Constraint-based genome-scale metabolic models are becoming an established tool for using genomic and biochemical information to predict cellular phenotypes. While these models provide quantitative predictions for individual reactions and are readily scalable for any biological system, they have inherent limitations. Using current methods, it is difficult to computationally elucidate a specific network state that directly depicts an in vivo state, especially in the instances where the organism might be functionally in a suboptimal state. In this study, we generated RNA sequencing data to characterize the transcriptional state of the cellulolytic anaerobe, Clostridium thermocellum, and algorithmically integrated these data with a genome-scale metabolic model. The phenotypes of each calculated metabolic flux state were compared to 13 experimentally determined physiological parameters to identify the flux mapping that best matched the in vitro growth of C. thermocellum. By this approach we found predicted fluxes for 88 reactions to be changed between the best solely computational prediction (flux balance analysis) and the best experimentally derived prediction. The alteration of these 88 reaction fluxes led to a detailed network-wide flux mapping that was able to capture the suboptimal cellular state of C. thermocellum.