Computational modelling of genome-scale metabolic networks and its application to CHO cell cultures
Introduction
Ongoing development of systems biology in recent years derives mainly from the successful integration of many computational approaches into the experimental work. One of the most successful applications of computer science in biology is in the annotation of genomes from the vast data generated by DNA sequencing experiments. Here, the computational approaches have been essential for the analysis of large amounts of sequenced data as well as for their presentation and applications. The success stories of the annotation of genomes of different simple organisms [1] as well as of the human genome [2] were followed by the establishment of the first genome-scale metabolic models (GEM) [3].
GEMs are the most accurate in silico representation of the genotype-phenotype link [4]. These models are continuously improved with the accuracy of their descriptions as well as the strength of the predictions they make. For example, Recon, the GEM describing human metabolism, was first published in 2007 but it has now gone through six iterations of improvements [5], [6]. These improvements are a consequence of the evolution of experimental and computational approaches used in systems biology. They result from the publicly available large scale data through the literature and through different general as well as specific web databases, such as KEGG (Kyoto Encyclopaedia of Genes and Genomes) [7], BRENDA (BRaunschweig ENzyme DAtabase) [8] and BioCyc [9]. Publicly available computational models through databases such as BioModels [10] and BiGG Models [11] present an additional driving force for exchangeability of knowledge and computational tools. The main motivation towards the continuous development of GEMs is a vast scope of their applications. These range from (1) the design and optimisation of environmentally friendly bio-based production of fine chemicals with simple organisms [12], and so called genome-scale synthetic biology [13] to (2) the optimisation of biopharmaceutical manufacturing cell lines and processes in non-mammalian cells, such as Pichia pastoris [14], as well as in mammalian cells [15], such as Chinese Hamster Ovary (CHO) cells [16], and finally to (3) the identification and analysis of possible biomarkers of complex diseases, such as non-alcoholic fatty liver disease (NAFLD) [17], [18], [19] and cancer [20], [21].
Computational models can be used to identify the segments that need to be explained more accurately to obtain a valid representation of the system's response. This allows us to systematically increase the knowledge describing biological mechanisms governing complex networks. Integration of data obtained from experiments, literature and databases into metabolic computational models can be described with a circular iteration scheme of knowledge acquisition and model improvements as shown in Fig. 1. It consists of (1) data acquisition and refinement through experimental work, literature and publicly available databases; (2) establishment and optimisation of computational models using the acquired data; and (3) analysis and validation of computational models and their potential refinement through another iteration of the cycle. The computational approaches are essential in all three steps described in the scheme. Novel computational approaches, which can be used in the reconstruction, analysis, refinement and visualisation of metabolic models, are therefore vital for the continuous progress of systems biology.
Numerous computational methods are available in the field of metabolic modelling and analysis. Majority of these are derived from the constraint-based analysis. The development and application of these methods is driven by the publicly available toolboxes, such as Pathway Tools [22], RAVEN (Reconstruction, Analysis, and Visualisation of mEtabolic Networks) [23] and probably the most popular COBRA (COnstraint-Based Reconstruction and Analysis) toolbox [24], [25]. These toolboxes implement the majority of the available computational methods. They follow open source concepts and are easy to update with novel methods. Computational methods applied to the analysis of metabolic networks include basic analyses, which can predict the reaction fluxes that bring the network to its optimal state (for example flux balance analysis - FBA) [26]. Implemented methods can be used to tailor the metabolic model with a specific context (see for example [27], [28]), and automatic identification of reactions, which need to be blocked in order to achieve the optimal state of the metabolic network, for example the state in which the production of selected metabolite is optimal [29], [30]. Large attention has also been devoted to the development of different visualisation approaches (see for example Escher [31]). Visualisation is, however, still mostly performed manually (see for example ReconMap [32] for the visualisation of human metabolism model).
