[en] MOTIVATION: Sensitivity analysis is a useful tool to identify key parameters in metabolic models. It is typically only applied to the growth rate, disregarding the sensitivity of other solution variables to parameters. Further, sensitivity analysis of elementary flux modes could provide low-dimensional insights into optimal solutions, but they are not defined when a model is subject to inhomogeneous flux constraints, such as the frequently used ATP maintenance reaction.
RESULTS: We introduce optimal flux modes (OFMs), an analogue to EFMs, but specifically applied to optimal solutions of constraint-based models. Further, we prove that implicit differentiation can always be used to efficiently calculate the sensitivities of both whole-model solutions and OFM-based solutions to model parameters. This allows for fine-grained sensitivity analysis of the optimal solution, and investigation of how these parameters exert control on the optimal composition of OFMs. This novel framework is implemented in DifferentiableMetabolism.jl, a software package designed to efficiently differentiate solutions of constraint-based models. To demonstrate scalability, we differentiate solutions of 342 yeast models; additionally we show that sensitivities of specific subsystems can guide metabolic engineering. Applying our scheme to an Escherichia coli model, we find that OFM sensitivities predict the effect of knockout experiments on waste product accumulation. Sensitivity analysis of OFMs also provides key insights into metabolic changes resulting from parameter perturbations.
AVAILABILITY AND IMPLEMENTATION: Software introduced here is available as open-source Julia packages DifferentiableMetabolism.jl (https://github.com/stelmo/DifferentiableMetabolism.jl) and ElementaryFluxModes.jl (https://github.com/HettieC/ElementaryFluxModes.jl), which both work on all major operating systems and computer architectures. Code to reproduce all results is available from https://github.com/HettieC/DifferentiableOFMPaper, and as an archive from https://doi.org/10.5281/zenodo.15183208.
SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Research center :
Luxembourg Centre for Systems Biomedicine (LCSB): Bioinformatics Core (R. Schneider Group)
Abd-El-Haleem DA, Elkatory MR, Abu-Elreesh GM. Uncovering novel polyhydroxyalkanoate biosynthesis genes and unique pathway in yeast Hanseniaspora valbyensis for sustainable bioplastic production. Sci Rep 2024; 14: 27162.
Agrawal A, Barratt S, Boyd S et al. Differentiating through a cone program. arXiv: 1904.09043, 2019, preprint: not peer reviewed.
Basan M, Hui S, Okano H et al. Overflow metabolism in Escherichia coli results from efficient proteome allocation. Nature 2015; 528: 99-104.
Beg QK, Vazquez A, Ernst J et al. Intracellular crowding defines the mode and sequence of substrate uptake by Escherichia coli and constrains its metabolic activity. Proc Natl Acad Sci USA 2007; 104: 12663-8.
Bellut K, Krogerus K, Arendt EK. Lachancea fermentati strains isolated from kombucha: fundamental insights, and practical application in low alcohol beer brewing. Front Microbiol 2020; 11: 764.
Bruggeman FJ, Planqué R, Molenaar D et al. Searching for principles of microbial physiology. FEMS Microbiol Rev 2020; 44: 821-44.
Brunk E, Sahoo S, Zielinski DC et al. Recon3d enables a three-dimensional view of gene variation in human metabolism. Nat Biotechnol 2018; 36: 272-81.
Chiba Y, Akeboshi H. Glycan engineering and production of 'humanized'glycoprotein in yeast cells. Biol Pharm Bull 2009; 32: 786-95.
Contiero J, Beatty C, Kumari S et al. Effects of mutations in acetate metabolism on high-cell-density growth of Escherichia coli. J Ind Microbiol Biotechnol 2000; 24: 421-30.
Davidi D, Milo R. Lessons on enzyme kinetics from quantitative proteomics. Curr Opin Biotechnol 2017; 46: 81-9.
De Groot DH, Van Boxtel C, Planqué R et al. The number of active metabolic pathways is bounded by the number of cellular constraints at maximal metabolic rates. PLoS Comput Biol 2019; 15: e1006858.
De Maeseneire S, De Mey M, Vandedrinck S et al. Metabolic characterisation of E. coli citrate synthase and phosphoenolpyruvate carboxylase mutants in aerobic cultures. Biotechnol Lett 2006; 28: 1945-53.
De Mey M, De Maeseneire S, Soetaert W et al. Minimizing acetate formation in E. coli fermentations. J Ind Microbiol Biotechnol 2007; 34: 689-700.
Edwards JS, Palsson BO. Systems properties of the Haemophilus influenzae Rd metabolic genotype. J Biol Chem 1999; 274: 17410-6.
Fong SS, Nanchen A, Palsson BO et al. Latent pathway activation and increased pathway capacity enable Escherichia coli adaptation to loss of key metabolic enzymes. J Biol Chem 2006; 281: 8024-33.
Gagneur J, Klamt S. Computation of elementary modes: a unifying framework and the new binary approach. BMC Bioinformatics 2004; 5: 1-21.
Heckmann D, Campeau A, Lloyd CJ et al. Kinetic profiling of metabolic specialists demonstrates stability and consistency of in vivo enzyme turnover numbers. Proc Natl Acad Sci U S A 2020; 117: 23182-90.
