Cell Culture Process Research Articles

Silver nanoparticles induce formation of multi-protein aggregates that contain cadherin but do not colocalize with nanoparticles

Silver nanoparticles (AgNPs) are increasingly incorporated in diverse products to confer antimicrobial properties. They are released into the environment during manufacture, after disposal, and from the products during use. Because AgNPs bioaccumulate in brain, it is important to understand how they interact with neural cell physiology. We found that the focal adhesion (FA)-associated protein cadherin aggregated in a dose-dependent response to AgNP exposure in differentiating cultured B35 neuroblastoma cells. These aggregates tended to colocalize with F-actin inclusions that form in response to AgNP and also contain β-catenin. However, using hyperspectral microscopy, we demonstrate that these multi-protein aggregates did not colocalize with the AgNPs themselves. Furthermore, expression and organization of the FA protein vinculin did not change in cells exposed to AgNP. Our findings suggest that AgNPs activate an intermediate mechanism which leads to formation of aggregates via specific protein-protein interactions. Finally, we detail the changes in hyperspectral profiles of AgNPs during different stages of cell culture and immunocytochemistry processing. AgNPs in citrate-stabilized solution present mostly blue with some rainbow spectra and these are maintained upon mounting in Prolong Gold. Exposure to tissue culture medium results in a uniform green spectral shift that is not further altered by fixation and protein block steps of immunocytochemistry.

Integration of just-in-time learning with variational autoencoder for cell culture process monitoring based on Raman spectroscopy.

Protein production in the biopharmaceutical industry necessitates the utilization of multiple analytical techniques and control methodologies to ensure both safety and consistency. To facilitate real-time monitoring and control of cell culture processes, Raman spectroscopy has emerged as a versatile analytical technology. This technique, categorized as a Process Analytical Technology, employs chemometric models to establish correlations between Raman signals and key variables of interest. One notable approach for achieving real-time monitoring is through the application of just-in-time learning (JITL), an industrial soft sensor modeling technique that utilizes Raman signals to estimate process variables promptly. The conventional Raman-based JITL method relies on the K-nearest neighbor (KNN) algorithm with Euclidean distance as the similarity measure. However, it falls short of addressing the impact of data uncertainties. To rectify this limitation, this study endeavors to integrate JITL with a variational autoencoder (VAE). This integration aims to extract dominant Raman features in a nonlinear fashion, which are expressed as multivariate Gaussian distributions. Three experimental runs using different cell lines were chosen to compare the performance of the proposed algorithm with commonly utilized methods in the literature. The findings indicate that the VAE-JITL approach consistently outperforms partial least squares, convolutional neural network, and JITL with KNN similarity measure in accurately predicting key process variables.

Real-time model correction using Kalman filter for Raman-controlled cell culture processes

Raman spectroscopy has found extensive use in monitoring and controlling cell culture processes. In this context, the prediction accuracy of Raman-based models is of paramount importance. However, models established with data from manually fed-batch cultures often exhibit poor performance in Raman-controlled cultures. Thus, there is a need for effective methods to rectify these models. The objective of this paper is to investigate the efficacy of Kalman filter (KF) algorithm in correcting Raman-based models during cell culture. Initially, partial least squares (PLS) models for different components were constructed using data from manually fed-batch cultures, and the predictive performance of these models was compared. Subsequently, various correction methods including the PLS-KF-KF method proposed in this study were employed to refine the PLS models. Finally, a case study involving the auto-control of glucose concentration demonstrated the application of optimal model correction method. The results indicated that the original PLS models exhibited differential performance between manually fed-batch cultures and Raman controlled cultures. For glucose, the RMSEP of manually fed-batch culture and Raman-controlled culture was 0.23 and 0.40 g·L-1. With the implementation of model correction methods, there was a significant improvement in model performance within Raman-controlled cultures. The RMSEP for glucose from updating-PLS, KF-PLS, and PLS-KF-KF was 0.38, 0.36, and 0.17 g·L-1, respectively. Notably, the proposed PLS-KF-KF model correction method was found to be more effective and stable, playing a vital role in the automated nutrient feeding of cell cultures.

