A large number of genome-scale models of cellular metabolism are available for various organisms. These models include all known metabolic reactions based on the genome annotation. However, the reactions that are active are dependent on the cellular metabolic function or environmental condition. Constraint-based methods that integrate condition-specific transcriptomics data into models have been used extensively to investigate condition-specific metabolism. Here, we present a method (TEX-FBA) for modeling condition-specific metabolism that combines transcriptomics and reaction thermodynamics data to generate a thermodynamically-feasible condition-specific metabolic model. TEX-FBA is an extension of thermodynamic-based flux balance analysis (TFA), which allows the simultaneous integration of different stages of experimental data (e.g., absolute gene expression, metabolite concentrations, thermodynamic data, and fluxomics) and the identification of alternative metabolic states that maximize consistency between gene expression levels and condition-specific reaction fluxes. We applied TEX-FBA to a genome-scale metabolic model of Escherichia coli by integrating available condition-specific experimental data and found a marked reduction in the flux solution space. Our analysis revealed a marked correlation between actual gene expression profile and experimental flux measurements compared to the one obtained from a randomly generated gene expression profile. We identified additional essential reactions from the membrane lipid and folate metabolism when we integrated transcriptomics data of the given condition on the top of metabolomics and thermodynamics data. These results show TEX-FBA is a promising new approach to study condition-specific metabolism when different types of experimental data are available.Author summaryCells utilize nutrients via biochemical reactions that are controlled by enzymes and synthesize required compounds for their survival and growth. Genome-scale models of metabolism representing these complex reaction networks have been reconstructed for a wide variety of organisms ranging from bacteria to human cells. These models comprise all possible biochemical reactions in a cell, but cells choose only a subset of reactions for their immediate needs and functions. Usually, these models allow for a large flux solution space and one can integrate experimental data in order to reduce it and potentially predict the physiology for a specific condition. We developed a method for integrating different types of omics data, such as fluxomics, transcriptomics, metabolomics into genome-scale metabolic models that reduces the flux solution space. Using gene expression data, the algorithm maximizes the consistency between the predicted and experimental flux for the reactions and predicts biologically relevant flux ranges for the remaining reactions in the network. This method is useful for determining fluxes of metabolic reactions with reduced uncertainty and suitable for performing context- and condition-specific analysis in metabolic models using different types of experimental data.