Kacser H, Burns J. Rate control of biological processes. Symp Soc Exp Biol 1973; 27: 65-104.
Kim HU, Kim TY, Lee SY. Genome-scale metabolic network analysis and drug targeting of multi-drug resistant pathogen Acinetobacter baumannii AYE. Mol Biosyst 2010; 6: 339-48.
Klamt S, Regensburger G, Gerstl MP et al. From elementary flux modes to elementary flux vectors: metabolic pathway analysis with arbitrary linear flux constraints. PLoS Comput Biol 2017; 13: e1005409.
Kratochvíl M, Wilken SE, Ebenhöh O et al. Cobrexa 2: tidy and scalable construction of complex metabolic models. Bioinformatics 2025; 41: btaf056. 10.1093/bioinformatics/btaf056.
Kroll A, Rousset Y, Hu X-P et al. Turnover number predictions for kinetically uncharacterized enzymes using machine and deep learning. Nat Commun 2023; 14: 4139.
Li F, Yuan L, Lu H et al. Deep learning-based k cat prediction enables improved enzyme-constrained model reconstruction. Nat Catal 2022; 5: 662-72.
Liebermeister W. The value structure of metabolic states. bioRxiv, 10.1101/483891, 2018, preprint: not peer reviewed.
Liu G, Li S, Shi Q et al. Engineering of Saccharomyces pastorianus old yellow enzyme 1 for the synthesis of pharmacologically active (s)-profen derivatives. Mol Catal 2021; 507: 111568.
Long CP, Gonzalez JE, Sandoval NR et al. Characterization of physiological responses to 22 gene knockouts in Escherichia coli central carbon metabolism. Metab Eng 2016; 37: 102-13.
Lozano Terol G, Gallego-Jara J, Sola Martínez RA et al. Engineering protein production by rationally choosing a carbon and nitrogen source using e. coli bl21 acetate metabolism knockout strains. Microb Cell Fact 2019; 18: 19.
Meldal BHM, Forner-Martinez O, Costanzo MC et al. The complex portal-an encyclopaedia of macromolecular complexes. Nucleic Acids Res 2015; 43: D479-84. 10.1093/nar/gku975
Monk JM, Lloyd CJ, Brunk E et al. iML1515, a knowledgebase that computes Escherichia coli traits. Nat Biotechnol 2017; 35: 904-8.
Mori M, Hwa T, Martin OC et al. Constrained allocation flux balance analysis. PLoS Comput Biol 2016; 12: e1004913.
Nakano K, Rischke M, Sato S et al. Influence of acetic acid on the growth of escherichia coli k12 during high-cell-density cultivation in a dialysis reactor. Appl Microbiol Biotechnol 1997; 48: 597-601.
Orth JD, Thiele I, Palsson BØ. What is flux balance analysis? Nat Biotechnol 2010; 28: 245-8.
Pan JG, Rhee JS, Lebeault JM. Physiological constraints in increasing biomass concentration of Escherichia coli b in fed-batch culture. Biotechnol Lett 1987; 9: 89-94.
Park JM, Kim TY, Lee SY. Constraints-based genome-scale metabolic simulation for systems metabolic engineering. Biotechnol Adv 2009; 27: 979-88.
Piirainen MA, Salminen H, Frey AD. Production of galactosylated complex-type n-glycans in glycoengineered Saccharomyces cerevisiae. Appl Microbiol Biotechnol 2022; 106: 301-15.
Sánchez BJ, Zhang C, Nilsson A et al. Improving the phenotype predictions of a yeast genome-scale metabolic model by incorporating enzymatic constraints. Mol Syst Biol 2017; 13: 935.
Schmidt A, Kochanowski K, Vedelaar S et al. The quantitative and condition-dependent escherichia coli proteome. Nat Biotechnol 2016; 34: 104-10.
Schuster R, Schuster S. Refined algorithm and computer program for calculating all non-negative fluxes admissible in steady states of biochemical reaction systems with or without some flux rates fixed. Comput Appl Biosci 1993; 9: 79-85.
Terzer M. Large scale methods to enumerate extreme rays and elementary modes. Ph.D. thesis, ETH Zurich, 2009.
The UniProt Consortium. UniProt: the Universal Protein Knowledgebase in 2025. Nucleic Acids Res 2025; 53: D609-17. 10.1093/nar/gkae1010
Urbanczik R. Enumerating constrained elementary flux vectors of metabolic networks. IET Syst Biol 2007; 1: 274-9.
van den Bogaard S, Saa PA, Alter TB. Sensitivities in protein allocation models reveal distribution of metabolic capacity and flux control. Bioinformatics 2024; 40: btae691.
Varma A, Palsson BO. Metabolic flux balancing: basic concepts, scientific and practical use. Nat Biotechnol 1994; 12: 994-8.
Wilken SE, Besancon M, Kratochvíl M et al. Interrogating the effect of enzyme kinetics on metabolism using differentiable constraint-based models. Metab Eng 2022; 74: 72-82.
Zhao J, Baba T, Mori H et al. Effect of zwf gene knockout on the metabolism of Escherichia coli grown on glucose or acetate. Metab Eng 2004; 6: 164-74.