Comprehensive cell culture monitoring: Leveraging in-line Raman spectroscopy for enhanced biopharmaceutical manufacturing insights

Many biopharmaceutical manufacturers and researchers seek a comprehensive understanding and real-time monitoring of cell growth and metabolism to enhance the cell culture process. However, the biological complexity of cells and the intricacies of cell culture process have impeded the development of analytical and monitoring techniques. This study aims to present a complete picture of cell culture processes using in-line Raman spectroscopy. Initially, we proposed an efficient method for discriminating and removing anomalous spectra. Subsequently, Raman-based models for 27 components including amino acid, organic acid, lipid, organic alkali, alkaloid, alcohol, saccharide, biomass, physicochemical parameter, ion, and protein titer were established. The Q2 obtained by these models was greater than 0.8, and the relative percent difference (RPD) from external validation was greater than 2.0 except for glucose. Additionally, the proposed control charts effectively monitored normal batches and abnormal conditions such as bacterial contamination and insufficient feeding. In cases of unexplained culture abnormalities, the proposed workflow offered explanations.

3D Artificial Skin Platform for Investigating Pregnancy-Related Skin Pigmentation.

In this study, we created a 3D Artificial Skin Platform that can be used for the treatment of pigmentation by artificially realizing the skin of pregnant women. For the stable realization of 3D artificial skin, a bilayer hydrogel composed of collagen type I and fibrin was designed and applied to the study to reduce the tension-induced contraction of collagen type I, the extracellular matrix (ECM) of artificial skin, by dynamic culture. Oxygen concentration and 17β-Estradiol (E2) concentration, which are highly related to melanin production, were selected as parameters of the pregnancy environment and applied to cell culture. Oxygen concentration, which is locally reduced in the first trimester (2.5-3%), and E2, which is upregulated in the third trimester, were applied to the cell culture process. We analyzed whether the 3D artificial skin implemented in the 3D Artificial Skin Platform could better represent the tendency of melanin expression in pregnant women than cells cultured under the same conditions in 2D. The expression levels of melanin and melanin-related genes in the 2D cell culture did not show a significant trend that was similar to the melanin expression trend in pregnant women. However, the 3D artificial skin platform showed a significant trend towards a 2-6-fold increase in melanin expression in response to low oxygen concentrations (2.5%) and E2 concentrations (17 ng/mL), which was similar to the trend in pregnant women in vivo. These results suggest that 3D artificial skin cultured on the Artificial Skin Platform has the potential to be used as a substitute for human pregnant skin in various research fields related to the treatment of pigmentation.

Investigation of fouling in perfusion cell culture processes using alternating tangential flow filtration

Microfiltration membranes are increasingly used to retain the cells inside bioreactors while continuous harvest of the desired biopharmaceutical occurs in perfusion processes. One method of microfiltration is Alternating Tangential Flow (ATF) filtration, which involves moving the cultivation broth tangentially back and forth across the membrane, which results in a backwash effect that reduces fouling. In this study, fouling was investigated with asymmetric polysulfone hollow fibers operated in ATF mode attached to a bioreactor producing a recombinant protein in Chinese Hamster Ovary (CHO) cells. Fouling was assessed through different approaches, including determination of critical flux using an improved flux-step method. Fouling was studied through measurements of Transmembrane Pressure (TMP), protein transmission, membrane pore size and staining of the membrane after operation to visualize the distribution of biological fouling inside the membrane. For critical flux determination, fluxes of up to 69 LMH were used without exceeding the critical flux. No sign of fouling was observed for the short-term (<3.5 h) critical flux experiment. However, during prolonged operation at 8.3 LMH the TMP jumped to 0.9–0.95 bar, indicating fouling. At this state, the protein transmission remains at the same high level (>88 %).

Image‐based hybrid model incorporating initial spatial distribution for mesenchymal stem cell cultivation process design

AbstractThis work presents an image‐based hybrid model incorporating the initial spatial distribution for mesenchymal stem cell (MSC) cultivation process design. First, three levels of seeding bias in static MSC cultivation were experimentally investigated, and the resulting initial distribution was quantified using phase contrast microscopy image analysis. Second, the observed spatial heterogeneity was defined as a new parameter by calculating the standard deviation of the seeded cell number fraction among numerical 8 × 8 tiles on two‐dimensional plates. Third, the parameter was incorporated into kinetic models to consider spatial growth limitation. The model was then applied to simulate MSC cultivation processes with experimental data. Using dynamic and stochastic simulation outputs, feasible ranges of cell‐harvesting time could be determined to satisfy a given requirement for the minimum cell number and acceptable confluency level. The developed hybrid model could serve as a basis for quantitative decision‐making in the design of MSC cultivation processes.

Selection of Leptin Surrogates by a General Phenotypic Screening Method for Receptor Agonists.

