Hybrid Dynamic Models of Bioprocesses Based on Elementary Flux Modes and Multilayer Perceptrons
Abstract
:1. Introduction
2. Hybrid Modeling
2.1. EFM Selection
2.1.1. Metabolic Network Analysis
2.1.2. EFM Reduction Procedure
- Generation of the initial EFM set: The first step concerns the generation of elementary flux vectors. If the size of the metabolic network is not too big, the whole set of modes can be computed and enumerated using software tools such as EFMtool. Otherwise, if the number of EFMs becomes prohibitive, alternative methods to identify only subsets are required. A fast generation algorithm [13] can be used that requires also the knowledge of experimental measurements of uptake and excretion rates. Nevertheless, regardless the EFMs generation method, a matrix of elementary flux modes E can be obtained.
- Biological interpretation: The second step consists in ensuring a biological interpretation of the matrix K, which is defined by
- Main EFM reduction: This step allows reducing the number of modes up to a target . This value is generally set close to, and sometimes slightly greater than, to avoid computational issues during the final step of the procedure. First, collinearity between vectors is evaluated and the collinear modes are discarded. Second, an optimization-based reduction is achieved where a randomly selected vector is removed if the following inequality is satisfied:
- Selection of a minimal bioreaction model: For this final step, an even smaller number of EFMs is selected among the previous set of modes ( < ). This target is chosen below the number of extracellular measured species in order to derive macroscopic models with less reactions than components. This step is no longer based on random successive eliminations of modes but is a selection step of the best combination of EFMs among the previous modes. Hence, the performance index of all possible EFM combinations is computed and the final set of modes is the one with the smallest value of the indicator, which represents the distance to the experimental data. Note that the number of possible combinations is given byAs a consequence, has to be chosen small enough to avoid an unmanageable number of EFM combinations and, at the same time, large enough to have more flexibility in the selection, i.e., to have a sufficient pool of EFMs for the selection of the best EFM combination.
2.2. Dynamic Mass Balance Model
NN Kinetic Modeling
- one input layer, which distributes the input values to the first hidden layer;
- one or several hidden layers of perceptrons;
- one output layer, which recovers the output of each perceptron of the last hidden layer.
3. Case Study: Perfusion Cultures of Hybridoma Cell Line HB-58
3.1. Metabolic Network
3.2. Measurement Configuration
4. Numerical Results
4.1. EFM Selection
4.2. Kinetic Modeling
4.3. Interpretation of Overflow Metabolism
5. Conclusions
Author Contributions
Funding
Data Availability Statement
Conflicts of Interest
References
- Hodgson, B.; Taylor, C.; Ushio, J.; Leigh, J. Intelligent modelling of bioprocesses: A comparison of structured and unstructured approaches. Bioprocess Biosyst. Eng. 2005, 26, 353–359. [Google Scholar]
- Haag, J.; Vande Wouwer, A.; Bogaerts, P. Dynamic modeling of complex biological systems: A link between metabolic and macroscopic description. Math. Biosci. 2005, 193, 25–49. [Google Scholar] [CrossRef] [PubMed]
- Haag, J.; Vande Wouwer, A.; Bogaerts, P. Systematic procedure for the reduction of complex biological reaction pathways and the generation of macroscopic equivalents. Chem. Eng. Sci. 2005, 60, 459–465. [Google Scholar]
- Baroukh, C.; Bernard, O. Metabolic modeling of C. sorokiniana diauxic heterotrophic growth. IFAC-PapersOnLine 2016, 49, 330–335. [Google Scholar] [CrossRef]
- Schuster, S.; Hilgetag, C. On elementary flux modes in biochemical reaction systems at steady state. J. Biol. Syst. 1994, 2, 165–182. [Google Scholar] [CrossRef] [Green Version]
