No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Synergising stoichiometric modelling with artificial neural networks to predict antibody glycosylation patterns in Chinese hamster ovary cells
Abstract
In-process quality control of biotherapeutics, such as monoclonal antibodies, requires computationally efficient process models that use readily measured process variables to compute product quality. Existing kinetic cell culture models can effectively describe the underlying mechanisms but require considerable development and parameterisation effort. Stoichiometric models, on the other hand, provide a generic, parameter-free means for describing metabolic behaviour but do not extend to product quality prediction. We have overcome this limitation by integrating a stoichiometric model of Chinese hamster ovary (CHO) cell metabolism with an artificial neural network that uses the fluxes of nucleotide sugar donor synthesis to compute the profile of Fc N-glycosylation, a critical quality attribute of antibody therapeutics. We demonstrate that this hybrid framework accurately computes glycan distribution profiles using a set of easy-to-obtain experimental data, thus providing a platform for process control applications.
... They identified the optimal kinetic model structure to predict the key state variables, and optimize the production process of lutein from microalgae [122]. In the process of quality control of biotherapeutics, such as monoclonal antibodies, Antonakoudis et al. integrated a stoichiometric model with an artificial neural network to predict the glycosylation profile in CHO cell cultivations [123]. With this hybrid model, the glycan distribution profiles can be computed with accuracy and thus a platform is provided for process control in biotherapeutics production [123]. ...
... In the process of quality control of biotherapeutics, such as monoclonal antibodies, Antonakoudis et al. integrated a stoichiometric model with an artificial neural network to predict the glycosylation profile in CHO cell cultivations [123]. With this hybrid model, the glycan distribution profiles can be computed with accuracy and thus a platform is provided for process control in biotherapeutics production [123]. ...
The Valley of Death confronts industrial biotechnology with a significant challenge to the commercialization of products. Fortunately, with the integration of computation, automation and artificial intelligence (AI) technology, the industrial biotechnology accelerates to cross the Valley of Death. The Fourth Industrial Revolution (Industry 4.0) has spurred advanced development of intelligent biomanufacturing, which has evolved the industrial structures in line with the worldwide trend. To achieve this, intelligent biomanufacturing can be structured into three main parts that comprise digitalization, modeling and intellectualization, with modeling forming a crucial link between the other two components. This paper provides an overview of mechanistic models, data-driven models and their applications in bioprocess development. We provide a detailed elaboration of the hybrid model and its applications in bioprocess engineering, including strain design, process control and optimization, as well as bioreactor scale-up. Finally, the challenges and opportunities of biomanufacturing towards Industry 4.0 are also discussed.
... process control (Schinn et al., 2021a), identifying burdensome host cell protein for knockout (Kol et al., 2020), predicting product glycosylation (Antonakoudis et al., 2021), and designing feeding strategies to optimize monoclonal antibody production (Fouladiha et al., 2020). While some attempts have been made to measure CHO cell GEM predictive performance (Schinn et al., 2021b;Széliová, 2021aSzéliová, , 2021b, the effectiveness of CHO cell GEMs at capturing both intracellular flux distributions and extracellular phenotypes has not been comprehensively assessed. ...
... Cell line-specific models can, in theory, better capture metabolic features of the cell line at hand leading to improved model predictions and therefore more accurate model-led decision-making, for example, for cell line and process engineering. Furthermore, it has recently been demonstrated that a small-scale CHO cell stoichiometric model can outperform full-size CHO cell GEMs in growth rate predictions(Antonakoudis et al., 2021). Reducing the number of reactions within the GEM may therefore bring predictive performance in line with smaller stoichiometric models by reducing how underdetermined the GEM is while still maintaining the in-depth metabolic descriptive ability of a full GEM. ...
Genome‐scale metabolic models (GEMs) possess the power to revolutionize bioprocess and cell line engineering workflows thanks to their ability to predict and understand whole‐cell metabolism in silico. Despite this potential, it is currently unclear how accurately GEMs can capture both intracellular metabolic states and extracellular phenotypes. Here, we investigate this knowledge gap to determine the reliability of current Chinese hamster ovary (CHO) cell metabolic models. We introduce a new GEM, iCHO2441, and create CHO‐S and CHO‐K1 specific GEMs. These are compared against iCHO1766, iCHO2048, and iCHO2291. Model predictions are assessed via comparison with experimentally measured growth rates, gene essentialities, amino acid auxotrophies, and ¹³C intracellular reaction rates. Our results highlight that all CHO cell models are able to capture extracellular phenotypes and intracellular fluxes, with the updated GEM outperforming the original CHO cell GEM. Cell line‐specific models were able to better capture extracellular phenotypes but failed to improve intracellular reaction rate predictions in this case. Ultimately, this work provides an updated CHO cell GEM to the community and lays a foundation for the development and assessment of next‐generation flux analysis techniques, highlighting areas for model improvements.
... Also, prospective applications of artificial NN for modeling cell culture (Antonakoudis et al., 2021), optimizing media composition (Bashokouh et al., 2019), predicting protein aggregation (Budholiya et al., 2020), and of recurrent NN for predicting mAb titer, VCC, cell viability and glucose/lactate concentration (Smiatek et al., 2021) are being explored (refer Illustration 3). Of late, tree-based models, able to model nonlinear systems with insensitivity to missing values/ outliers, are being implemented in fault detection heuristics (Shrivastava et al., 2017) and predicting biomass and cell metabolites (Thompson et al., 2019;Xu et al., 2021). ...
... To compute the profile of Fc N-glycosylation, a critical quality attribute of antibody therapeutics Accurately computes glycan distribution profiles Provides a platform for process control applications Antonakoudis et al. (2021) 8 Reinforcement learning ...
Advanced control strategies are well established in chemical, pharmaceutical, and food processing industries. Over the past decade, the application of these strategies is being explored for control of bioreactors for manufacturing of biotherapeutics. Most of the industrial bioreactor control strategies apply classical control techniques, with the control system designed for the facility at hand. However, with the recent progress in sensors, machinery, and industrial internet of things, and advancements in deeper understanding of the biological processes, coupled with the requirement of flexible production, the need to develop a robust and advanced process control system that can ease process intensification has emerged. This has further fuelled the development of advanced monitoring approaches, modeling techniques, process analytical technologies, and soft sensors. It is seen that proper application of these concepts can significantly improve bioreactor process performance, productivity, and reproducibility. This review is on the recent advancements in bioreactor control and its related aspects along with the associated challenges. This study also offers an insight into the future prospects for development of control strategies that can be designed for industrial‐scale production of biotherapeutic products.
... By illustrating the influence on glycosylation stereoselectivity from factors like different nucleophiles, electrophiles, catalysts, and solvents, this model provides researchers new insights into the glycan biosynthesis reactions. Another method was developed by Antonakoudis et al. (2021), which employed a Powerful and easy to use glycoinformatics tools are being developed for exploring across different glycosylation databases (e.g., GlycanFormatConverter ( Tsuchiya et al., 2019) and GlycoGlyph (Mehta and Cummings, 2020)). ...
... Biologists usually aim for understanding molecular bases of phenotypes, so deciphering the hidden mechanisms driving changes in glycosylation in diseases is important for improving the efficacy and safety of treatments and therapeutic proteins (Pörtner, 2014). It also can help avoid diseases associated with malfunctioned glycosylation , and save time and resources from erroneous experiments (Antonakoudis et al., 2021). Here we discuss the state-ofthe-art approaches that could be deployed to open the 'black box' in AI-based glycan analyses (Saghaleyni, 2021). ...
Glycans are complex, yet ubiquitous across biological systems. They are involved in diverse essential organismal functions. Aberrant glycosylation may lead to disease development, such as cancer, autoimmune diseases, and inflammatory diseases. Glycans, both normal and aberrant, are synthesized using extensive glycosylation machinery, and understanding this machinery can provide invaluable insights for diagnosis, prognosis, and treatment of various diseases. Increasing amounts of glycomics data are being generated thanks to advances in glycoanalytics technologies, but to maximize the value of such data, innovations are needed for analyzing and interpreting large-scale glycomics data. Artificial intelligence (AI) provides a powerful analysis toolbox in many scientific fields, and here we review state-of-the-art AI approaches on glycosylation analysis. We further discuss how models can be analyzed to gain mechanistic insights into glycosylation machinery and how they shape glycans under different scenarios. Finally, we further propose how to leverage the gained knowledge for developing predictive AI-based models of glycosylation. Thus, guiding future research of AI-based glycosylation model development will provide insightful insights into glycosylation and glycan machinery.
... This has been demonstrated by the recent surge in their application for large-scale multi-omics analysis of cancer datasets (Chaudhary et al., 2018;Li et al., 2022;Nicora et al., 2020;Opdam et al., 2017;Poirion et al., 2021;Wang et al., 2021). In the context of biotherapeutic production, ML has been applied for genome-scale modeling of CHO metabolic networks (Schinn et al., 2021;Zampieri et al., 2019) and for predicting protein glycosylation patterns in CHO cells (Antonakoudis et al., 2021;Kotidis and Kontoravdi, 2020). ...
