The AGORA resource of genome-scale metabolic reconstructions for 773 human gut microorganisms (Magnúsdóttir et al., 2017) made its strain-resolved models compatible with the human reconstruction Recon 2, yet explicitly left open the methodological and computational challenge of integrating large-scale microbial community models with host metabolic models to predict host phenotypes and systematically map microbial metabolic effects onto host metabolism. Subsequent work substantially advanced parts of this agenda — whole-body metabolic (WBM) models embedding personalized AGORA communities in an organ-resolved human host (2020), and AGORA2 with 7,302 strains and validated drug-metabolism coverage (2023) — but a scalable framework whose host-phenotype predictions are quantitatively validated against independent experimental data, at full strain resolution, with a systematic mapping of microbial community effects onto host metabolism, does not yet exist: 2025 reviews of host–microbe and community-scale metabolic modeling still identify quantitative validation and scalability as the central unmet challenges. The problem is to construct, scale, and benchmark such a coupled host–microbiome simulation capability.

Background

A genome-scale metabolic model (GEM) is a mathematical representation, assembled from an organism's annotated genome and biochemical literature, of all metabolic reactions a cell can catalyze; flux balance analysis (FBA) then predicts feasible reaction fluxes by optimizing a cellular objective (such as growth) subject to the steady-state mass-balance constraint S⋅v=0S\cdot v=0 — where SS is the stoichiometric matrix (the matrix of metabolite coefficients in each reaction) and vv is the vector of reaction fluxes — and to nutrient-availability limits. Recon 2 (Thiele et al., 2013) is a community-driven GEM of human metabolism covering thousands of genes and reactions, usable for cell-type- and tissue-level simulation. Magnúsdóttir et al. (2017) constructed AGORA (Assembly of Gut Organisms through Reconstruction and Analysis), 773 strain- and species-resolved GEMs of human gut microorganisms built by a semi-automated comparative curation pipeline, and made them compartment-compatible with Recon 2 so that microbial exchange metabolites can be matched to host transport and metabolism. This matters because the gut microbiota chemically modifies diet and host-derived molecules before and after host absorption (host–microbiome co-metabolism): microbial production of short-chain fatty acids (bacterial fermentation products such as butyrate and propionate, which the host absorbs and uses as an energy source), neurotransmitters, bile-acid transformations, and drug metabolism all influence host physiology, yet the responsible strain-to-host-pathway links are hard to resolve experimentally in humans. A coupled in silico model — a strain-resolved microbial community model exchanging metabolites with a human host GEM through a shared gut-lumen compartment — would let one predict, from metagenomic and clinical data, how a given individual's microbiome alters host organ metabolism and blood metabolites. The AGORA paper itself stated the open problem: joining large-scale microbial community models to host metabolic reconstructions for predictive host–microbiome simulation, and mapping microbial effects onto host metabolism, was a methodological and computational challenge the resource enables but does not solve. It further acknowledged that its semi-automated curation introduces consistent biases requiring strain-specific refinement (for example experimentally determined species-specific biomass objective functions, i.e., the reaction defining growth-associated macromolecular requirements, and cell-wall and lipopolysaccharide (the lipid–sugar macromolecules of Gram-negative bacterial outer membranes) biosynthesis pathways), and that xenobiotic (drug and foreign-compound) metabolism was out of scope. Progress since has reshaped the landscape: the Microbiome Modeling Toolbox (Baldini et al., 2019) supports personalized community models but at the community level only; MICOM (Diener et al., 2020), a method that builds community models directly from metagenomic sequencing data, scales community modeling to metagenomes but likewise without a host model; the whole-body metabolic (WBM) reconstructions (Thiele et al., 2020) provide organ-resolved human host models (over 80,000 reactions, 26 organs, six blood cell types) and demonstrated 'myWBM' models that embed personalized AGORA communities of roughly 131 strains per individual from 149 Human Microbiome Project participants, predicting microbiome-dependent neurotransmitter biosynthesis potential, butyrate supply to colonocytes, and ethanol-related fluxes; AGORA2 (Heinken et al., 2023) expanded coverage to 7,302 strains and added strain-resolved metabolism of 98 drugs validated against independent experimental datasets with accuracies of 0.72–0.84; and personalized myWBM analyses have been applied to Parkinson's and Alzheimer's disease cohorts (Hensen & Thiele, 2025; Hensen et al., 2024), while Best et al. (2025) coupled 181 mouse gut microorganism models to a mouse host model. However, reviews in 2025 (Srinak et al.; Quinn-Bohmann et al.) conclude that capturing the complexity and dynamic nature of host–microbe metabolic interaction with GEMs remains a significant challenge and that no robust validation framework yet establishes quantitatively accurate community- and host-scale predictions — so the coupling capability exists in demonstration form, but validated, scalable, predictive host–microbiome simulation does not.

