A comparative analysis of metabolic regulators (CAROM) found a conserved, attribute-based partitioning of metabolic regulation among four mechanisms in yeast and E. coli: growth-limiting (essential) enzymes are preferentially acetylated, high-flux reactions and futile cycles are preferentially phosphorylated, highly connected reactions attract multiple mechanisms, and reversible reactions are largely unregulated. However, a subset of essential reactions (about 14% in the yeast stationary-phase analysis of the preprint) is regulated by none of the four mechanisms, the analysis covers only a few conditions, and the mechanistic origin of the partitioning is unexplained. The open problem is to complete this regulatory architecture: determine which mechanisms — allosteric regulation, additional post-translational modifications such as ubiquitination, methylation, or succinylation, or non-coding-RNA regulation — govern the unassigned reactions; establish whether the attribute-based partitioning persists across environmental and genetic conditions and extends to additional mechanisms; and explain mechanistically how the partitioning arises.

Background

Cellular metabolism is regulated at several layers: transcriptional regulation changes enzyme mRNA levels, translational and protein-level regulation change enzyme abundance, and post-translational modifications (PTMs) — chemical modifications of amino-acid residues, such as lysine acetylation and phosphorylation — reversibly alter enzyme activity without changing protein abundance. Genome-scale metabolic models (GEMs) catalogue every known reaction and gene of an organism's metabolism (for example the Yeast 7 reconstruction of Saccharomyces cerevisiae and the iJO1366 model of E. coli). Constraint-based methods applied to GEMs predict metabolic behavior without kinetic parameters: flux balance analysis (FBA) computes reaction fluxes that maximize an objective such as growth, gene-knockout simulations identify reactions essential for growth, and flux variability analysis (FVA) computes the maximum feasible flux (Vmax) through each reaction.

A complementary systems-level literature had connected single regulatory layers to metabolism: most prominently, Hackett et al. (2016) mapped known metabolite-enzyme (allosteric) regulatory interactions in yeast onto measured fluxes and enzyme properties, finding for instance that reversible reactions are underrepresented among allosterically regulated targets. What had been missing was a systematic comparison of several regulatory mechanisms against each other on the same network.

The CAROM framework (Comparative Analysis of Regulators of Metabolism; Chandrasekaran 2019 preprint, published as Smith, Shen, Lee and Chandrasekaran 2022) filled this gap. CAROM takes as input, for a given condition, lists of targets of different regulatory processes (from transcriptome, proteome, acetylome and phosphoproteome datasets — measurements of mRNA, protein, lysine-acetylation and phosphorylation changes) and compares the target sets using thirteen reaction attributes derived from the GEM, including FBA essentiality, FVA maximum flux, network connectivity (the number of neighboring reactions, closeness within the network, and PageRank centrality, a node-importance measure that weights a reaction's links by how well connected its neighbors are), reversibility, and enzyme molecular weight. Applied to the transition to stationary phase (a non-dividing, nutrient-limited state) in yeast and E. coli, and to yeast cell-cycle and nitrogen-starvation data, CAROM found a conserved 'division of labor': essential, growth-limiting reactions are strongly enriched among acetylation targets; reactions with high maximum flux, kinase-phosphatase pairs (a kinase and the opposing phosphatase that attach and remove the same phosphate group), isozymes (alternative enzymes that catalyze the same reaction), and energy-wasting 'futile cycles' are enriched among phosphorylation targets; highly connected reactions tend to be targeted by multiple mechanisms; and reversible reactions are predominantly unregulated or regulated by unknown mechanisms. A single machine-learning model using these features classified acetylation versus phosphorylation targets across yeast, E. coli and mammalian cells with AUC above 0.8, and its phosphorylation predictions during the mammalian cell cycle were validated against independent phosphoproteomics. Interpreting the model with game-theory-based feature attribution (Shapley analysis) identified network connectivity, essentiality and condition-specific maximum flux as the main discriminating attributes.

Three gaps remained. First, a fraction of reactions — including, in the preprint's yeast stationary-phase analysis, about 14% of essential reactions — was not targeted by any of the four mechanisms, and the authors proposed allosteric regulation or less-characterized PTMs as candidates while noting that context-specific datasets for those mechanisms were lacking; the published version's 'unregulated' category explicitly also covers unknown regulation, so part of the gap may reflect incomplete annotation rather than true absence of regulation. Second, the pattern was established from only a few conditions in two microbes, so its generality across environments, genetic backgrounds and organisms was untested. Third, the analysis establishes correlations between reaction attributes and regulatory targeting, not mechanisms: it does not explain how a cell's regulatory machinery comes to target high-flux reactions by phosphorylation and growth-limiting enzymes by acetylation. Resolving these gaps jointly would amount to completing the basic regulatory architecture of metabolic networks, the explicit goal stated in the published paper's conclusion.