Gene therapy is poised to transition from niche to mainstream medicine, with recombinant adeno-associated virus (rAAV) as the vector of choice. However, this requires robust, scalable, industrialized production to meet demand and provide affordable patient access, which has thus far failed to materialize. Closing the chasm between demand and supply requires innovation in biomanufacturing to achieve the essential step change in rAAV product yield and quality. Omics provides a rich source of mechanistic knowledge that can be applied to HEK293, the prevailing cell line for rAAV production. In this review, the findings from a growing number of disparate studies that apply genomics, epigenomics, transcriptomics, proteomics, and metabolomics to HEK293 bioproduction are explored. Learnings from CHO-Omics, application of omics approaches to improve CHO bioproduction, provide context for the potential of "HEK-Omics" as a multiomics-informed approach providing actionable mechanistic insights for improved transient and stable production of rAAV and other recombinant products in HEK293.
... In particular, they appear to demonstrate superior performance for smaller datasets with a reduced risk of overfitting. Finally, hybrid approaches that combine stoichiometric modelling and ANNs have also been used to predict glycan abundance and provide metabolic insight, though the ANNs were relatively small due to data size constraints and the risk of overfitting [22]. ...
N-glycosylation can have a profound effect on the quality of mAb therapeutics. In biomanufacturing, one of the ways to influence N-glycosylation patterns is by altering the media used to grow mAb cell expression systems. Here, we explore the potential of machine learning (ML) to forecast the abundances of N-glycan types based on variables related to the growth media. The ML models exploit a dataset consisting of detailed glycomic characterisation of Anti-HER fed-batch bioreactor cell cultures measured daily under 12 different culture conditions, such as changes in levels of dissolved oxygen, pH, temperature, and the use of two different commercially available media. By performing spent media quantitation and subsequent calculation of pseudo cell consumption rates (termed media markers) as inputs to the ML model, we were able to demonstrate a small subset of media markers (18 selected out of 167 mass spectrometry peaks) in a Chinese Hamster Ovary (CHO) cell cultures are important to model N-glycan relative abundances (Regression - correlations between 0.80–0.92; Classification - AUC between 75.0–97.2). The performances suggest the ML models can infer N-glycan critical quality attributes from extracellular media as a proxy. Given its accuracy, we envisage its potential applications in biomaufactucuring, especially in areas of process development, downstream and upstream bioprocessing.
... Similarly, Antonakoudis et al. developed a hybrid modeling framework that combines mechanistic information in the form of a stoichiometric model with a deep artificial neural network to predict antibody glycosylation patterns in Chinese hamster ovary cells [19]. By training the neural network with data on bioprocess variables such as metabolite fluxes, cell culture parameters, antibody quality parameters, and glycosylation pathways, the researchers leveraged deep neural network principles to enhance the prediction of product quality (glycan distribution) in bioprocesses involving CHO cells. ...
The use of machine learning and deep learning has become prominent within various fields of bioprocessing for countless modeling and prediction tasks. Previous reviews have emphasized machine learning applications in various fields of bioprocessing, including biomanufacturing. This comprehensive review highlights many of the different machine learning and multivariate analysis techniques that have been utilized within Chinese hamster ovary cell biomanufacturing, specifically due to their rising significance in the industry. Applications of machine and deep learning within other bioprocessing industries are also briefly discussed.
... IVCD is an important metric in cell culture that is directly related to the specific rates of product formation, nutrient consumption and cell growth. The cumulative viable cell concentration over the culture period up to time t-denoted by IVCD t -was calculated using the trapezoidal rule by Equation (1): [34] ...
Bioreactor scale‐up and scale‐down have always been a topical issue for the biopharmaceutical industry and despite considerable effort, the identification of a fail‐safe strategy for bioprocess development across scales remains a challenge. With the ubiquitous growth of digital transformation technologies, new scaling methods based on computer models may enable more effective scaling. This study aimed to evaluate the potential application of machine learning (ML) algorithms for bioreactor scale‐up, with a specific focus on the prediction of scaling parameters. Factors critical to the development of such models were identified and data for bioreactor scale‐up studies involving CHO cell‐generated mAb products collated from the literature and public sources for the development of unsupervised and supervised ML models. Comparison of bioreactor performance across scales identified similarities between the different processes and primary differences between small‐ and large‐scale bioreactors. A series of three case studies were developed to assess the relationship between cell growth and scale‐sensitive bioreactor features. An embedding layer improved the capability of artificial neural network models to predict cell growth at a large‐scale, as this approach captured similarities between the processes. Further models constructed to predict scaling parameters demonstrated how ML models may be applied to assist the scaling process. The development of data sets that include more characterization data with greater variability under different gassing and agitation regimes will also assist the future development of ML tools for bioreactor scaling.
... The framework was tested in three independent fed-batch experiments and demonstrated an error rate of only 0.27%. This platform has potential applications in process control, cell line selection, and metabolic engineering [101]. Since cutting-edge biopharmaceuticals including virus-like particles, exosomes, cell, and gene therapies, along with recombinant proteins and peptides lack any platform production approach, there is a pressing need for intensified and quicker product development procedures [102]. ...
The increasing demand for biosimilar monoclonal antibodies (mAbs) has prompted the development of stable high-producing cell lines while simultaneously decreasing the time required for screening. Existing platforms have proven inefficient, resulting in inconsistencies in yields, growth characteristics, and quality features in the final mAb products. Selecting a suitable expression host, designing an effective gene expression system, developing a streamlined cell line generation approach, optimizing culture conditions, and defining scaling-up and purification strategies are all critical steps in the production of recombinant proteins, particularly monoclonal antibodies, in mammalian cells. As a result, an active area of study is dedicated to expression and optimizing recombinant protein production. This review explores recent breakthroughs and approaches targeted at accelerating cell line development to attain efficiency and consistency in the synthesis of therapeutic proteins, specifically monoclonal antibodies. The primary goal is to bridge the gap between rising demand and consistent, high-quality mAb production, thereby benefiting the healthcare and pharmaceutical industries.
Graphical Abstract
... Future research should focus on developing methods and frameworks for Explainable AI (XAI) that enable users to understand the decision-making processes of complex neural networks. This is crucial, especially in applications where trust, accountability, and ethical considerations are paramount [8]. ...
Artificial Intelligence (AI) has witnessed a paradigm shift with the integration of Neural Networks (NN) and Machine Learning (ML). This paper explores the synergies between NN and ML, delving into their combined potential to enhance AI capabilities. We investigate the historical context, current state, and future prospects of this integration, addressing key challenges and opportunities. By presenting a comprehensive overview, this paper aims to contribute to the evolving landscape of AI research and development.
... In this case, the DOE can run in silico which will significantly reduce the resource consumption for lab experiments and at the same time accelerate process development. Data driven model can be also integrated with stoichiometric model to understand metabolic shifts during the cell culture [180], and predict amino acid concentration [181], or quality attributes [182]. ...
The capabilities of digital twins (DT) and the successes they have demonstrated in various industries have attracted significant attention from the biopharmaceutical manufacturing field. The realization of DT could help reduce cost, improve productivity, maintain high and constant product quality, and meet Quality-by-Design (QbD) guidelines. Nonetheless, the development of DT is still at its infancy stage and an integrated DT in biopharma is rarely reported. To assist with further advancement in the development of DT, this book chapter gives an overview of the current state of key components involved in a digital twin for biopharmaceutical manufacturing. Currently available software for DT development, process analytical technologies (PATs) for data acquisition and information communication, various modeling strategies to construct the digital replica, and the integration between physical and virtual plants are comprehensively reviewed.
... To the reduce the complexity of the procedure, the use of artificial intelligence (AI) in the analysis of data was introduced and developed by Thoedoratouet et al. 253 An algorithm for prediction of stereochemistry in glycosylation has been developed by Moon et al. 254 Stoichiometric metabolic models have also been developed to predict glycoforms. 255 All these developments have helped in the strategic planning of the oligosaccharide synthesis with greater control and precision. This enzymatic glycosylation coupled with chemical glycosylation 256 has the potential to solve the various unsolved synthetic routes of complex and long oligosaccharide units. ...
... A number of authors have reported the successful implementation of hybrid model approaches to predict glycosylation distribution changes based on alterations in culture parameters and media supplements. [67,[79][80][81][82] Understanding the influential factors via modeling can identify the root cause of batch-to-batch variability, influences of scale-up on afucosylation, or appropriate control strategies. ...
Core fucosylation is a highly prevalent and significant feature of N‐glycosylation in therapeutic monoclonal antibodies produced by mammalian cells where its absence (afucosylation) plays a key role in treatment safety and efficacy. Notably, even slight changes in the level of afucosylation can have a considerable impact on the antibody‐dependent cell‐mediated cytotoxicity. Therefore, implementing control over afucosylation levels is important in upstream manufacturing to maintain consistent quality across batches of product, since standard downstream processing does not change afucosylation. In this review, the influences and strategies to control afucosylation are presented. In particular, there is emphasis on upstream manufacturing culture parameters and media supplementation, as these offer particular advantages as control strategies over alternative approaches such as cell line engineering and chemical inhibitors. The review discusses the relationship between the afucosylation influences and the underlying cellular metabolism to promote increased process understanding. Also, briefly highlighted is the value of empirical and mechanistic models in evaluating and designing control methods for core fucosylation.
... The smaller metabolic network size of microbial model systems such as, for example, E.coli cells, has also led to the development of a variety of solution methodologies and optimisation algorithms for the design of cells with desired phenotype (summarised in [15]). Although the transfer of the entire repertoire of techniques to mammalian cell systems is often hampered by increased model size and complexity leading to highly underdetermined models, there are already developments in the use of mammalian cell GEMs for strain and process engineering (e.g., [16][17][18]), as well as recent algorithm development work applied for understanding the nutritional needs of bioprocessing-relevant organisms but also Atlantic salmon (Salmo salar) [19]. ...
