Enforcing Detailed Balance Reveals Systematic Thermodynamic Inconsistencies in Whole-Cell Models
Whole-cell computational models represent the frontier of systems biology, yet none enforce the detailed balance condition—the thermodynamic constraint that k_f/k_r = (- G^/RT) for every reaction (the Wegscheider condition). Here we show that enforcing this constraint as a simulator invariant reveals systematic thermodynamic inconsistencies in the state-of-the-art JCVI-syn3A whole-cell model, producing qualitatively different biological predictions. We develop a stochastic simulation framework in which every reaction’s forward and reverse propensities satisfy local detailed balance at every simulation event, parameterized from eQuilibrator standard transformed Gibbs energies ( G^). Applied to JCVI-syn3A, we find that of 77 metabolic reactions with complete kinetic and thermodynamic data, 70% violate detailed balance by more than 3-fold (median discrepancy: 12-fold), and 27% (21 of 77) have their equilibrium direction reversed at standard state—of which 12 persist at physiological intracellular concentrations. Five of ten glycolytic steps have their equilibrium direction reversed, including lactate dehydrogenase, where the thermodynamic equilibrium constant (K_eq = 48,743) differs from the model’s kinetic value (K_eq = 0.97) by 50,000-fold—predicting dominant lactate fermentation consistent with known Mycoplasma biology. Our framework, validated on a 13-reaction proof-of-concept and scaled to a 2,069-reaction model of syn3A (198 metabolic + 51 transport + 1,820 gene expression reactions, 1,671 species), provides a principled foundation for thermodynamically consistent whole-cell simulation.