In the following chapters the review of the state-of-the-art methods for the analysis, reconstruction and visualisation of metabolic networks is described. We begin with the description of some general approaches for the modelling and analysis of biological systems with the emphasis on the GEMs (see Section 2). Furthermore, we describe the most comprehensive publicly available databases containing large experimental datasets and computational models (see Section 3). We comment on the approaches that can be used in the process of the reconstruction and visualisation of GEMs (see Sections 4 Reconstruction of genome-scale metabolic models, 5 Model visualisation). We overview the progress in the development of CHO GEMs in recent years (see Section 6). We demonstrate the application of selected computational methods on the analysis of the most recent and most complete CHO GEM, i.e. iCHO1766 [33] (see Section 7).
Section snippets
Constraint-based methods for the analysis of metabolic networks
Numerous computational methods have been developed for the computational reconstruction and analysis of metabolic networks in recent years. Most of these approaches have been integrated within different publicly available computational toolboxes, such as COBRA [24], [25] and RAVEN [23].
Computational analysis of molecular networks is usually performed on the basis of their stoichiometric description [34]. Here, each reaction is described with its stoichiometric coefficients [35]. Stoichiometric
Biological databases
Large experimental datasets together with different computational models have been made available in the form of publicly available databases in recent years. Some of these cover metabolic pathways for different organisms (e.g. KEGG [7] and MetaCyc [9]), while others focus on experimental data or computational implementations of different metabolic models (e.g. BiGG Models [45] and BioModels [10]). Table 1 lists some of the commonly used databases in the field of GEMs reconstruction and
Reconstruction of genome-scale metabolic models
GEMs systematically incorporate multi-omic data into an unified representation [46], [47], [48]. These data are often referred to as BiGG (Biochemical, Genetic and Genomic) data and represent the metabolic network of a specific organism [38]. GEM reconstruction has to describe every enzyme and its corresponding metabolic reactions within the metabolic network. The reconstruction must contain information about (1) substrates and products of each enzymatic reaction, (2) stoichiometric
Model visualisation
Visualisation of metabolic networks is important for the interpretation and understanding of their composition and comparison with similar networks. Visualisation is included within the majority of the existing pathway databases such as KEGG [7]. Visualised metabolic pathways and networks are however mostly manually drawn and stored in a static form [57]. During data updates, these images have to be modified manually. Moreover, manual visualisation of new metabolic pathways is extremely time
Computational modelling of CHO metabolism
CHO cells have become prevalent in the production of recombinant proteins for clinical applications [64]. These proteins should be therapeutically active, human-compatible, and target-specific [65]. In contrast to bacterial or yeast cells, mammalian cells are able to provide the proper protein folding, assembly and post-translational modifications, which are necessary in order to achieve high quality products [16]. In the last decades CHO cells have been widely applied to the production of
Case study: computational analysis of iCHO1766 model
We can use the FBA and its alternatives to predict the optimal cell growth and optimal product formation in dependency of different conditions. These include cell culture media composition and activity of enzymes catalysing the metabolic reactions within the network. Here we demonstrate the application of constraint-based approaches on three different analyses of iCHO1766 model, namely (1) basic FBA of metabolic network to assess the reference state of the network, (2) qualitative perturbation
Conclusion
Although the response of metabolic networks is mainly derived from simple enzymatic reactions, they possess complex and rich dynamical properties. Their study requires complex systems approaches. The dynamics of metabolic networks can be partially reproduced with the GEMs in combination with computational approaches we described. Even though the accuracy of these reconstructions are far from being perfect, our journey does not stop with the flawless GEMs. Combining computational models of
Acknowledgements
The research was partially supported by the scientific-research programme Pervasive Computing (P2-0359) financed by the Slovenian Research Agency in the years from 2009 to 2017 and by the basic research and application project Designed cellular logic (J1-6740) financed by the Slovenian Research Agency in the years from 2014 to 2017. We acknowledge also resources of FP7 CASyM (Coordinating Action Systems Medicine Europe, Grant no. 305033), and the Slovenian Research Agency grants P1-0390 and the
References (95)
- et al.
Using genome-scale models to predict biological capabilities
Cell
(2015) - et al.