With the plethora of omics data becoming available for mammalian cell and, increasingly, human cell systems, Genome-scale metabolic models (GEMs) have emerged as a useful tool for their organisation and analysis. The systems biology community has developed an array of tools for the solution, interrogation and customisation of GEMs as well as algorithms that enable the design of cells with desired phenotypes based on the multi-omics information contained in these models. However, these tools have largely found application in microbial cells systems, which benefit from smaller model size and ease of experimentation. Herein, we discuss the major outstanding challenges in the use of GEMs as a vehicle for accurately analysing data for mammalian cell systems and transferring methodologies that would enable their use to design strains and processes. We provide insights on the opportunities and limitations of applying GEMs to human cell systems for advancing our understanding of health and disease. We further propose their integration with data-driven tools and their enrichment with cellular functions beyond metabolism, which would, in theory, more accurately describe how resources are allocated intracellularly.
... This alternative is the use of neural networks-based predictive models that could work as a source of data when a reasonable accuracy has been reached. A wide range of these kinds of models can be found in the literature with application in different fields of science [52][53][54][55][56][57][58][59][60][61][62]. ...
The capability analysis of a process against requirements is often an instrument of change. The traditional and fuzzy process capability approaches are the most useful statistical techniques for determining the intrinsic spread of a controlled process for establishing realistic specifications and use for comparative processes. In the industry, the traditional approach is the most commonly used instrument to assess the impact of continuous improvement projects. However, these methods used to evaluate process capability indices could give misleading results because the dataset employed corresponds to the final product/service measures. This paper reviews an alternative procedure to assess the fuzzy process capability indices based on the statistical methodology involved in the modeling and design of experiments. Firstly, a model with reasonable accuracy is developed using a neural network approach. This model is embedded in a graphic user interface (GUI). Using the GUI, an experimental design is carried out, first to know the membership function of the process variability and then include this variability in the model. Again, an experimental design identifies the improved operating conditions for the significative independent variables. A new dataset is generated with these operating conditions, including the minimum error reached for each independent variable. Finally, the GUI is used to get a new prediction for the response variable. The fuzzy process capability indices are determined using the triangular membership function and the predicted response values. The feasibility of the proposed method was validated using a random data set corresponding to the basis weight of a papermaking process. The results indicate that the proposed method provides a better overview of the process performance, showing its true potential. The proposed method can be considered non-invasive.
... In the last few decades, several research efforts have been undertaken to mathematically model N-linked protein glycosylation, with the aim to predict the glycosylation profiles of biotherapeutic products, such as monoclonal antibodies, and to provide insight into the glycosylation machinery itself [186]. Recently, such mathematical models, varying in complexity and modelling approaches, have had appreciable success in the prediction of glycoforms [187][188][189][190], as well as in the successful reconstruction of the secretory pathway [191] in Chinese Hamster Ovary (CHO) cell cultures. In these mathematical models, the prediction of the N-linked glycosylation profiles for a given protein by a specific cell line was based on a given glycosylation reaction network that associated the intracellular glycosylation mechanisms with the final N-glycan structures. ...
The effective treatment of autoimmune disorders can greatly benefit from disease-specific biomarkers that are functionally involved in immune system regulation and can be collected through minimally invasive procedures. In this regard, human serum IgG N-glycans are promising for uncovering disease predisposition and monitoring progression, and for the identification of specific molecular targets for advanced therapies. In particular, the IgG N-glycome in diseased tissues is considered to be disease-dependent; thus, specific glycan structures may be involved in the pathophysiology of autoimmune diseases. This study provides a critical overview of the literature on human IgG N-glycomics, with a focus on the identification of disease-specific glycan alterations. In order to expedite the establishment of clinically-relevant N-glycan biomarkers, the employment of advanced computational tools for the interpretation of clinical data and their relationship with the underlying molecular mechanisms may be critical. Glycoinformatics tools, including artificial intelligence and systems glycobiology approaches, are reviewed for their potential to provide insight into patient stratification and disease etiology. Challenges in the integration of such glycoinformatics approaches in N-glycan biomarker research are critically discussed.
... Insight can be gained from earlier-generation biopharmaceutical processes, such as monoclonal antibody (mAb) development, which have already demonstrated the utility of mathematical modeling in cell and process understanding, process optimization, process design, and scale-up [6,7]. For example, modeling approaches in Chinese hamster ovary (CHO) cells have been recently developed to predict cell culture performance [8] and protein glycosylation [9,10], elucidate the effect of ammonium on the sialylation process [11], predict amino acid concentrations in cell cultures [12], and identify metabolic features and engineering targets for productivity improvement [13]. In addition, various types of models have been implemented for cell culture media formulation or feeding strategy optimization [14][15][16][17], process design and control to achieve a desired product quality profile [18], and physical characterization of bioreactors that contribute to process scale-up and operating conditions optimization across scales [19,20]. ...
With a high required dosing and relatively low process efficiency, providing recombinant adeno-associated virus (rAAV) supply for clinical trials and commercial demand remains challenging. One of the most widely used protocols to produce rAAVs is transient transfection of HEK293 cells. Despite decades of research, there are relatively few studies that focus on key factors in rAAV manufacturing processes. Only about 5–30% of rAAV capsids produced by cells contain the therapeutic element, causing low yields and an immense burden on downstream purification. Modeling approaches are valuable tools for bioprocess understanding and optimization. This review summarizes the recent mechanistic models of HEK293 metabolism and rAAV production. We also provide an outline for future development of hybrid mechanistic and data-driven models for understanding the underlying mechanisms and identifying critically relevant process variables.
... Another study tried to clarify the glycosylation process by training Artificial Neural Networks (ANNs) with the fluxes of the reactions involved in nucleotide sugar donor synthesis, which were calculated by a stoichiometric model of CHO cells, to predict the glycan distribution of the antibodies produced [117]. ...
As plants produce an enormous diversity of metabolites to help them adapt to the environment, the study of plant metabolism is of utmost importance to understand different plant phenotypes. Omics data have been generated at an unprecedented rate for several organisms, including plants, and are widely used to study the central dogma of molecular biology, connecting the genome to phenotypes. Constraint-based modelling (CBM) methods, working over genome-scale metabolic models (GSMMs), have been crucial for organising and analysing omics data by integrating them with biochemical knowledge. In 2009, the first plant GSMM was reconstructed and, since then, several advances have been made, including the creation of context- and multi-tissue models that have supported the study of plant metabolism. Nevertheless, plant metabolic modelling remains very challenging. In parallel, as omics datasets are complex and heterogeneous, machine learning (ML) models have been applied in their interpretation to foster knowledge discovery. Recently, the first studies combining both CBM and ML approaches have emerged and have shown promising results. Here, we present the major advances in plant metabolic modelling and review the main CBM-ML hybrid studies. Finally, we discuss the application of machine learning to address the unique challenges of plant metabolic modelling.
Efficaciously assessing product quality remains time- and resource-intensive. Online Process Analytical Technologies (PATs), encompassing real-time monitoring tools and soft-sensor models, are indispensable for understanding process effects and real-time product quality. This research study evaluated three modeling approaches for predicting CHO cell growth and production, metabolites (extracellular, nucleotide sugar donors (NSD) and glycan profiles): Mechanistic based on first principle Michaelis-Menten kinetics (MMK), data-driven orthogonal partial least square (OPLS) and neural network machine learning (NN). Our experimental design involved galactose-fed batch cultures. MMK excelled in predicting growth and production, demonstrating its reliability in these aspects and reducing the data burden by requiring fewer inputs. However, it was less precise in simulating glycan profiles and intracellular metabolite trends. In contrast, NN and OPLS performed better for predicting precise glycan compositions but displayed shortcomings in accurately predicting growth and production. We utilized time in the training set to address NN and OPLS extrapolation challenges. OPLS and NN models demanded more extensive inputs with similar intracellular metabolite trend prediction. However, there was a significant reduction in time required to develop these two models. The guidance presented here can provide valuable insight into rapid development and application of soft-sensor models with PATs for ipurposes. Therefore, we examined three model typesmproving real-time product CHO therapeutic product quality. Coupled with emerging -omics technologies, NN and OPLS will benefit from massive data availability, and we foresee more robust prediction models that can be advantageous to kinetic or partial-kinetic (hybrid) models.
The production of therapeutic proteins in mammalian cell culture is an essential unit operation in biopharmaceutical manufacture that can benefit from the predictive insights of effective process models, leading to accelerated process development and improved process control. This review outlines and evaluates current approaches to predictive model development for mammalian cell culture and protein production. Classical mechanistic and data driven approaches are analysed, together with potential challenges in model development and application, including the experimental requirements for parameter estimation. Hybrid models, which may offer greater robustness, are then explored along with hybrid model architecture and the steps involved in model development. Successful examples from other cell fermentation processes are also considered, for application to the development, monitoring and control of mammalian processes.
