A large part of a cell's energy budget is determined by the operation of its internal biochemical machinery, or metabolism.  Predicting its behaviour at genome-scale (an issue with many important practical implications) requires the computation of non-equilibrium steady states for reaction fluxes that integrate basic regulatory, stoichiometric and thermodynamic constraints. The latter point is particularly delicate for a number of reasons.  We describe the problem and some recent advances in the field that make use of both topological and dynamical aspects to reconstruct the intracellular free energy landscape, and (hopefully) improve the predictive ability of existing schemes.