Impact of synthetic biology and metabolic engineering on industrial production of fine chemicals
Biotechnol. Adv.
(2015) - et al.
Systems biology as a foundation for genome-scale synthetic biology
Curr. Opin. Biotechnol.
(2006) - et al.
Integration of systems biology in cell line and process development for biopharmaceutical manufacturing
Biochem. Eng. J.
(2016) - et al.
Dynamic flux balance analysis of diauxic growth in Escherichia coli
Biophys. J.
(2002) - et al.
Regulation of gene expression in flux balance models of metabolism
J. Theor. Biol.
(2001) - et al.
Whole-genome metabolic network reconstruction and constraint-based modeling?
Methods Enzym.
(2011) - et al.
Paint4net: COBRA toolbox extension for visualization of stoichiometric models of metabolism
Biosystems
(2012) - et al.
What can mathematical modelling say about CHO metabolism and protein glycosylation?
Comput. Struct. Biotechnol. J.
(2017) - et al.
The emerging CHO systems biology era: harnessing the omics revolution for biotechnology
Curr. Opin. Biotechnol.
(2013)
Optimizing amino acid composition of CHO cell culture media for a fusion protein production
Process Biochem.
Metabolic flux analysis of CHO cells at growth and non-growth phases using isotopic tracers and mass spectrometry
Metab. Eng.
Metabolic control at the cytosolmitochondria interface in different growth phases of CHO cells
Metab. Eng.
Dynamic model of CHO cell metabolism
Metab. Eng.
Parallel labeling experiments with [1,2-13C]glucose and [U-13C]glutamine provide new insights into CHO cell metabolism
Metab. Eng.
Dynamic metabolic flux analysis using B-splines to study the effects of temperature shift on CHO cell metabolism
Metab. Eng. Commun.
Reducing Recon 2 for steady-state flux analysis of HEK cell culture
J. Biotechnol.
The biomass objective function
Curr. Opin. Microbiol.
Responses of CHO-DHFR cells to ratio of asparagine to glutamine in feed media: cell growth, antibody production, metabolic waste, glutamate, and energy metabolism
Bioresour. Bioprocess.
Integrated network analysis reveals an association between plasma mannose levels and insulin resistance
Cell Metab.
The principles of whole-cell modeling
Curr. Opin. Microbiol.
Biomedical applications of cell- and tissue-specific metabolic network models
J. Biomed. Inf.
The complete genome sequence of escherichia coli K-12
Science
Initial sequencing and analysis of the human genome
Nature
Basic and applied uses of genome-scale metabolic network reconstructions of Escherichia coli
Mol. Syst. Biol.
Global reconstruction of the human metabolic network based on genomic and bibliomic data
Proc. Natl. Acad. Sci.
Recon 2.2: from reconstruction to model of human metabolism
Metabolomics
From genomics to chemical genomics: new developments in KEGG
Nucleic Acids Res.
BRENDA, AMENDA and FRENDA: the enzyme information system in 2007
Nucleic Acids Res.
The MetaCyc database of metabolic pathways and enzymes and the BioCyc collection of pathway/genome databases
Nucleic Acids Res.
BioModels: content, features, functionality and use
CPT Pharmacomet. Syst. Pharmacol.
BiGG models: a platform for integrating, standardizing and sharing genome-scale models
Nucleic Acids Res.
Genome-scale metabolic model of Pichia pastoris with native and humanized glycosylation of recombinant proteins
Biotechnol. Bioeng.
Recombinant protein therapeutics from CHO cells – 20 years and counting
Chem. Eng. Prog.
SteatoNet: the first integrated human metabolic model with multi-layered regulation to investigate liver-associated pathologies
PLoS Comput. Biol.
Genome-scale metabolic modelling of hepatocytes reveals serine deficiency in patients with non-alcoholic fatty liver disease
Nat. Commun.
Large-scale computational models of liver metabolism: how far from the clinics?
Hepatology
Flux balance analysis predicts essential genes in clear cell renal cell carcinoma metabolism
Sci. Rep.