Problem Statement

Construct and validate a scalable computational framework that couples strain-resolved gut microbial community metabolic models at the scale of the AGORA/AGORA2 resource — AGORA being the published collection of 773 strain-resolved gut-microbial genome-scale reconstructions and AGORA2 its 7,302-strain successor — with a human genome-scale metabolic reconstruction of the Recon 2 lineage (Recon 2 and its descendants, including organ-resolved whole-body models), so that, given an individual's metagenomic and relevant host data, the coupled model (i) predicts host metabolic phenotypes — such as biofluid metabolite levels and organ-level metabolic flux states — and (ii) systematically maps microbial community composition and strain- or reaction-level activity onto host metabolic pathways, quantifying which microbial contributions drive which host effects. The framework must be demonstrated at a scale of at least a thousand strain reconstructions per individual — substantially beyond the ~131 strains per individual used in current whole-body-model applications and toward the thousands of strain reconstructions available in resources such as AGORA2 — and its host-phenotype predictions must be quantitatively benchmarked against independent experimental measurements (for example, held-out cohort metabolomics) under a transparent validation protocol. The common objective uniting these deliverables is predictive, validated host–microbiome metabolic simulation; a demonstration that models can be joined without validated host-phenotype prediction does not complete the problem, nor does a community-scale model without a host component.