- Gao, J.; Gorenflo, V.; Scharer, J.; Budman, H. Dynamic metabolic modeling for a mAb bioprocess. Biotechnol. Prog. 2007, 23, 168–181. [Google Scholar] [CrossRef]
- Naderi, S.; Meshram, M.; Wei, C.; McConkey, B.; Ingalls, B.; Budman, H.; Scharer, J. Metabolic flux and nutrient uptake modeling of normal and apoptotic CHO cells. IFAC Proc. Vol. 2010, 43, 395–400. [Google Scholar]
- von Kamp, A.; Schuster, S. Metatool 5.0: Fast and flexible elementary modes analysis. Bioinformatics 2006, 22, 1930–1931. [Google Scholar] [CrossRef] [Green Version]
- Terzer, M.; Stelling, J. Large-scale computation of elementary flux modes with bit pattern trees. Bioinformatics 2008, 24, 2229–2235. [Google Scholar]
- Klamt, S.; Stelling, J. Combinatorial complexity of pathway analysis in metabolic networks. Mol. Biol. Rep. 2002, 29, 233–236. [Google Scholar]
- Figueiredo, L.; Podhorski, A.; Rubio, A.; Kaleta, C.; Beasley, J.; Schuster, S.; Planes, F. Computing the shortest elementary flux modes in genome-scale metabolic networks. Bioinformatics 2009, 25, 3158–3165. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Kaleta, C.; de Figueiredo, L.; Schuster, S. Can the whole be less than the sum of its parts ? Pathway analysis in genome-scale metabolic networks using elementary flux patterns. Genome Res. 2009, 19, 1872–1883. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Jungers, R.; Zamorano, F.; Blondel, V.; Vande Wouwer, A.; Bastin, G. Fast computation of minimal elementary decompositions of metabolic vectors. Automatica 2011, 47, 1255–1259. [Google Scholar] [CrossRef]
- Machado, D.; Soons, Z.; Patil, K.R.; Ferreira, E.C.; Rocha, I. Random sampling of elementary flux modes in large-scale metabolic networks. Bioinformatics 2012, 28, i515–i521. [Google Scholar] [CrossRef] [Green Version]
- Tabe-Bordbar, S.; Marashi, S. Finding elementary flux modes in metabolic networks based on flux balance analysis and flux coupling analysis: Application to the analysis of Escherichia coli metabolism. Biotechnol. Lett. 2013, 35, 2039–2044. [Google Scholar] [CrossRef]
- Soons, Z.; Rocha, E.; Ferreira, I. Selection of elementary modes for bioprocess control. Comput. Appl. Biotechnol. 2010, 11, 156–161. [Google Scholar] [CrossRef] [Green Version]
- Soons, Z.; Ferreira, I.; Rocha, E. Identification of minimal metabolic pathway models consistent with phenotypic data. J. Process Control 2011, 21, 1483–1492. [Google Scholar] [CrossRef] [Green Version]
- Oddsdottir, H.; Hagrot, E.; Chotteau, V.; Forsgren, A. On dynamically generating relevant elementary flux modes in a metabolic network using optimization. J. Math. Biol. 2015, 71, 903–920. [Google Scholar] [CrossRef] [Green Version]
- Oddsdottir, H.; Hagrot, E.; Chotteau, V.; Forsgren, A. Robustness analysis of elementary flux modes generated by column generation. Math. Biosci. 2016, 273, 45–56. [Google Scholar] [CrossRef]
- Provost, A.; Bastin, G. Dynamic metabolic modelling under the balanced growth condition. J. Process Control 2004, 14, 717–728. [Google Scholar] [CrossRef]
- Zamorano, F.; Vande Wouwer, A.V.; Jungers, R.M.; Bastin, G. Dynamic metabolic models of CHO cell cultures through minimal sets of elementary flux modes. J. Biotechnol. 2013, 164, 409–422. [Google Scholar] [CrossRef] [PubMed]
- Hebing, L.; Neymann, T.; Thüte, T.; Jockwer, A.; Engell, S. Efficient generation of models of fed-batch fermentations for process design and control. IFAC-PapersOnLine 2016, 49, 621–626. [Google Scholar] [CrossRef]
- Abbate, T.; Fernandes de Sousa, S.; Dewasme, L.; Bastin, G.; Vande Wouwer, A. Inference of dynamical macroscopic models of cell metabolism based on elementary flux modes analysis. Biochem. Eng. J. 2019, 151, 107325. [Google Scholar] [CrossRef]
- Maton, M.; Vande Wouwer, A.; Bogaerts, P. Selection of a minimal suboptimal set of EFMs for dynamic metabolic modelling. IFAC-PapersOnLine 2021, 54, 667–672. [Google Scholar] [CrossRef]