In order to meet the rising demand for biologics and become competitive on the developing biosimilar market, there is a need for process intensification of biomanufacturing processes. Process development of biologics has historically relied on extensive experimentation to develop and optimize biopharmaceutical manufacturing. Experimentation to optimize media formulations, feeding schedules, bioreactor operations and bioreactor scale up is expensive, labor intensive and time consuming. Mathematical modeling frameworks have the potential to enable process intensification while reducing the experimental burden. This review focuses on mathematical modeling of cellular metabolism and N-linked glycosylation as applied to upstream manufacturing of biologics. We review developments in the field of modeling cellular metabolism of mammalian cells using kinetic and stoichiometric modeling frameworks along with their applications to simulate, optimize and improve mechanistic understanding of the process. Interest in modeling N-linked glycosylation has led to the creation of various types of parametric and non-parametric models. Most published studies on mammalian cell metabolism have performed experiments in shake flasks where the pH and dissolved oxygen cannot be controlled. Efforts to understand and model the effect of bioreactor-specific parameters such as pH, dissolved oxygen, temperature, and bioreactor heterogeneity are critically reviewed. Most modeling efforts have focused on the Chinese Hamster Ovary (CHO) cells, which are most commonly used to produce monoclonal antibodies (mAbs). However, these modeling approaches can be generalized and applied to any mammalian cell-based manufacturing platform. Current and potential future applications of these models for Vero cell-based vaccine manufacturing, CAR-T cell therapies, and viral vector manufacturing are also discussed. We offer specific recommendations for improving the applicability of these models to industrially relevant processes.
Early and accurate diagnosis of gastric cancer is vital for effective and targeted treatment. It is known that glycosylation profiles differ in the cancer tissue development process. This study aimed to profile the N-glycans in gastric cancer tissues to predict gastric cancer using machine learning algorithms. The (glyco-) proteins of formalin-fixed parafilm embedded (FFPE) gastric cancer and adjacent control tissues were extracted by chloroform/methanol extraction after the conventional deparaffinization step. The N-glycans were released and labeled with a 2-amino benzoic (2-AA) tag. The MALDI-MS analysis of the 2-AA labeled N-glycans was performed in negative ionization mode, and fifty-nine N-glycan structures were determined. The relative and analyte areas of the detected N-glycans were extracted from the obtained data. Statistical analyses identified significant expression levels of 14 different N-glycans in gastric cancer tissues. The data were separated based on the physical characteristics of N-glycans and used to test in machine-learning models. It was determined that the multilayer perceptron (MLP) was the most appropriate model with the highest sensitivity, specificity, accuracy, Matthews correlation coefficient, and f1 scores for each dataset. The highest accuracy score (96.0 ± 1.3) was obtained from the whole N-glycans relative area dataset, and the AUC value was determined as 0.98. It was concluded that gastric cancer tissues could be distinguished from adjacent control tissues with high accuracy using mass spectrometry-based N-glycomic data.
Multi-strain probiotics are widely regarded as effective products for improving gut microbiota stability and host health, providing advantages over single-strain probiotics. However, in general, it is unclear to what extent different strains would cooperate or compete for resources, and how the establishment of a common biofilm microenvironment could influence their interactions. In this work, we develop an integrative experimental and computational approach to comprehensively assess the metabolic functionality and interactions of probiotics across growth conditions. Our approach combines co-culture assays with genome-scale modelling of metabolism and multivariate data analysis, thus exploiting complementary data- and knowledge-driven systems biology techniques. To show the advantages of the proposed approach, we apply it to the study of the interactions between two widely used probiotic strains of Lactobacillus reuteri and Saccharomyces boulardii, characterising their production potential for compounds that can be beneficial to human health. Our results show that these strains can establish a mixed cooperative-antagonistic interaction best explained by competition for shared resources, with an increased individual exchange but an often decreased net production of amino acids and short-chain fatty acids. Overall, our work provides a strategy that can be used to explore microbial metabolic fingerprints of biotechnological interest, capable of capturing multifaceted equilibria even in simple microbial consortia.
N-linked is the most common type of glycosylation which plays a significant role in identifying various diseases such as type I diabetes and cancer and helps in drug development. Most of the proteins cannot perform their biological and psychological functionalities without undergoing such modification. Therefore, it is essential to identify such sites by computational techniques because of experimental limitations. This study aims to analyze and synthesize the progress to discover N-linked places using machine learning methods. It also explores the performance of currently available tools to predict such sites. Almost seventy research articles published in recognized journals of the N-linked glycosylation field have shortlisted after the rigorous filtering process. The findings of the studies have been reported based on multiple aspects: publication channel, feature set construction method, training algorithm, and performance evaluation. Moreover, a literature survey has developed a taxonomy of N-linked sequence identification. Our study focuses on the performance evaluation criteria, and the importance of N-linked glycosylation motivates us to discover resources that use computational methods instead of the experimental method due to its limitations.
Glycosylation has been shown to define the safety and efficacy of biopharmaceuticals, thus classified as a critical quality attribute. However, controlling glycan heterogeneity has always been a major challenge owing to the multivariate factors that govern the glycosylation process. Conventional approaches for controlling glycosylation such as gene editing and metabolic control have succeeded in obtaining desired glycan profiles in accordance with the Quality by Design paradigm. Nonetheless, the development of smart algorithms and omics‐enabled complete cell characterization has made it possible to predict glycan profiles beforehand, and manipulate process variables accordingly. This review thus discusses the various approaches available for control and prediction of glycosylation in biopharmaceuticals. Further, the futuristic goal of integrating such technologies is discussed in order to attain an automated and digitized continuous bioprocess for control of glycosylation. Given, control of a process as complex as glycosylation requires intense monitoring intervention, we examine the current technologies that enable automation. Finally, we discuss the challenges and the technological gap that currently limits incorporation of an automated process in routine bio‐manufacturing, with a glimpse into the economic bearing.
With recent advances in digitisation and big data analytics, more pharmaceutical firms are adopting digital tools to achieve modernisation. The biological phenomena within bioreactors are a key target for such digital approaches, as these processes are often complicated and difficult to scale. Historically, rules of thumb have been used to match performance indicators across bioreactor scales. Although such methods are well-established and frequently employed by industry, no universal solution has been developed to overcome the many challenges faced in process development and scale-up. Several computer-based methodologies can potentially be applied to bioreactor scale-up, including knowledge driven and data-driven techniques. This review assesses the state of the art in digital advances in scaling bioreactors and the advantages and limitations of scaling techniques. Traditional approaches and their constraints are outlined. The application of knowledge-based techniques is then considered and compared to data-driven models. The ability to transfer processes across bioreactor scales, to compare data and predict process indicators across scales are then examined. Finally, the role of hybrid modelling and digital twins and their potential in bioprocess development are explored.
Due to high demand, monoclonal antibodies (mAbs) production needs to be efficient, as well as maintaining a high product quality. Quality by design (QbD) via predictive process modeling greatly facilitates process understanding and can be used to adjust process parameters to further improve the unit operations. In this work, mechanistic and dynamic kriging models are developed to capture the protein productivity and glycan fractions under different temperatures and pH levels. The design of experiments is used to generate input and output data for model training. The dynamic kriging model shows good performance in capturing the dynamic profiles of cell cultures and glycosylation using only limited input data. The developed model is further used for feasibility analysis, and successfully identifies the operating design space, maintaining high productivity and guaranteed product quality.
The control of nutrient availability is critical to large‐scale manufacturing of biotherapeutics. However, the quantification of proteinogenic amino acids is time‐consuming and thus is difficult to implement for real‐time in situ bioprocess control. Genome‐scale metabolic models describe the metabolic conversion from media nutrients to proliferation and recombinant protein production, and therefore are a promising platform for in silico monitoring and prediction of amino acid concentrations. This potential has not been realized due to unresolved challenges: (1) the models assume an optimal and highly efficient metabolism, and therefore tend to underestimate amino acid consumption, and (2) the models assume a steady state, and therefore have a short forecast range. We address these challenges by integrating machine learning with the metabolic models. Through this we demonstrate accurate and time‐course dependent prediction of individual amino acid concentration in culture medium throughout the production process. Thus, these models can be deployed to control nutrient feeding to avoid premature nutrient depletion or provide early predictions of failed bioreactor runs.
With omics data being generated at an unprecedented rate, genome-scale modelling has become pivotal in its organisation and analysis. However, machine learning methods have been gaining ground in cases where knowledge is insufficient to represent the mechanisms underlying such data or as a means for data curation prior to attempting mechanistic modelling. We discuss the latest advances in genome-scale modelling and the development of optimisation algorithms for network and error reduction, intracellular constraining and applications to strain design. We further review applications of supervised and unsupervised machine learning methods to omics datasets from microbial and mammalian cell systems and present efforts to harness the potential of both modelling approaches through hybrid modelling.
Kinetic models offer incomparable insight on cellular mechanisms controlling protein glycosylation. However, their ability to reproduce site-specific glycoform distributions depends on accurate estimation of a large number of protein-specific kinetic parameters and prior knowledge of enzyme and transport protein levels in the Golgi membrane. Herein we propose an artificial neural network (ANN) for protein glycosylation and apply this to four recombinant glycoproteins produced in Chinese hamster ovary (CHO) cells, two monoclonal antibodies and two fusion proteins. We demonstrate that the ANN model accurately predicts site-specific glycoform distributions of up to eighteen glycan species with an average absolute error of 1.1%, correctly reproducing the effect of metabolic perturbations as part of a hybrid, kinetic/ANN, glycosylation model (HyGlycoM), manganese supplementation and glycosyltransferase knock out experiments as a stand-alone machine learning algorithm. These results showcase the potential of machine learning and hybrid approaches for rapidly developing performance-driven models of protein glycosylation.