There is a high demand for agonist biomolecules such as cytokine surrogates in both biological and medicinal research fields. These are typically sourced through natural ligand engineering or affinity-based screening, followed by individual functional validation. However, efficient screening methods for identifying rare hits within immense libraries are very limited. In this research article, we introduce a phenotypic screening method utilizing biological receptor activation-dependent cell survival (BRADS). This method offers a high-throughput, low-background, and cost-effective approach that can be implemented in virtually any biochemical laboratory setting. As a proof-of-concept, we successfully identified a surrogate for human leptin following a two-week cell culture process, without the need for specialized high-throughput equipment or reagents. This surrogate effectively emulates the activity of native human leptin in cell validation assays. Our findings not only underscore the effectiveness of BRADS but also suggest its potential applicability to a broad range of biological receptors, including Notch and GPCRs.

Autonomous differentiation of transgenic cells requiring no external hormone application: the endogenous gene expression and phytohormone behaviors.

The ectopic overexpression of developmental regulator (DR) genes has been reported to improve the transformation in recalcitrant plant species because of the promotion of cellular differentiation during cell culture processes. In other words, the external plant growth regulator (PGR) application during the tissue and cell culture process is still required in cases utilizing DR genes for plant regeneration. Here, the effect of Arabidopsis BABY BOOM (BBM) and WUSCHEL (WUS) on the differentiation of tobacco transgenic cells was examined. We found that the SRDX fusion to WUS, when co-expressed with the BBM-VP16 fusion gene, significantly influenced the induction of autonomous differentiation under PGR-free culture conditions, with similar effects in some other plant species. Furthermore, to understand the endogenous background underlying cell differentiation toward regeneration, phytohormone and RNA-seq analyses were performed using tobacco leaf explants in which transgenic cells were autonomously differentiating. The levels of active auxins, cytokinins, abscisic acid, and inactive gibberellins increased as cell differentiation proceeded toward organogenesis. Gene Ontology terms related to phytohormones and organogenesis were identified as differentially expressed genes, in addition to those related to polysaccharide and nitrate metabolism. The qRT-PCR four selected genes as DEGs supported the RNA-seq data. This differentiation induction system and the reported phytohormone and transcript profiles provide a foundation for the development of PGR-free tissue cultures of various plant species, facilitating future biotechnological breeding.

Abstract 2060: Automation of 3D cancer spheroid assay for compound screening

Abstract Finding efficient drug combinations to treat cancer patients is critical for therapy success. Accordingly, there is a critical need to develop methods for efficient testing drug efficacy to discover new therapeutic targets. 3D cancer models are highly valuable tools for cancer research and drug development. However, the complexity of performing 3D assays remains a hurdle for adopting these methods for compound screening. To alleviate the bottlenecks that come with labor-intensive manual protocols we developed cell culture automation methods using CellXpress.ai instrument. The instrument enables automation of 2D or 3D culture for prolonged complex workflows. The CellXpress.ai provides automated plating, passaging, media exchanges, monitoring organoids, alone with compound treatment and the end point assays with high content image analysis. In the present study we describe full automation of imaging, analysis, and cell culture process which enables scaling up complex 3D cell-based assays and compound screening. As an example of 3D protocol, we automated culture and imaging of colorectal cancer 3D spheroids formed from HCT116 cell line in U-shape low attachment plates. HCT116 cells were expanded in 2D, then spheroids were formed after automated dispensing of cell suspension into U-shape 96 or 384 plates. After 48h spheroids were treated with 12 anti-cancer compounds, at multiple concentrations for 3-5 days, then stained and imaged. Cell plating, compound additions, media exchange, and staining were performed automatically by CellXpressAI instrument. During the culture spheroids were monitored using transmitted light, with immediate analysis of phenotypic changes, including inhibition of growth or spheroid disintegration. For endpoint assay spheroids were stained with combination of Hoechst nuclear stain and viability dyes Calcein AM and EtHD, then imaged, and analyzed for multiple read-outs including spheroid size, marker intensities, and live-dead cell scoring. We observed concentration-dependent inhibition of spheroid growth and cell death in response to anti-cancer compounds and evaluated effective compound concentrations for cytotoxic or cytostatic effects. Cell culture process automation for 3D assays, powered by imaging and image analysis has a potential to bring 3D biology into another level, allowing to increase throughput and reproducibility, and enabling variety of high throughput drug discovery and precision medicine applications. Citation Format: Oksana Sirenko, Angeline Lim, Astrid Michlmayr, Zhisong Tong, Emilie Keidel, Felix Spira. Automation of 3D cancer spheroid assay for compound screening [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2060.