- Savageau, M. Biochemical systems analysis. I. Some mathematical properties of the rate law for the component enzymatic reactions. J. Theor. Biol. 1969, 25, 365–369. [Google Scholar] [CrossRef]
- Savageau, M. Biochemical systems analysis. II. The steady-state solutions for n-pool system using a power-law approximation. J. Theor. Biol. 1969, 25, 370–379. [Google Scholar] [CrossRef]
- Voit, E.; Savageau, M. Equivalence between S-systems and Volterra-systems. Math. Biosci. 1982, 78, 47–55. [Google Scholar] [CrossRef] [Green Version]
- Savageau, M. Introduction to S-systems and the underlying power-law formalism. Math. Comput. Model. 1979, 11, 546–551. [Google Scholar] [CrossRef] [Green Version]
- Shiraishi, F.; Savageau, M. The tricarboxylic acid cycle in Dictiostelium discoideum. Formulation of the alternative kinetic representations. J. Biol. Chem. 1992, 267, 22912–22918. [Google Scholar] [CrossRef]
- Curto, R.; Sorribas, A.; Cascante, M. Comparative characterization of the fermentation pathway of Saccharomyces cerevisiae using biochemical systems theory and metabolic control analysis: Model definition and nomenclature. Math. Biosci. 1995, 130, 25–50. [Google Scholar] [CrossRef]
- Torres, N.; Voit, E.; Gonzalez-Alcon, C. Optimization of nonlinear biotechnological processes with linear programming: Application to citric acid production by Aspergillus niger. Biotechnol. Bioeng. 1996, 49, 247–258. [Google Scholar] [CrossRef]
- Hernandez-Bermejo, B.; Fairen, V.; Sorribas, A. Power-law modeling based on least-squares minimization criteria. Math. Biosci. 1999, 161, 83–94. [Google Scholar] [CrossRef]
- Haag, J.; Vande Wouwer, A.; Remy, M. A general model of reaction kinetics in biological systems. Bioprocess Biosyst. Eng. 2005, 27, 303–309. [Google Scholar] [CrossRef] [PubMed]
- Naderi, S.; Meshram, M.; Wei, C.; McConkey, B.I.B.; Budman, H.; Scharer, J. Development of a mathematical model for evaluating the dynamics of normal and apoptotic Chinese hamster ovary cells. Biotechnol. Prog. 2011, 27, 1197–1205. [Google Scholar] [CrossRef] [PubMed]
- Hagrot, E.; Oddsdottir, H.; Hosta, J.; Jacobsen, E.; Chotteau, V. Poly-pathway model, a novel approach to simulate multiple metabolic states by reaction network-based model—Application to amino acid depletion in CHO cell culture. J. Biotechnol. 2017, 259, 235–247. [Google Scholar] [CrossRef] [PubMed]
- Mailier, J.; Vande Wouwer, A. Identification of nested biological kinetic models using likelihood ratio tests. Chem. Eng. 2012, 84, 727–734. [Google Scholar] [CrossRef]
- Grosfils, A.; Vande Wouwer, A.; Bogaerts, P. On a general model structure for macroscopic biological reaction rates. J. Biotechnol. 2007, 130, 253–264. [Google Scholar] [CrossRef]
- Richelle, A.; Bogaerts, P. Systematic methodology for bioprocess model identification based on generalized kinetic functions. Biochem. Eng. J. 2015, 100, 41–49. [Google Scholar] [CrossRef]
- Montague, G.; Morris, J. Neural-network contributions in biotechnology. Trends Biotechnol. 1994, 12, 312–324. [Google Scholar] [CrossRef]
- Chen, L.; Bernard, O.; Bastin, G.; Angelov, P. Hybrid modeling of biotechnological processes using neural networks. Control Eng. Pract. 2000, 8, 821–827. [Google Scholar] [CrossRef]
- Vande Wouwer, A.; Renotte, C.; Bogaerts, P. Biological reaction modeling using radial basis function networks. Comput. Chem. Eng. 2004, 28, 2157–2164. [Google Scholar] [CrossRef]
- Maton, M.; Vande Wouwer, A.; Bogaerts, P. A systematic elementary flux mode selection procedure for deriving macroscopic bioreaction models from metabolic networks. J. Process Control 2022, 118, 170–184. [Google Scholar] [CrossRef]
- Amribt, Z.; Hongxing, N.; Bogaerts, P. Macroscopic modelling of overflow metabolism and model based optimization of hybridoma cell fed-batch cultures. Biochem. Eng. J. 2013, 70, 196–209. [Google Scholar] [CrossRef]