Multivariate latent variable methods have become a popular and versatile toolset to analyze bioprocess data in industry and academia. This work spans such applications from the evaluation of the role of the standard process variables and metabolites to the metabolomics level, that is, to the extensive number metabolic compounds detectable in the extracellular and intracellular domains. Given the substantial effort currently required for the measurement of the latter groups, a tailored methodology is presented that is capable of providing valuable process insights as well as predicting the glycosylation profile based on only four experiments measured over 12 cell culture days. An important result of the work is the possibility to accurately predict many of the glycan variables based on the information of three experiments. An additional finding is that such predictive models can be generated from the more accessible process and extracellular information only, that is, without including the more experimentally cumbersome intracellular data. With regards to the incorporation of omics data in the standard process analytics framework in the future, this works provides a comprehensive data analysis pathway which can efficiently support numerous bioprocessing tasks.
Systems biology has long been interested in models capturing both metabolism and expression in a cell. We propose here an implementation of the metabolism and expression model formalism (ME-models), which we call ETFL, for Expression and Thermodynamics Flux models. ETFL is a hierarchical model formulation, from metabolism to RNA synthesis, that allows simulating thermodynamics-compliant intracellular fluxes as well as enzyme and mRNA concentration levels. ETFL formulates a mixed-integer linear problem (MILP) that enables both relative and absolute metabolite, protein, and mRNA concentration integration. ETFL is compatible with standard MILP solvers and does not require a non-linear solver, unlike the previous state of the art. It also accounts for growth-dependent parameters, such as relative protein or mRNA content. We present ETFL along with its validation using results obtained from a well-characterized E. coli model. We show that ETFL is able to reproduce proteome-limited growth. We also subject it to several analyses, including the prediction of feasible mRNA and enzyme concentrations and gene essentiality. Accounting for the effects of genetic expression in genome-scale metabolic models is challenging. Here, the authors introduce a model formulation that efficiently simulates thermodynamic-compliant fluxes, enzyme and mRNA concentration levels, allowing omics integration and broad analysis of in silico cellular physiology.
In mammalian cells, >25% of synthesized proteins are exported through the secretory pathway. The pathway complexity, however, obfuscates its impact on the secretion of different proteins. Unraveling its impact on diverse proteins is particularly important for biopharmaceutical production. Here we delineate the core secretory pathway functions and integrate them with genome-scale metabolic reconstructions of human, mouse, and Chinese hamster ovary cells. The resulting reconstructions enable the computation of energetic costs and machinery demands of each secreted protein. By integrating additional omics data, we find that highly secretory cells have adapted to reduce expression and secretion of other expensive host cell proteins. Furthermore, we predict metabolic costs and maximum productivities of biotherapeutic proteins and identify protein features that most significantly impact protein secretion. Finally, the model successfully predicts the increase in secretion of a monoclonal antibody after silencing a highly expressed selection marker. This work represents a knowledgebase of the mammalian secretory pathway that serves as a novel tool for systems biotechnology. The secretory pathway is used in the production of most biopharmaceuticals, but the associated biosynthetic costs are little understood. Here, the authors integrate the core secretory pathway into genome-scale metabolic models of human, mouse, and CHO cells, enabling in silico analysis.
Animal cell culture processes have become the standard platform to produce therapeutic proteins such as recombinant monoclonal antibodies (mAb). Since the mAb quality could be subject to significant changes depending on manufacturing process conditions, real time monitoring and control systems are required to ensure mAb specifications mainly glycosylation and patient safety. Up to now, real time monitoring glycosylation of proteins has received scarce attention. In this article, the use of near infrared (NIR) to monitor mAb glycosylation has been reported for the first time. Whereas monitoring models are mainly constructed using linear partial least squares regressions (PLSR), evidences presented in this study indicate nonlinearity relationship between in situ captured spectra and compound concentrations, compromising the PLSR performances. A novel and simple approach was proposed to fit nonlinearity using the locally weighted regression (LWR). The LWR models were found to be more appropriate for handling information contained in spectra so that real time monitoring of cultures were accurately performed. Moreover, for the first time, the LWR calibration models allowed mAb glycosylation to be monitored, in a real time manner, by using in situ NIR spectroscopy. These results represent a further step toward developing active‐control feedback of animal cell processes, particularly for ensuring properties of biologics.
Exerting control over the glycan moieties of antibody therapeutics is highly desirable from a product safety and batch‐to‐batch consistency perspective. Strategies to improve antibody productivity may compromise quality, while interventions for improving glycoform distribution can adversely affect cell growth and productivity. Process design therefore needs to consider the trade‐off between preserving cellular health and productivity while enhancing antibody quality. In this work, we present a modeling platform that quantifies the impact of glycosylation precursor feeding – specifically that of galactose and uridine – on cellular growth, metabolism as well as antibody productivity and glycoform distribution. The platform has been parameterized using an initial training data set yielding an accuracy of ±5% with respect to glycoform distribution. It was then used to design an optimized feeding strategy that enhances the final concentration of galactosylated antibody in the supernatant by over 90% compared with the control without compromising the integral of viable cell density or final antibody titer. This work supports the implementation of Quality by Design towards higher‐performing bioprocesses.
Kinetic modeling is the most suitable framework to describe the dynamic behavior of mammalian cell culture although its industrial application is still in its infancy. Herein, we review mammalian bioprocess relevant kinetic models and found that the simple unstructured-unsegregated approach utilizing empirical Monod type kinetics based on limiting substrates and inhibitory metabolites is commonly used due to the traceability and simple formalism. Notably, the available kinetic models are typically small to moderate in size and the development of large-scale models is severely hampered by the scarcity of kinetic data and limitations in current parameter estimation methods. The recent availability of abundant high-throughput multi-omics datasets from mammalian cell culture have now paved ways to improve parameterization of kinetic models, and integrate regulatory, signaling, and product quality related intracellular events, as well as cellular metabolism within the modeling framework. Ultimately, we foresee that multi-scale modeling is the way forward in building predictive kinetic models of mammalian cell culture to advance biomaufacturing.
Glycosylation greatly influences the safety and efficacy of many of the highest-selling recombinant therapeutic proteins (rTPs). In order to define optimal cell culture feeding strategies that control rTP glycosylation, it is necessary to know how nucleotide sugars (NSs) are consumed towards host cell and rTP glycosylation. Here, we present a theoretical framework that integrates the reported glycoproteome of CHO cells, the number of N-linked and O-GalNAc glycosylation sites on individual host cell proteins (HCPs), and the carbohydrate content of CHO glycosphingolipids to estimate the demand of NSs towards CHO cell glycosylation. We have identified the most abundant N-linked and O-GalNAc CHO glycoproteins, obtained the weighted frequency of N-linked and O-GalNAc glycosites across the CHO cell proteome, and have derived stoichiometric coefficients for NS consumption towards CHO cell glycosylation. By combining the obtained stoichiometric coefficients with previously reported data for specific growth and productivity of CHO cells, we observe that the demand of NSs towards glycosylation is significant and, thus, is required to better understand the burden of glycosylation on cellular metabolism. The estimated demand of NSs towards CHO cell glycosylation can be used to rationally design feeding strategies that ensure optimal and consistent rTP glycosylation.
Background:
Several methods have been developed to predict system-wide and condition-specific intracellular metabolic fluxes by integrating transcriptomic data with genome-scale metabolic models. While powerful in many settings, existing methods have several shortcomings, and it is unclear which method has the best accuracy in general because of limited validation against experimentally measured intracellular fluxes.
Results:
We present a general optimization strategy for inferring intracellular metabolic flux distributions from transcriptomic data coupled with genome-scale metabolic reconstructions. It consists of two different template models called DC (determined carbon source model) and AC (all possible carbon sources model) and two different new methods called E-Flux2 (E-Flux method combined with minimization of l2 norm) and SPOT (Simplified Pearson cOrrelation with Transcriptomic data), which can be chosen and combined depending on the availability of knowledge on carbon source or objective function. This enables us to simulate a broad range of experimental conditions. We examined E. coli and S. cerevisiae as representative prokaryotic and eukaryotic microorganisms respectively. The predictive accuracy of our algorithm was validated by calculating the uncentered Pearson correlation between predicted fluxes and measured fluxes. To this end, we compiled 20 experimental conditions (11 in E. coli and 9 in S. cerevisiae), of transcriptome measurements coupled with corresponding central carbon metabolism intracellular flux measurements determined by 13C metabolic flux analysis (13C-MFA), which is the largest dataset assembled to date for the purpose of validating inference methods for predicting intracellular fluxes. In both organisms, our method achieves an average correlation coefficient ranging from 0.59 to 0.87, outperforming a representative sample of competing methods. Easy-to-use implementations of E-Flux2 and SPOT are available as part of the open-source package MOST (http://most.ccib.rutgers.edu/).
Conclusion:
Our method represents a significant advance over existing methods for inferring intracellular metabolic flux from transcriptomic data. It not only achieves higher accuracy, but it also combines into a single method a number of other desirable characteristics including applicability to a wide range of experimental conditions, production of a unique solution, fast running time, and the availability of a user-friendly implementation.