Potential Druggability of Mesenchymal Stem/Stromal Cell-derived Exosomes.

Exosomes secreted by mesenchymal stem/stromal cells (MSC-Exos) are advantageous candidate sources for novel acellular therapy. Despite the current standards of good manufacturing practice (GMP), the deficiency of suitable quality-control methods and the difficulties in large-scale preparation largely restrict the development of therapeutic products and their clinical applications worldwide. Herein, we mainly focus on three dominating issues commonly encountered in exosomal GMP, including issues upstream of the cell culture process, downstream of the purification process, exosomes quality control, and the drug properties of exosomes and their druggability from a corporate perspective. Collectively, in this review article, we put forward the issues of preparing clinical exosome drugs for the treatment of diverse diseases and provide new references for the clinical application of GMP-grade MSC-Exos.

Integer optimization models and algorithms for the multi-period non-shareable resource allocation problem

The resource allocation problem (RAP) determines a solution to optimally allocate limited resources to several activities or tasks. In this study, we propose a novel resource allocation problem referred to as multi-period non-shareable resource allocation problem (MNRAP), which is motivated by the characteristics of resources considered in the stem cell culture process for producing stem cell therapeutics. A resource considered in the MNRAP has the following three characteristics: (i) resource consumption required to perform an activity and available resource capacity may change over time; (ii) multiple activities cannot share one resource; and (iii) resource requirements can be satisfied through the combination of different types of resources. The MNRAP selects some of the given activities to maximize the overall profit under limited resources with these characteristics. To address this problem, pattern-based integer programming formulations based on the concept of resource patterns are proposed. These formulations attempt to overcome the limitations of a compact integer programming formulation, the utilization of which is challenging for large-scale problems owing to their complexity. Further, based on a branch-and-price approach to solving pattern-based formulations, effective heuristic algorithms are proposed to provide high-quality solutions for large instances. Moreover, through computational experiments on a wide range of instances, including real-world instances, the superiority of the proposed formulations and heuristic algorithms is demonstrated.

Elucidating uptake and metabolic fate of dipeptides in CHO cell cultures using 13C labeling experiments and kinetic modeling

The rapidly growing market of biologics including monoclonal antibodies has stimulated the need to improve biomanufacturing processes including mammalian host systems such as Chinese Hamster Ovary (CHO) cells. Cell culture media formulations continue to be enhanced to enable intensified cell culture processes and optimize cell culture performance. Amino acids, major components of cell culture media, are consumed in large amounts by CHO cells. Due to their low solubility and poor stability, certain amino acids including tyrosine, leucine, and phenylalanine can pose major challenges leading to suboptimal bioprocess performance. Dipeptides have the potential to replace amino acids in culture media. However, very little is known about the cleavage, uptake, and utilization kinetics of dipeptides in CHO cell cultures. In this study, replacing amino acids, including leucine and tyrosine by their respective dipeptides including but not limited to Ala-Leu and Gly-Tyr, supported similar cell growth, antibody production, and lactate profiles. Using 13C labeling techniques and spent media studies, dipeptides were shown to undergo both intracellular and extracellular cleavage in cultures. Extracellular cleavage increased with the culture duration, indicating cleavage by host cell proteins that are likely secreted and accumulate in cell culture over time. A kinetic model was built and for the first time, integrated with 13C labeling experiments to estimate dipeptide utilization rates, in CHO cell cultures. Dipeptides with alanine at the N-terminus had a higher utilization rate than dipeptides with alanine at the C-terminus and dipeptides with glycine instead of alanine at N-terminus. Simultaneous supplementation of more than one dipeptide in culture led to reduction in individual dipeptide utilization rates indicating that dipeptides compete for the same cleavage enzymes, transporters, or both. Dipeptide utilization rates in culture and cleavage rates in cell-free experiments appeared to follow Michaelis-Menten kinetics, reaching a maximum at higher dipeptide concentrations. Dipeptide utilization behavior was found to be similar in cell-free and cell culture environments, paving the way for future testing approaches for dipeptides in cell-free environments prior to use in large-scale bioreactors. Thus, this study provides a deeper understanding of the fate of dipeptides in CHO cell cultures through an integration of cell culture, 13C labeling, and kinetic modeling approaches providing insights in how to best use dipeptides in media formulations for robust and optimal mammalian cell culture performance.

Enhancing real-time cell culture monitoring: Automated Raman model optimization with Taguchi method.