References

  1. Magnúsdóttir S, Heinken A, Kutt L, Ravcheev DA, Bauer E, Noronha A, Greenhalgh K, Jäger C, Baginska J, Wilmes P, Fleming RMT, Thiele I. Generation of genome-scale metabolic reconstructions for 773 members of the human gut microbiota. Nature Biotechnology 35:81–89 (2017). DOI: 10.1038/nbt.3703. https://doi.org/10.1038/nbt.3703
  2. Thiele I, Swainston N, Fleming RMT, Hoppe A, Sahoo S, Aurich MK, Haraldsdóttir H, Mo ML, Rolfsson O, Stobbe MD, Thorleifsson SG, Agren R, Bölling C, Bordel S, Chavali AK, Dobson P, Dunn WB, Endler L, Hala D, Hucka M, Hull D, Jameson D, Jamshidi N, Jónsson JJ, Juty N, Keating S, Nookaew I, Le Novère N, Malys N, Mazein A, Papin JA, Price ND, Selkov E, Sigurdsson MI, Simeonidis E, Sonnenschein N, Smallbone K, Sorokin A, van Beek JH, Weichart D, Goryanin I, Nielsen J, Westerhoff HV, Kell DB, Mendes P, Palsson BØ. A community-driven global reconstruction of human metabolism. Nature Biotechnology 31:419–425 (2013). DOI: 10.1038/nbt.2488. https://doi.org/10.1038/nbt.2488
  3. Babaei P, Shoaie S, Ji B, Nielsen J. Challenges in modeling the human gut microbiome. Nature Biotechnology 36:682–686 (2018). DOI: 10.1038/nbt.4213. https://doi.org/10.1038/nbt.4213
  4. Magnúsdóttir S, Heinken A, Fleming RMT, Thiele I. Reply to 'Challenges in modeling the human gut microbiome'. Nature Biotechnology 36:686–691 (2018). DOI: 10.1038/nbt.4212. https://doi.org/10.1038/nbt.4212
  5. Baldini F, Heinken A, Heirendt L, Magnúsdóttir S, Fleming RMT, Thiele I. The Microbiome Modeling Toolbox: from microbial interactions to personalized microbial communities. Bioinformatics 35:2332–2334 (2019). DOI: 10.1093/bioinformatics/bty941. https://doi.org/10.1093/bioinformatics/bty941
  6. Diener C, Gibbons SM, Resendis-Antonio O. MICOM: Metagenome-Scale Modeling To Infer Metabolic Interactions in the Gut Microbiota. mSystems 5:e00606-19 (2020). DOI: 10.1128/msystems.00606-19. https://doi.org/10.1128/msystems.00606-19
  7. Thiele I, Sahoo S, Heinken A, Hertel J, Heirendt L, Aurich MK, Fleming RMT. Personalized whole-body models integrate metabolism, physiology, and the gut microbiome. Molecular Systems Biology 16:e9898 (2020). DOI: 10.15252/msb.20198982. https://doi.org/10.15252/msb.20198982
  8. Heinken A, Hertel J, Acharya G, Ravcheev DA, Nyga M, Okpala OE, Hogan M, Magnúsdóttir S, Martinelli F, Nap B, Preciat G, Edirisinghe JN, Henry CS, Fleming RMT, Thiele I. Genome-scale metabolic reconstruction of 7,302 human microorganisms for personalized medicine. Nature Biotechnology 41:1320–1331 (2023). DOI: 10.1038/s41587-022-01628-0. https://doi.org/10.1038/s41587-022-01628-0
  9. Hensen T, Ahmad S, Kastenmüller G, Kraaij R, Ghanbari M, Ikram A, Kaddurah-Daouk R, Thiele I. In silico metabolic modelling links microbiome-derived metabolites to risk factors of Alzheimer's disease. Gut Microbes Reports 2:2443171 (2024). DOI: 10.1080/29933935.2024.2443171. https://doi.org/10.1080/29933935.2024.2443171
  10. Best L, Dost T, Esser D, Flor S, Gamarra AM, Haase M, Kadibalban AS, Marinos G, Walker A, Zimmermann J, Simon R, Schmidt S, Taubenheim J, Künzel S, Häsler R, Franzenburg S, Groth M, Waschina S, Rosenstiel P, Sommer F, Witte OW, Schmitt-Kopplin P, Baines JF, Frahm C, Kaleta C. Metabolic modelling reveals the aging-associated decline of host–microbiome metabolic interactions in mice. Nature Microbiology 10:973–991 (2025). DOI: 10.1038/s41564-025-01959-z. https://doi.org/10.1038/s41564-025-01959-z
  11. Hensen T, Thiele I. Metabolic modeling links gut microbiota to metabolic markers of Parkinson's disease. Gut Microbes 17:2554195 (2025). DOI: 10.1080/19490976.2025.2554195. https://doi.org/10.1080/19490976.2025.2554195
  12. Srinak N, Krüger F, Kaleta C, Taubenheim J. Metabolic modeling of host-microbe interactions. Computational and Structural Biotechnology Journal 27:4304–4319 (2025). DOI: 10.1016/j.csbj.2025.10.006. https://doi.org/10.1016/j.csbj.2025.10.006
  13. Quinn-Bohmann N, Carr AV, Diener C, Gibbons SM. Moving from genome-scale to community-scale metabolic models for the human gut microbiome. Nature Microbiology 10:1055–1066 (2025). DOI: 10.1038/s41564-025-01972-2. https://doi.org/10.1038/s41564-025-01972-2