NADPH regeneration capacity is attracting growing research attention due to its important role in resisting oxidative stress. Besides, NADPH availability has been regarded as a limiting factor in production of industrially valuable compounds. The central carbon metabolism carries the carbon skeleton flux supporting the operation of NADPH-regenerating enzyme and offers flexibility in coping with NADPH demand for varied intracellular environment. To acquire an insightful understanding of its NADPH regeneration capacity, the elementary mode method was employed to compute all elementary flux modes (EFMs) of a network representative of central carbon metabolism. Based on the metabolic flux distributions of these modes, a cluster analysis of EFMs with high NADPH regeneration rate was conducted using the self-organizing map clustering algorithm. The clustering results were used to study the relationship between the flux of total NADPH regeneration and the flux in each NADPH producing enzyme. The results identified several reaction combinations supporting high NADPH regeneration, which are proven to be feasible in cells via thermodynamic analysis and coincident with a great deal of previous experimental report. Meanwhile, the reaction combinations showed some common characteristics: there were one or two decarboxylation oxidation reactions in the combinations that produced NADPH and the combination constitution included certain gluconeogenesis pathways. These findings suggested cyclization pathways as a powerful way for NADPH regeneration capacity of bacterial central carbon metabolism.
Metabolic flux analysis (MFA) is widely used to estimate intracellular fluxes. Conventional MFA, however, is limited to continuous cultures and the mid-exponential growth phase of batch cultures. Dynamic MFA (DMFA) has emerged to characterize time-resolved metabolic fluxes for the entire culture period. Here, the linear DMFA approach was extended using B-spline fitting (B-DMFA) to estimate mass balanced fluxes. Smoother fits were achieved using reduced number of knots and parameters. Additionally, computation time was greatly reduced using a new heuristic algorithm for knot placement. B-DMFA revealed that Chinese hamster ovary cells shifted from 37°C to 32°C maintained a constant IgG volume-specific productivity, whereas the productivity for the controls peaked during mid-exponential growth phase and declined afterward. The observed 42% increase in product titer at 32°C was explained by a prolonged cell growth with high cell viability, a larger cell volume and a more stable volume-specific productivity.
Matthew Croughan first author articles in Biotechnology & Bioengineering, 1987 – March 2015
Free access articles available to everyone for free (via link). Others depend upon subscription status.
Croughan, M. S., Hamel, J.-F. and Wang, D. I. C. (1987), Hydrodynamic effects on animal cells grown in microcarrier cultures. Biotechnol. Bioeng., 29: 130–141. doi: 10.1002/bit.260290117
http://onlinelibrary.wiley.com/doi/10.1002/bit.260290117/abstract
Croughan, M. S., Hamel, J.-F. P. and Wang, D. I. C. (1988), Effects of microcarrier concentration in animal cell culture. Biotechnol. Bioeng., 32: 975–982. doi: 10.1002/bit.260320805
http://onlinelibrary.wiley.com/doi/10.1002/bit.260320805/abstract
Croughan, M. S. and Wang, D. I. C. (1989), Growth and death in overagitated microcarrier cell cultures. Biotechnol. Bioeng., 33: 731–744. doi: 10.1002/bit.260330611
http://onlinelibrary.wiley.com/doi/10.1002/bit.260330611/abstract
Croughan, M. S., Sayre, E. S. and Wang, D. I. C. (1989), Viscous reduction of turbulent damage in animal cell culture. Biotechnol. Bioeng., 33: 862–872. doi: 10.1002/bit.260330710
http://onlinelibrary.wiley.com/doi/10.1002/bit.260330710/abstract
FREE Access:
Croughan, M. S. and Wang, D. I. C. (1990), Reversible removal and hydrodynamic phenomena in CHO microcarrier cultures. Biotechnol. Bioeng., 36: 316–319. doi: 10.1002/bit.260360314
http://onlinelibrary.wiley.com/doi/10.1002/bit.260360314/abstract
Croughan, M. S., Hamel, J.-F. and Wang, D. I. C. (2000), Hydrodynamic effects on animal cells grown in microcarrier cultures. Biotechnol. Bioeng., 67: 841–852. doi: 10.1002/(SICI)1097-0290(20000320)67:6<841::AID-BIT19>3.0.CO;2-K
http://onlinelibrary.wiley.com/doi/10.1002/(SICI)1097-0290(20000320)67:6%3C841::AID-BIT19%3E3.0.CO;2-K/abstract
Croughan, M. S. and Hu, W.-S. (2006), From microcarriers to hydrodynamics: Introducing engineering science into animal cell culture. Biotechnol. Bioeng., 95: 220–225. doi: 10.1002/bit.21088
http://onlinelibrary.wiley.com/doi/10.1002/bit.21088/abstract
FREE Access:
Croughan, M. S., Konstantinov, K. B. and Cooney, C. (2015), The future of industrial bioprocessing: Batch or continuous?. Biotechnol. Bioeng., 112: 648–651. doi: 10.1002/bit.25529
http://onlinelibrary.wiley.com/doi/10.1002/bit.25529/abstract
COnstraint-Based Reconstruction and Analysis (COBRA) methods are widely used for genome-scale modeling of metabolic networks in both prokaryotes and eukaryotes. Due to the successes with metabolism, there is an increasing effort to apply COBRA methods to reconstruct and analyze integrated models of cellular processes. The COBRA Toolbox for MATLAB is a leading software package for genome-scale analysis of metabolism; however, it was not designed to elegantly capture the complexity inherent in integrated biological networks and lacks an integration framework for the multiomics data used in systems biology. The openCOBRA Project is a community effort to promote constraints-based research through the distribution of freely available software.
Here, we describe COBRA for Python (COBRApy), a Python package that provides support for basic COBRA methods. COBRApy is designed in an object-oriented fashion that facilitates the representation of the complex biological processes of metabolism and gene expression. COBRApy does not require MATLAB to function; however, it includes an interface to the COBRA Toolbox for MATLAB to facilitate use of legacy codes. For improved performance, COBRApy includes parallel processing support for computationally intensive processes.
COBRApy is an object-oriented framework designed to meet the computational challenges associated with the next generation of stoichiometric constraint-based models and high-density omics data sets.
http://opencobra.sourceforge.net/
After hundreds of generations of adaptive evolution at exponential growth, Escherichia coli grows as predicted using flux balance analysis (FBA) on genome-scale metabolic models (GEMs). However, it is not known whether the predicted pathway usage in FBA solutions is consistent with gene and protein expression in the wild-type and evolved strains. Here, we report that >98% of active reactions from FBA optimal growth solutions are supported by transcriptomic and proteomic data. Moreover, when E. coli adapts to growth rate selective pressure, the evolved strains upregulate genes within the optimal growth predictions, and downregulate genes outside of the optimal growth solutions. In addition, bottlenecks from dosage limitations of computationally predicted essential genes are overcome in the evolved strains. We also identify regulatory processes that may contribute to the development of the optimal growth phenotype in the evolved strains, such as the downregulation of known regulons and stringent response suppression. Thus, differential gene and protein expression from wild-type and adaptively evolved strains supports observed growth phenotype changes, and is consistent with GEM-computed optimal growth states. Molecular Systems Biology 6: 390; published online 27 July 2010; doi:10.1038/msb.2010.47 Subject Categories: functional genomics; simulation and data analysis
Advances in the field of bioinformatics have led to reconstruction of genome-scale networks for a number of key organisms. The application of physicochemical constraints to these stoichiometric networks allows researchers, through methods such as flux balance analysis, to highlight key sets of reactions necessary to achieve particular objectives. The key benefits of constraint-based analysis lie in the minimal knowledge required to infer systemic properties. However, network degeneracy leads to a large number of flux distributions that satisfy any objective; moreover, these distributions may be dominated by biologically irrelevant internal cycles. By examining the geometry underlying the problem, we define two methods for finding a unique solution within the space of all possible flux distributions; such a solution contains no internal cycles, and is representative of the space as a whole. The first method draws on typical geometric knowledge, but cannot be applied to large networks because of the high computational complexity of the problem. Thus a second method, an iteration of linear programs which scales easily to the genome scale, is defined. The algorithm is run on four recent genome-scale models, and unique flux solutions are found. The algorithm set out here will allow researchers in flux balance analysis to exchange typical solutions to their models in a reproducible format. Moreover, having found a single solution, statistical analyses such as correlations may be performed.
Genome-scale metabolic network reconstructions in microorganisms have been formulated and studied for about 8 years. The constraint-based
approach has shown great promise in analyzing the systemic properties of these network reconstructions. Notably, constraint-based
models have been used successfully to predict the phenotypic effects of knock-outs and for metabolic engineering. The inherent
uncertainty in both parameters and variables of large-scale models is significant and is well suited to study by Monte Carlo
sampling of the solution space. These techniques have been applied extensively to the reaction rate (flux) space of networks,
with more recent work focusing on dynamic/kinetic properties. Monte Carlo sampling as an analysis tool has many advantages,
including the ability to work with missing data, the ability to apply post-processing techniques, and the ability to quantify
uncertainty and to optimize experiments to reduce uncertainty. We present an overview of this emerging area of research in
systems biology.