Raman spectroscopy has found widespread usage in monitoring cell culture processes both in research and practical applications. However, commonly, preprocessing methods, spectral regions, and modeling parameters have been chosen based on experience or trial-and-error strategies. These choices can significantly impact the performance of the models. There is an urgent need for a simple, effective, and automated approach to determine a suitable procedure for constructing accurate models. This paper introduces the adoption of a design of experiment (DoE) method to optimize partial least squares models for measuring the concentration of different components in cell culture bioreactors. The experimental implementation utilized the orthogonal test table L25(56). Within this framework, five factors were identified as control variables for the DoE method: the window width of Savitzky-Golay smoothing, the baseline correction method, the order of preprocessing steps, spectral regions, and the number of latent variables. The evaluation method for the model was considered as a factor subject to noise. The optimal combination of levels was determined through the signal-to-noise ratio response table employing Taguchi analysis. The effectiveness of this approach was validated through two cases, involving different cultivation scales, different Raman spectrometers, and different analytical components. The results consistently demonstrated that the proposed approach closely approximated the global optimum, regardless of data set size, predictive components, or the brand of Raman spectrometer. The performance of models recommended by the DoE strategy consistently surpassed those built using raw data, underscoring the reliability of models generated through this approach. When compared to exhaustive all-combination experiments, the DoE approach significantly reduces calculation times, making it highly practical for the implementation of Raman spectroscopy in bioprocess monitoring.

An innovative hybrid modeling approach for simultaneous prediction of cell culture process dynamics and product quality.

The use of hybrid models is extensively described in the literature to predict the process evolution in cell cultures. These models combine mechanistic and machine learning methods, allowing the prediction of complex process behavior, in the presence of many process variables, without the need to collect a large amount of data. Hybrid models cannot be directly used to predict final product critical quality attributes, or CQAs, because they are usually measured only at the end of the process, and more mechanistic knowledge is needed for many classes of CQAs. The historical models can instead predict the CQAs better; however, they cannot directly relate manipulated process parameters to final CQAs, as they require knowledge of the process evolution. In this work, we propose an innovative modeling approach based on combining a hybrid propagation model with a historical data-driven model, that is, the combined hybrid model, for simultaneous prediction of full process dynamics and CQAs. The performance of the combined hybrid model was evaluated on an industrial dataset and compared to classical black-box models, which directly relate manipulated process parameters to CQAs. The proposed combined hybrid model outperforms the black-box model by 33% on average in predicting the CQAs while requiring only around half of the data for model training to match performance. Thus, in terms of model accuracy and experimental costs, the combined hybrid model in this study provides a promising platform for process optimization applications.

Fed-batch performance profiles for mAb production using different intensified N - 1 seed strategies are CHO cell-line dependent.

Recent optimizations of cell culture processes have focused on the final seed scale-up step (N - 1 stage) used to inoculate the production bioreactor (N-stage bioreactor) to enable higher inoculation cell densities (2-20 × 106 cells/mL), which could shorten the production culture duration and/or increase the volumetric productivity. N - 1 seed process intensification can be achieved by either non-perfusion (enriched-batch or fed-batch) or perfusion culture to reach those higher final N - 1 viable cell densities (VCD). In this study, we evaluated how different N - 1 intensification strategies, specifically enriched-batch (EB) N - 1 versus perfusion N - 1, affect cell growth profiles and monoclonal antibody (mAb) productivity in the final N-stage production bioreactor operated in fed-batch mode. Three representative Chinese Hamster Ovary (CHO) cell lines producing different mAbs were cultured using either EB or perfusion N - 1 seeds and found that the N-stage cell growth and mAb productivities were comparable between EB N - 1 and perfusion N - 1 conditions for two of the cell lines but were very different for the third. In addition, within the two similar cell growth cell lines, differences in cell-specific productivity were observed. This suggests that the impact of the N - 1 intensification process on production was cell-line dependent. This study revealed that the N - 1 intensification strategy and the state of seeds from the different N - 1 conditions may affect the outcome of the N production stage, and thus, the choice of N - 1 intensification strategy could be a new target for future upstream optimization of mAb production.

Small extracellular vesicles purification and scale-up.