Problem Statement

Complete the division-of-labor map of metabolic regulation established by the CAROM analysis. Concretely, for metabolism as represented by genome-scale models such as Yeast 7 and iJO1366: (1) identify which regulatory mechanism or mechanisms govern the reactions that are targeted by none of transcription, translation (protein-level change), lysine acetylation, and phosphorylation — in the yeast stationary-phase analysis of the preprint this includes a subset of essential reactions (about 14%) — evaluating the candidates proposed in the source (allosteric regulation and less-characterized PTMs such as ubiquitination, methylation or succinylation, as well as non-coding-RNA regulation) against context-specific datasets, and separating genuine regulatory gaps from gaps in current annotations; (2) determine whether the attribute-based partitioning — enrichment of essential reactions among acetylation targets, of high-maximum-flux, isozyme and futile-cycle reactions among phosphorylation targets, of highly connected reactions among multiply regulated ones, and depletion of reversible reactions among all regulated ones — is conserved across environmental and genetic conditions beyond the few studied (stationary phase, cell cycle, nitrogen starvation) and across organisms, and whether each additional mechanism admits its own distinct, predictable reaction-attribute profile; and (3) provide a mechanistic account of how the partitioning arises, going beyond correlation — the two examples given (how maximal flux is sensed in phosphorylation targeting, and why growth-limiting enzymes are preferentially acetylated) illustrate the required general account, which is not restricted to them, while an evolutionary account of why the partitioning arose is preferred but not required. Deliverable (2) contains two claims that stand or fall independently: the conservation claim is a universal over conditions and organisms, so a single reproducible counterexample — an additional condition or organism that reverses the reported enrichments or depletions — resolves that claim negatively; whether each added mechanism admits its own predictable attribute profile is a claim that stands independently of the conservation claim. These three deliverables jointly support one objective, namely a complete, condition-resolved and mechanistically interpretable assignment of metabolic reactions to regulatory mechanisms. Construction of fully kinetic or dynamic whole-network models with embedded regulation is a related longer-term modeling program recorded in the source's limitations.

References

  1. Chandrasekaran S., 'Division of labor in metabolic regulation by transcription, translation, acetylation and phosphorylation', bioRxiv preprint, version 1, 11 November 2019. DOI: 10.1101/838243, https://doi.org/10.1101/838243. Version 2 of the same preprint (posted 3 June 2021 by Smith K., Shen F., Lee H.-J. and Chandrasekaran S.) was retitled 'Metabolic signatures of regulation by phosphorylation and acetylation' and published as the iScience 2022 paper listed below.
  2. Smith K., Shen F., Lee H.-J., Chandrasekaran S., 'Metabolic signatures of regulation by phosphorylation and acetylation', iScience, 2022. DOI: 10.1016/j.isci.2021.103730, https://doi.org/10.1016/j.isci.2021.103730
  3. Smith K., Rhoads N.M., Chandrasekaran S., 'Protocol for CAROM: A machine learning tool to predict post-translational regulation from metabolic signatures', STAR Protocols, 2022. DOI: 10.1016/j.xpro.2022.101799, https://doi.org/10.1016/j.xpro.2022.101799
  4. Chung C.H., Lin D.-W., Eames A., Chandrasekaran S., 'Next-Generation Genome-Scale Metabolic Modeling through Integration of Regulatory Mechanisms', Metabolites, 2021. DOI: 10.3390/metabo11090606, https://doi.org/10.3390/metabo11090606
  5. Hackett S.R., Zanotelli V.R., Xu W., Goya J., Park J.O., Perlman D.H., Gibney P.A., Botstein D., Storey J.D., Rabinowitz J.D., 'Systems-level analysis of mechanisms regulating yeast metabolic flux', Science, 2016. DOI: 10.1126/science.aaf2786, https://doi.org/10.1126/science.aaf2786
  6. Lee H.-J., Shen F., Eames A., Jedrychowski M.P., Chandrasekaran S., 'Dynamic metabolic network modeling of a mammalian cell cycle using time-course multi-omics data', bioRxiv preprint, 2021. DOI: 10.1101/2021.10.11.464012, https://doi.org/10.1101/2021.10.11.464012
  7. Al Zubaidi S.M., Nasar M.I., Notebaart R.A., Ralser M., Alam M.T., 'An enzyme activation network reveals extensive regulatory crosstalk between metabolic pathways', Molecular Systems Biology, 2025. DOI: 10.1038/s44320-025-00111-7, https://doi.org/10.1038/s44320-025-00111-7
  8. Varghese S., Soman S.P., Ahmed M., John L., Ramesh P., Soman S., Pai V.R., Raju R., 'Integrative Analysis of Post-Translational Modifications Identifies a PTM-Enriched Regulatory Core in Human Metabolic Enzymes', Metabolites, 16(3), 2026. DOI: 10.3390/metabo16030163, https://doi.org/10.3390/metabo16030163