Chinese hamster ovary (CHO) cells are most prevalently used for producing recombinant therapeutics in biomanufacturing. Recently, more rational and systems approaches have been increasingly exploited to identify key metabolic bottlenecks and engineering targets for cell line engineering and process development based on the CHO genome-scale metabolic model which mechanistically characterizes cell culture behaviours. However, it is still challenging to quantify plausible intracellular fluxes and discern metabolic pathway usages considering various clonal traits and bioprocessing conditions. Thus, we newly incorporated enzyme kinetic information into the updated CHO genome-scale model (iCHO2291) and added enzyme capacity constraints within the flux balance analysis framework (ecFBA) to significantly reduce the flux variability in biologically meaningful manner, as such improving the accuracy of intracellular flux prediction. Interestingly, ecFBA could capture the overflow metabolism under the glucose excess condition where the usage of oxidative phosphorylation is limited by the enzyme capacity. In addition, its applicability was successfully demonstrated via a case study where the clone- and media-specific lactate metabolism was deciphered, suggesting that the lactate-pyruvate cycling could be beneficial for CHO cells to efficiently utilize the mitochondrial redox capacity. In summary, iCHO2296 with ecFBA can be used to confidently elucidate cell cultures and effectively identify key engineering targets, thus guiding bioprocess optimization and cell engineering efforts as a part of digital twin model for advanced biomanufacturing in future.
Constraint based modelling methods, such as Flux Balance Analysis (FBA), have been extensively used to decipher complex, information rich ‐omics datasets in order to elicit system‐wide behavioral patterns of cellular metabolism. FBA has been successfully used to gain insight in a wide range of applications, such as range of substrate utilization, product yields and to design metabolic engineering strategies to improve bioprocess performance. A well‐known challenge associated with large genome‐scale metabolic networks (GEMs) is that they result in underdetermined problem formulations. Consequently, rather than unique solutions, FBA and related methods examine ranges of reaction flux values that are consistent with the studied physiological conditions. The wider the reported flux ranges, the higher the uncertainty in the determination of basic reaction properties, limiting interpretability of and confidence in the results. Herein we propose a new, computationally efficient approach that refines flux range predictions by constraining reaction fluxes based on the elemental balance of carbon. We compared carbon constraint FBA (ccFBA) against experimentally measured intracellular fluxes using the latest CHO GEM (iCHO1766) and were able to substantially improve the accuracy of predicted flux values compared to FBA. ccFBA can be used as a stand‐alone method but is also compatible with and complimentary to other constraint‐based approaches.
This article is protected by copyright. All rights reserved.
The manufacture of protein-based therapeutics presents unique challenges due to limited control over the biotic phase. This typically gives rise to a wide range of protein structures of varying safety and in vivo efficacy. Herein we propose a computational methodology, enabled by the application of constrained Global Sensitivity Analysis, for efficiently exploring the operating range of process inputs in silico and identifying a design space that meets output constraints. The methodology was applied to an antibody-producing Chinese hamster ovary (CHO) cell culture system: we explored >8000 feeding strategies to identify a subset of manufacturing conditions that meet constraints on antibody titre and glycan distribution as an attribute of product quality. Our computational findings were then verified experimentally, confirming the applicability of this approach to a challenging production system. We envisage that this methodology can significantly expedite bioprocess development and increase operational flexibility.
Cell culture media (CCM) used in industrial biotechnology are complex mixtures of molecules and elements that are inherently difficult to analyze comprehensively. CCM quality analysis is of utmost importance for efficient production of protein-based therapeutics with the correct Critical Quality Attributes (CQAs). Here we discuss the use of rapid spectroscopic methods for routine screening of CCM molecular variance which include electronic (UV–vis absorption and fluorescence) and vibrational (Raman, FT-IR, and Near-Infra-red) spectroscopies. CCM analysis needs to provide: identity testing, compositional variance analysis (i.e. lot-to-lot variation), validation of media preparation protocols, and correlations with Critical Process Parameters (CPPs) or product CQAs. Rapid spectroscopic methods can fulfil some of these requirements but only with correct sample handling and preparation. Accurate analysis requires the use of chemometrics combined with a detailed knowledge of sample behavior such as water absorption and chemical stability.
The glycosylation of therapeutic monoclonal antibodies (mAbs), a known critical quality attribute, is often greatly modified during the production process by animal cells. It is essential for biopharmaceutical industries to monitor and control this glycosylation. However, current glycosylation characterization techniques involve time- and labor-intensive analyses, often carried out at the end of the culture when the product is already synthesized. This study proposes a novel methodology for real-time monitoring of antibody glycosylation site occupancy using Raman spectroscopy. It was first observed in CHO cell batch culture that when low nutrient concentrations were reached, a decrease in mAb glycosylation was induced, which made it essential to rapidly detect this loss of product quality. By combining in situ Raman spectroscopy with chemometric tools, efficient prediction models were then developed for both glycosylated and non-glycosylated mAbs. By comparing Variable Importance in Projection profiles of the prediction models, it was confirmed that Raman spectroscopy is a powerful method to distinguish extremely similar molecules, despite the high complexity of the culture medium. Finally, the Raman prediction models were used to monitor batch and feed-harvest cultures in situ. For the first time, it was demonstrated that the concentrations of glycosylated and non-glycosylated mAbs could be successfully and simultaneously estimated in real-time with high accuracy, including their sudden variations due to medium exchanges. Raman spectroscopy can thus be considered as a promising PAT tool for feedback process control dedicated to on-line optimization of mAb quality. This article is protected by copyright. All rights reserved.
The development of cell culture processes is highly complex and requires a large number of experiments on various scales to define the design space of the final process and fulfil the regulatory requirements. This work follows an almost complete process development cycle for a biosimilar monoclonal antibody, from high throughput screening and optimization to scale-up and process validation. The key goal of this analysis is to apply tailored multivariate tools to support decision-making at every stage of process development. A toolset mainly based on Principal Component Analysis, Decision Trees and Partial Least Square Regression combined with a Genetic Algorithm is presented. It enables to visualize the sequential improvement of the high-dimensional quality profile towards the target, provides a solid basis for the selection of effective process variables and allows to dynamically predict the complete set of product quality attributes. Additionally, this work shows the deep level of process knowledge which can be deduced from small scale experiments through such multivariate tools. The presented methodologies are generally applicable across various processes and substantially reduce the complexity, experimental effort as well as the costs and time of process development.
The application of PAT for in-line monitoring of biopharmaceutical manufacturing operations has a central role in developing more robust and consistent processes. Various spectroscopic techniques have been applied for collecting real-time data from cell culture processes. Among these, Raman spectroscopy has been shown to have advantages over other spectroscopic techniques, especially in aqueous culture solutions. Measurements of several process parameters such as glucose, lactate, glutamine, glutamate, ammonium, osmolality and VCD using Raman-based chemometrics models have been reported in literature. The application of Raman spectroscopy, coupled with calibration models for amino acid measurement in cell cultures, has been assessed. The developed models cover four amino acids important for cell growth and production: tyrosine, tryptophan, phenylalanine and methionine. The chemometrics models based on Raman spectroscopy data demonstrate the significant potential for the quantification of tyrosine, tryptophan and phenylalanine. The model for methionine would have to be further refined to improve quantification.
Highlights
•A consensus genome-scale model of CHO cell metabolism was constructed
•Model has 1,766 genes and 6,663 reactions describing metabolism and protein production
•Integration of omic data reveals amino acid auxotrophy bases in various cell lines
•Cell engineering can yield more efficient resource allocation than bioprocess methods
Summary
Chinese hamster ovary (CHO) cells dominate biotherapeutic protein production and are widely used in mammalian cell line engineering research. To elucidate metabolic bottlenecks in protein production and to guide cell engineering and bioprocess optimization, we reconstructed the metabolic pathways in CHO and associated them with >1,700 genes in the Cricetulus griseus genome. The genome-scale metabolic model based on this reconstruction, iCHO1766, and cell-line-specific models for CHO-K1, CHO-S, and CHO-DG44 cells provide the biochemical basis of growth and recombinant protein production. The models accurately predict growth phenotypes and known auxotrophies in CHO cells. With the models, we quantify the protein synthesis capacity of CHO cells and demonstrate that common bioprocess treatments, such as histone deacetylase inhibitors, inefficiently increase product yield. However, our simulations show that the metabolic resources in CHO are more than three times more efficiently utilized for growth or recombinant protein synthesis following targeted efforts to engineer the CHO secretory pathway. This model will further accelerate CHO cell engineering and help optimize bioprocesses.
Objectives
Infliximab, trastuzumab and bevacizumab are among the most frequently prescribed therapeutic proteins, and like most other therapeutic proteins, are glycosylated. As differences in glycosylation may significantly change the safety and efficacy of therapeutic glycoproteins, it is extremely important to control N-glycosylation consistency. In the first part of this study, the batch-to-batch consistency of the N-glycosylation of infliximab, trastuzumab and bevacizumab was analysed. In the second part, the consistency of the N-glycosylation of bevacizumab stored in polycarbonate syringes (for off-label drug use) for 3 months was examined.
Methods
N-glycans were (i) enzymatically released using peptide-N-glycosidase F, (ii) reduced, and (iii) analysed using hydrophilic interaction liquid chromatography coupled with high-resolution mass spectrometry. Mass spectrometry data were interpreted using principal component analysis combined with two-way analysis of variance and Tukey post hoc tests. The biological activity of infliximab and trastuzumab was examined using enzyme-linked immunosorbent assays.
Results
The results of both studies make important contributions to the field of hospital pharmacy. All batches of the studied therapeutic glycoproteins (infliximab, trastuzumab and bevacizumab) varied considerably (especially in galactosylation), while the N-glycosylation of bevacizumab remained unchanged during 3-month storage.