- Meyer-Baese, A.; Schmid, V. Foundations of neural networks. In Pattern Recognition and Signal Analysis in Medical Imaging, 2nd ed.; Academic Press: Cambridge, MA, USA, 2014; pp. 197–243. [Google Scholar]
- Hecht-Nielsen, R. Theory of the backpropagation neural network. In Neural Networks for Perception: Computation, Learning, and Architectures; Academic Press: Cambridge, MA, USA, 1992; pp. 65–93. [Google Scholar]
- Janowsky, S. Pruning versus clipping in neural networks. Phys. Rev. A 1989, 39, 6600. [Google Scholar] [CrossRef] [PubMed]
- Reed, R. Pruning algorithms: A survey. IEEE Trans. Neural Netw. 1993, 4, 740–747. [Google Scholar] [CrossRef]
- Hang, S.; Pool, J.; Tran, J.; Dally, W. Learning both weights and connections for efficient neural network. Adv. Neural Inf. Process. Syst. 2015, 28, 1135–1143. [Google Scholar]
- Frankle, J.; Carbin, M. The lottery ticket hypothesis: Finding sparse, trainable neural networks. In Proceedings of the International Conference on Learning Representations, New Orleans, LA, USA, 6–9 May 2019. [Google Scholar]
- Tanaka, H.; Kunin, D.; Yamins, D.; Ganguli, S. Pruning neural networks without any data by iteratively conserving synaptic flow. Adv. Neural Inf. Process. Syst. 2015, 33, 6377–6389. [Google Scholar]
- Silvestre, M.; Ling, L. Pruning methods to MLP neural networks considering proportional apparent error rate for classification problems with unbalanced data. Measurement 2014, 56, 88–94. [Google Scholar] [CrossRef]
- Niu, H.; Amribt, Z.; Fickers, P.; Tan, W.; Bogaerts, P. Metabolic pathway analysis and reduction for mammalian cell cultures—Towards macroscopic modeling. Chem. Eng. Sci. 2013, 102, 461–473. [Google Scholar] [CrossRef]
- Fernandes de Sousa, S.; Bastin, G.; Jolicoeur, M.; Vande Wouwer, A. Dynamic metabolic flux analysis using a convex analysis approach: Application to hybridoma cell cultures in perfusion. Biotechnol. Bioeng. 2015, 113, 1102–1121. [Google Scholar] [CrossRef]
- Dochain, D.; Vanrolleghem, P. Identification of bioprocess models. In Bioprocess Control; ISTE Ltd.: London, UK; John Wiley & Sons Inc.: Hoboken, NJ, USA, 2008; pp. 47–76. [Google Scholar]
- Quesney, S.; Marc, A.; Gerdil, C.; Gimenez, C.; Marvel, J.; Richard, Y.; Meignier, B. Kinetics and metabolic specificities of Vero cells in bioreactor cultures with serum-free medium. Cytotechnology 2003, 42, 1–11. [Google Scholar] [CrossRef] [PubMed]
Experiment | ( Cells/L) | (h) | Feed Stream (mM) | |
---|---|---|---|---|
1 | 0.19 | 54 | 11 | 5 |
2 | 0.23 | 56 | 15 | 11.5 |
3 | 0.36 | 48 | 28 | 4 |
4 | 0.36 | 44 | 28 | 9.5 |
Parameters | Value | Parameters | Value | Parameters | Value |
---|---|---|---|---|---|
0.0150 | 0.1459 | 0.0194 | |||
0.0258 | 0.2425 | 0.4851 | |||
0.0729 | 0.0092 | 0.2408 | |||
0.2425 | 0.1204 | 0.7925 |
# Neurons | # Parameters |
---|---|
10 | 125 |
5 | 65 |
4 | 53 |
3 | 41 |
2 | 29 |
1 | 17 |
5.8 | 15.5 | 33.5 | 20 | 7.1 | 77 | 6.6 | 47.1 | 26.5 | 105.3 | 58.7 | |||
7.1 | 4.5 | 10.4 | 14.5 | 8.6 | 8.3 | & 125.7 | 1 | 1 | 1.6 | 13.4 | |||
7.3 | 31.7 | 40.4 | 11.5 | 5.8 | 9.5 | 4.7 | 73.3 | 38.2 | 41.1 | 710.7 | |||
32.8 | 50.9 | 82 | 169.6 | ||||||||||
3.5 | 2.1 | 7.7 | 11.3 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Maton, M.; Bogaerts, P.; Vande Wouwer, A. Hybrid Dynamic Models of Bioprocesses Based on Elementary Flux Modes and Multilayer Perceptrons. Processes 2022, 10, 2084. https://doi.org/10.3390/pr10102084
Maton M, Bogaerts P, Vande Wouwer A. Hybrid Dynamic Models of Bioprocesses Based on Elementary Flux Modes and Multilayer Perceptrons. Processes. 2022; 10(10):2084. https://doi.org/10.3390/pr10102084
Chicago/Turabian StyleMaton, Maxime, Philippe Bogaerts, and Alain Vande Wouwer. 2022. "Hybrid Dynamic Models of Bioprocesses Based on Elementary Flux Modes and Multilayer Perceptrons" Processes 10, no. 10: 2084. https://doi.org/10.3390/pr10102084