Cell-free protein expression has emerged as an important approach in systems and synthetic biology, and a promising technology for personalized point of care medicine. Cell-free systems derived from crude whole cell extracts have shown remarkable utility as a protein synthesis technology. However, if cell-free platforms for on-demand biomanufacturing are to become a reality, the performance limits of these systems must be defined and optimized. Toward this goal, we modeled E. coli cell-free protein expression using a sequence specific dynamic constraint-based approach in which metabolite measurements were directly incorporated into the flux estimation problem. A cell-free metabolic network was constructed by removing growth associated reactions from the iAF1260 reconstruction of K-12 MG1655 E. coli. Sequence specific descriptions of transcription and translation processes were then added to this metabolic network to describe protein production. A linear programming problem was then solved over short time intervals to estimate metabolic fluxes through the augmented cell-free network, subject to material balances, time rate of change and metabolite measurement constraints. The approach captured the biphasic cell-free production of a model protein, chloramphenicol acetyltransferase. Flux variability analysis suggested that cell-free metabolism was potentially robust; for example, the rate of protein production could be met by flux through the glycolytic, pentose phosphate, or the Entner-Doudoroff pathways. Variation of the metabolite constraints revealed central carbon metabolites, specifically upper glycolysis, tricarboxylic acid (TCA) cycle, and pentose phosphate, to be the most effective at training a predictive model, while energy and amino acid measurements were less effective. Irrespective of the measurement set, the metabolic fluxes (for the most part) remained unidentifiable. These findings suggested dynamic constraint-based modeling could aid in the design of cell-free protein expression experiments for metabolite prediction, but the flux estimation problem remains challenging. Furthermore, while we modeled the cell-free production of only a single protein in this study, the sequence specific dynamic constraint-based modeling approach presented here could be extended to multi-protein synthetic circuits, RNA circuits or even small molecule production.