Conclusions
Threshold values for batch-to-batch N-glycosylation variations should be established and batch-to-batch glycosylation consistency should be regularly tested. In our study, samples with significantly different N-glycosylation profiles showed no significant variations in biological activity, suggesting that the differences are probably not therapeutically significant.
A methodology to calculate the required amount of amino acids (a.a.) and glucose in feeds for animal cell culture from monitoring their levels in batch experiments is presented herein. Experiments with the designed feeds on an antibody-producing Chinese hamster ovary cell line resulted in a 3-fold increase in titer compared to batch culture. Adding 40% more nutrients to the same feed further increases the yield to 3.5 higher than in batch culture. Our results show that above a certain threshold there is no linear correlation between nutrient addition and the integral of viable cell concentration. In addition, although high ammonia levels hinder cell growth, they do not appear to affect specific antibody productivity, while we hypothesize that high extracellular lactate concentration is the cause for the metabolic shift towards lactate consumption for the cell line used. Overall, the performance of the designed feeds is comparable to that of a commercial feed that was tested in parallel. Expanding this approach to more nutrients, as well as changing the ratio of certain amino acids as informed by flux balance analysis, could achieve even higher yields. Biotechnol. Bioeng. © 2014 Wiley Periodicals, Inc.
One of today's key challenges is the ability to decode the functions of complex carbo- hydrates in various biological contexts. To rapidly generate high-quality glycomics data in a high-throughput fashion, we developed a robotized and low-cost N-glycan analysis platform for glycoprofiling of immunoglobulin G antibodies, which are central players of the immune system and of vital importance in the biopharmaceutical industry. The key features include (a) rapid IgG afinity purification and sample concentration, (b) protein denaturation and glycan release on a multiwell filtration device (c) glycan purification on solid-supported hydrazide and (d) glycan quantification by ultra-performance liquid chromatography. The sample preparation workflow was automated using a robotic liquid handling workstation, allowing the preparation of 96 samples (or multiples thereof) in 22 hours with excellent reproducibility and thus should greatly facilitate biomarker discovery and glycosylation monitoring of therapeutic IgGs.
Chinese hamster ovary (CHO) cells are preferred hosts for the production of recombinant biopharmaceuticals. Efforts to optimize these bioprocesses have largely relied on empirical experience and our knowledge of cellular behavior in culture is highly incomplete. More recently, comprehensive investigations of metabolic network operation have started to be used to uncover traits associated with optimal growth and recombinant protein production. In this work, we used (1) H-nuclear magnetic resonance ((1) H-NMR) to analyze the supernatants of glutamine-synthetase (GS)-CHO cell clones expressing variable amounts of an IgG4 under control and butyrate-treated conditions. Exometabolomic data revealed accumulation of several metabolic by-products, indicating inefficiencies at different metabolic nodes. The data were contextualized in a detailed network of 117 reactions and the cellular fluxomes estimated through metabolic flux analysis (MFA). This approach allowed comparing metabolic activity across different clones, growth phases and culture conditions, in particular the efficiency pertaining to carbon lost to glycerol and lactate accumulation and the characteristic nitrogen metabolism involving high asparagine and serine uptake rates. Importantly, this study shows that early butyrate treatment has a marked effect on sustaining high nutrient consumption along culture time, being more pronounced during the stationary phase when extra energy generation and biosynthetic activity is fueled to increase IgG formation. Collectively, the information generated contributes to deepening our understanding of mammalian CHO cells metabolism in culture, facilitating future design of improved bioprocesses. Biotechnol. Bioeng. © 2013 Wiley Periodicals, Inc.
Monoclonal antibodies (mAbs) are one of the most important products of the biopharmaceutical industry. Their therapeutic efficacy depends on the post-translational process of glycosylation, which is influenced by manufacturing process conditions. Herein, we present a dynamic mathematical model for mAb glycosylation that considers cisternal maturation by approximating the Golgi apparatus to a plug flow reactor and by including recycling of Golgi-resident proteins (glycosylation enzymes and transport proteins [TPs]). The glycosylation reaction rate expressions were derived based on the reported kinetic mechanisms for each enzyme, and transport of nucleotide sugar donors [NSDs] from the cytosol to the Golgi lumen was modeled to serve as a link between glycosylation and cellular metabolism. Optimization-based methodologies were developed for estimating unknown enzyme and TP concentration profile parameters. The resulting model is capable of reproducing glycosylation profiles of commercial mAbs. It can further reproduce the effect gene silencing of the FucT glycosylation enzyme and cytosolic NSD depletion have on the mAb oligosaccharide profile. All novel elements of our model are based on biological evidence and generate more accurate results than previous reports. We therefore believe that the improvements contribute to a more detailed representation of the N-linked glycosylation process. The overall results show the potential of our model toward evaluating cell engineering strategies that yield desired glycosylation profiles. Additionally, when coupled to cellular metabolism, this model could be used to assess the effect of process conditions on glycosylation and aid in the design, control, and optimization of biopharmaceutical manufacturing processes.
Chinese hamster ovary (CHO) cells are the main platform for production of biotherapeutics in the biopharmaceutical industry. However, relatively little is known about the metabolism of CHO cells in cell culture. In this work, metabolism of CHO cells was studied at the growth phase and early stationary phase using isotopic tracers and mass spectrometry. CHO cells were grown in fed-batch culture over a period of six days. On days 2 and 4, [1,2-(13)C] glucose was introduced and the labeling of intracellular metabolites was measured by gas chromatography-mass spectrometry (GC-MS) at 6, 12 and 24h following the introduction of tracer. Intracellular metabolic fluxes were quantified from measured extracellular rates and (13)C-labeling dynamics of intracellular metabolites using non-stationary (13)C-metabolic flux analysis ((13)C-MFA). The flux results revealed significant rewiring of intracellular metabolic fluxes in the transition from growth to non-growth, including changes in energy metabolism, redox metabolism, oxidative pentose phosphate pathway and anaplerosis. At the exponential phase, CHO cell metabolism was characterized by a high flux of glycolysis from glucose to lactate, anaplerosis from pyruvate to oxaloacetate and from glutamate to α-ketoglutarate, and cataplerosis though malic enzyme. At the stationary phase, the flux map was characterized by a reduced flux of glycolysis, net lactate uptake, oxidative pentose phosphate pathway flux, and reduced rate of anaplerosis. The fluxes of pyruvate dehydrogenase and TCA cycle were similar at the exponential and stationary phase. The results presented here provide a solid foundation for future studies of CHO cell metabolism for applications such as cell line development and medium optimization for high-titer production of recombinant proteins.
Constraint-based modeling is an approach for quantitative prediction of net reaction flux in genome-scale biochemical networks. In vivo, the second law of thermodynamics requires that net macroscopic flux be forward, when the transformed reaction Gibbs energy is negative. We calculate the latter by using (i) group contribution estimates of metabolite species Gibbs energy, combined with (ii) experimentally measured equilibrium constants. In an application to a genome-scale stoichiometric model of Escherichia coli metabolism, iAF1260, we demonstrate that quantitative prediction of reaction directionality is increased in scope and accuracy by integration of both data sources, transformed appropriately to in vivo pH, temperature and ionic strength. Comparison of quantitative versus qualitative assignment of reaction directionality in iAF1260, assuming an accommodating reactant concentration range of 0.02-20mM, revealed that quantitative assignment leads to a low false positive, but high false negative, prediction of effectively irreversible reactions. The latter is partly due to the uncertainty associated with group contribution estimates. We also uncovered evidence that the high intracellular concentration of glutamate in E. coli may be essential to direct otherwise thermodynamically unfavorable essential reactions, such as the leucine transaminase reaction, in an anabolic direction.
A new form of metabolic flux analysis (MFA) called thermodynamics-based metabolic flux analysis (TMFA) is introduced with the capability of generating thermodynamically feasible flux and metabolite activity profiles on a genome scale. TMFA involves the use of a set of linear thermodynamic constraints in addition to the mass balance constraints typically used in MFA. TMFA produces flux distributions that do not contain any thermodynamically infeasible reactions or pathways, and it provides information about the free energy change of reactions and the range of metabolite activities in addition to reaction fluxes. TMFA is applied to study the thermodynamically feasible ranges for the fluxes and the Gibbs free energy change, ΔrG′, of the reactions and the activities of the metabolites in the genome-scale metabolic model of Escherichia coli developed by Palsson and co-workers. In the TMFA of the genome scale model, the metabolite activities and reaction ΔrG′ are able to achieve a wide range of values at optimal growth. The reaction dihydroorotase is identified as a possible thermodynamic bottleneck in E. coli metabolism with a ΔrG′ constrained close to zero while numerous reactions are identified throughout metabolism for which ΔrG′ is always highly negative regardless of metabolite concentrations. As it has been proposed previously, these reactions with exclusively negative ΔrG′ might be candidates for cell regulation, and we find that a significant number of these reactions appear to be the first steps in the linear portion of numerous biosynthesis pathways. The thermodynamically feasible ranges for the concentration ratios ATP/ADP, NAD(P)/NAD(P)H, and are also determined and found to encompass the values observed experimentally in every case. Further, we find that the NAD/NADH and NADP/NADPH ratios maintained in the cell are close to the minimum feasible ratio and maximum feasible ratio, respectively.
A genome-scale metabolic network model synergizes with statistical learning to predict amino acid concentrations in Chinese hamster ovary cell cultures
- Schinn
What is flux balance?
- Orth