Flux Transformer learns intracellular metabolic fluxes from isotope labeling and measured reaction rates
In the authors' words
Cells convert incoming nutrients into energy and biomass via metabolic fluxes. Quantifying these fluxes involves feeding cells isotope-labeled nutrients, quantifying metabolite labeling patterns, and inverting the labeling data to retrieve the underlying flux distribution. Traditional metabolic flux analysis (MFA) is both computationally expensive and frequently requires manual refinement of the cost function to obtain plausible results. For this reason, MFA has remained inaccessible to non-experts. Here we present Flux Transformer, an AI model that uses reaction-specific attention to integrate isotope-labeling data with available reaction-rate measurements, reconstructing mass-balanced steady-state fluxes. The model is trained on simulated data with added noise and identifies the underlying fluxes from these data more effectively than traditional MFA. Provided experimental data from wild-type and mutant bacteria, yeast, plants, and mammalian T cells, it efficiently and accurately recovers expert-curated MFA fluxes. By making flux analysis accessible, Flux Transformer is poised to accelerate metabolic research and engineering.
Appeared: Thursday, November 19. bioRxiv. Preprint, not yet peer-reviewed.