Cancer metabolism: a modeling perspective
Front. Physiol.
Pathway Tools version 19.0: integrated software for pathway/genome informatics and systems biology
Brief. Bioinforma.
The RAVEN toolbox and its use for generating a genome-scale metabolic model for Penicillium chrysogenum
PLoS Comput. Biol.
Quantitative prediction of cellular metabolism with constraint-based models: the COBRA toolbox
Nat. Protoc.
Quantitative prediction of cellular metabolism with constraint-based models: the COBRA Toolbox v2.0
Nat. Protoc.
What is flux balance analysis?
Nat. Biotechnol.
Context-specific metabolic networks are consistent with experiments
PLoS Comput. Biol.
A systematic evaluation of methods for tailoring genome-scale metabolic models
Cell Syst.
Optknock: a bilevel programming framework for identifying gene knockout strategies for microbial strain optimization
Biotechnol. Bioeng.
Cited by (20)
Integration of omics data to generate and analyse COVID-19 specific genome-scale metabolic models
2022, Computers in Biology and MedicineCitation Excerpt :GEMs have gained large importance in different fields ranging from biotechnology to systems biology and systems medicine [5,22]. For example, GEMs have been applied in cancer research [23,24], drug targeting in pathogens [25], prediction of drug side effects [26], and optimisation of production of recombinant proteins using Chinese hamster ovary (CHO) cells [27]. Reconstruction and analysis of GEMs are mostly based on constraint-based approaches.
What CHO is made of: Variations in the biomass composition of Chinese hamster ovary cell lines
2020, Metabolic EngineeringCitation Excerpt :Genome-scale metabolic models are comprehensive cell (line) specific reconstructions of cellular metabolism (Thiele and Palsson, 2010) that provide a mathematical framework for the analysis of metabolism and the integration of multi-omics data. In fact, first ‘‘real life’’ applications are already available (Calmels et al., 2019; Rejc et al., 2017), demonstrating their usefulness. Two key ingredients determine the predictive qualities of a genome-scale metabolic model, namely (i) a cell line-specific reconstruction of the metabolic capabilities (Calmels et al., 2019) as well as (ii) a cell line-and condition-specific biomass composition (Lachance et al., 2019; Tomàs-Gamisans et al., 2018).
A non-dominated sorting Differential Search Algorithm Flux Balance Analysis (ndsDSAFBA) for in silico multiobjective optimization in identifying reactions knockout
2019, Computers in Biology and MedicineCitation Excerpt :In silico metabolic engineering is a famous method that offers several advantages, including reducing the costs and time taken, providing prior knowledge for wet lab experiments, and less laborious methodology. These advantages have enabled researchers to simulate and re-engineer organisms with the aim of improving and exploiting their capabilities, for example, optimising the production of high-demand industrial metabolites [1,2]. Constraints-based modelling (CBM) methods have been used to predict the behaviour of organisms, contributing important knowledge for predicting the effects of phenotypic and genotypic perturbations on the organisms themselves.
Dynamic metabolic network modeling of mammalian Chinese hamster ovary (CHO) cell cultures with continuous phase kinetics transitions
2019, Biochemical Engineering JournalCitation Excerpt :The latter offer dynamic description of processes within CHO cells, based on kinetic expressions and parameter estimation algorithms [12]; see also recent reviews [13,14]. These approaches, together with bio-based constraints, provide a foundation of genome-scale metabolic models (GEM) [15,16]. To this day, over 100 theoretical methods were developed in this field [17], some of them are part of publicly available toolboxes, such as RAVEN (reconstruction, analysis, and visualization of metabolic networks) [18] and COBRA (constrained-based reconstruction and analysis) [19].
Identification of a dynamic metabolic flux model for a mammalian cell culture
2019, IFAC-PapersOnLineAdvancements in CHO metabolomics: techniques, current state and evolving methodologies
2024, Frontiers in Bioengineering and Biotechnology
- 1
Authors contributed equally to this work.