Thermodynamically enabled and reaction attuned estimation of metabolic fluxes

Mathematical models of metabolism based on mass-balance constraints are essential for analyzing cellular biochemistry, but their predictions are often compromised by thermodynamically infeasible flux cycles-unrealistic loops of biochemical reactions that circulate continuously without external energy input. Here, we introduce Teraflux, a method that generates transcriptome-specific metabolic flux estimations free of these cycles. Teraflux maximizes Shannon entropy weighted by gene expression data while enforcing relaxed thermodynamic consistency. By distinguishing between chemical potentials and reaction rates, it uses flux constraints to encode biological irreversibility without overfitting the chemical potential space. Across diverse experimental conditions in Escherichia coli, Teraflux achieves high prediction accuracy, preserves biologically essential feasible cycles like the glyoxylate shunt, and prevents implausible metabolic states such as artificial glucose excretion. Furthermore, thermodynamic variability analysis confirms that Teraflux's predictions fundamentally align with physiological limits on intracellular metabolite concentrations. Teraflux provides a robust, computationally efficient tool for obtaining reliable, condition-specific fluxomes.