Exosomes are small extracellular vesicles (sEVs) secreted by cells. With advances in the study of sEVs, they have shown great potential in the diagnosis and treatment of disease. However, sEV therapy usually requires a certain dose and purity of sEVs to achieve the therapeutic effect, but the existing sEV purification technology exists in the form of low yield, low purity, time-consuming, complex operation and many other problems, which greatly limits the application of sEVs. Therefore, how to obtain high-purity and high-quality sEVs quickly and efficiently, and make them realize large-scale production is a major problem in current sEV research. This paper discusses how to improve the purity and yield of sEVs from the whole production process of sEVs, including the upstream cell line selection and cell culture process, to the downstream isolation and purification, quality testing and the final storage technology.

Robust platform for inline Raman monitoring and control of perfusion cell culture.

Perfusion cell culture has been gaining increasing popularity for biologics manufacturing due to benefits such as smaller footprint, increased productivity, consistent product quality and manufacturing flexibility, cost savings, and so forth. Process Analytics Technologies tools are highly desirable for effective monitoring and control of long-running perfusion processes. Raman has been widely investigated for monitoring and control of traditional fed batch cell culture process. However, implementation of Raman for perfusion cell culture has been very limited mainly due to challenges with high-cell density and long running times during perfusion which cause extremely high fluorescence interference to Raman spectra and consequently it is exceedingly difficult to develop robust chemometrics models. In this work, a platform based on Raman measurement of permeate has been proposed for effective analysis of perfusion process. It has been demonstrated that this platform can effectively circumvent the fluorescence interference issue while providing rich and timely information about perfusion dynamics to enable efficient process monitoring and robust bioreactor feed control. With the highly consistent spectral data from cell-free sample matrix, development of chemometrics models can be greatly facilitated. Based on this platform, Raman models have been developed for good measurement of several analytes including glucose, lactate, glutamine, glutamate, and permeate titer. Performance of Raman models developed this way has been systematically evaluated and the models have shown good robustness against changes in perfusion scale and variations in permeate flowrate; thus models developed from small lab scale can be directly transferred for implementation in much larger scale of perfusion. With demonstrated robustness, this platform provides a reliable approach for automated glucose feed control in perfusion bioreactors. Glucose model developed from small lab scale has been successfully implemented for automated continuous glucose feed control of perfusion cell culture at much larger scale.

Effect of inner diameter, filter length, and pore size on hollow fiber filter fouling during perfusion cell culture.

As the need for higher volumetric productivity in biomanufacturing grows, biopharmaceutical companies are increasingly investing in a perfusion cell culture process, most commonly one that uses a hollow fiber filter as the cell retention device. A current challenge with using hollow fiber filters is fouling of the membrane, which reduces product sieving and can increase transmembrane pressure (TMP) past process limitations. In this work, the impact of hollow fiber filter geometries on product sieving and hydraulic membrane resistance profiles is evaluated in a tangential flow filtration (TFF) perfusion system. The hollow fibers tested had lengths ranging from 19.8 to 41.5 cm, inner diameters (IDs) ranging from 1.0 to 2.6 mm, and pore sizes of 0.2 or 0.65 μm. The results showed that the shortest hollow fibers experienced higher product sieving while larger IDs contributed to both higher product sieving and lower hydraulic membrane resistances, illustrating the impact of filter geometry on process performance. The results also showed 0.2 μm pore size filters maintain higher product sieving, but also higher membrane resistances compared to 0.65 μm pore size filters. This study highlights the need for optimized hollow fiber filter geometries to maximize use of the membrane area, which in turn can reduce production costs and increase scalability of the perfusion process.

Enhanced parameter estimation with improved particle swarm optimization algorithm for cell culture process modeling

AbstractMechanism models and model‐assisted process optimization play a vital role in biopharmaceutical industry. However, the parameter estimation remains a challenging task in modeling due to the presence of multiple local optima and ill‐conditioning. In this study, we firstly presented a comparative evaluation of particle swarm optimization (PSO), genetic algorithm (GA), and simulated annealing (SA) for parameter estimation and confirmed PSO was the most efficient approach. Then an improved PSO algorithm using second‐order oscillation and particle replacement strategies was developed to enhance fitting performance in fed‐batch cell culture modeling. The results of fitting and model validation demonstrated that the improved PSO could explore more reasonable parameters and leading to a significantly enhancement in predictive accuracy throughout the entire cell culture process under varying conditions. Moreover, 10 diverse fed‐batch experiments were conducted to validate the fitting abilities on different process condition and clones, the improved PSO method exhibited improvements in accuracy and universality of parameter estimation for modeling various cultivation processes, particularly those lacking any prior knowledge. This improved algorithm is implemented to make it available to both the scientific community and industry, offering customized solutions for specific projects.