHO cell growth, antibody titer and product quality by intracellular metabolite profiling and in vivo monitoring of redox state. Biotechnol Prog, 2023, 39(4): e3352

[33] Mellahi K, Brochu D, Gilbert M, et al.. Assessment of fed-batch cultivation strategies for an inducible CHO cell line. J Biotechnol, 2019, 298: 45-56

[34] Chen Z, Li X, Li P, et al.. Quantitative analysis of cancer cell morphology using digital holography technology under high temperature stimulation. Histochem Cell Biol, 2025, 163: 63

[35] Kalnoor G, Kadrolli V. Advanced leukocyte classification using attention mechanisms and dual channel U-Net architecture. Sci Rep, 2025, 15: 13825

[36] Bai H, Wang X, Guan Y, et al.. Blood cell counting based on U-Net + + and YOLOv5. Optoelectron Lett, 2023, 19: 370-6

[37] Wubineh BZ, Rusiecki A, Halawa K. Deep learning-based automatic segmentation and classification for cervical cancer detection using an improved U-Net and ensemble methods. Sci Rep, 2026, 16: 5184

[38] Siegl M, Geier D, Andreeßen B, et al.. Data synchronization techniques and their impact on the prediction performance of automated recalibrated soft sensors in bioprocesses. Biotechnol Bioproc E, 2024, 29: 929-41

[39] Bro R, Smilde AK. Principal component analysis. Anal Methods, 2014, 6(9): 2812-31

[40] Galvão RKH, Araújo MCU. Variable selection. Comprehensive Chemometrics. Elsevier; 2009. pp. 233–83. https://doi.org/10.1016/B978-044452701-1.00075-2.

[41] Wold S, Geladi P, Esbensen K, Öhman J. Multi-way principal components-and PLS-analysis. J Chemometrics, 1987, 1(1): 41-56

[42] Guo S, Popp J, Bocklitz T. Chemometric analysis in Raman spectroscopy from experimental design to machine learning based modeling. Nat Protoc, 2021, 16(12): 5426-59

[43] Breiman L. Random Forests Mach Learn, 2001, 45: 5-32

[44] Probst P. Hyperparameters and tuning strategies for random forest. WIREs Data Min Knowl Discov, 2019, 9: e1301

[45] Segal MR. Machine learning benchmarks and random forest regression. Center for Bioinformatics & Molecular Biostatistics, UCSF; 2004.

[46] Fernandez-Delgado M, et al.. Do we need hundreds of classifiers to solve real world classification problems?. J Mach Learn Res, 2014, 15(1): 3133-81

[47] Zavala-Ortiz DA, Denner A, Aguilar-Uscanga MG, et al.. Comparison of partial least square, artificial neural network, and support vector regressions for real-time monitoring of CHO cell culture processes using in situ near-infrared spectroscopy. Biotechnol Bioeng, 2022, 119(2): 535-49

[48] Alhuthali S, Kotidis P, Kontoravdi C. Osmolality effects on CHO cell growth, cell volume, antibody productivity and glycosylation. Int J Mol Sci, 2021, 22(7): 3290

[49] Hacker DL, et al.. 25 years of recombinant proteins from CHO cells. Pharm Bioprocess, 2009, 1(1): 89-104

Funding

National Key Research and Development Program of China(2022YFC3401304)

National Natural Science Foundation of China(32471540, 32371540, 21878128, and 31701582)

RIGHTS & PERMISSIONS

Jiangnan University