Constraint-based modeling has been shown, in many instances, to be useful for metabolic engineering by allowing the prediction of the metabolic phenotype resulting from genetic manipulations. But the basic premise of constraint-based modeling-that of applying constraints to preclude certain behaviors-only makes sense for certain genetic manipulations (such as knockouts and knockdowns). In particular, when genes (such as those associated with a heterologous pathway) are introduced under artificial control, it is unclear how to predict the correct behavior. In this paper, we introduce a modeling method that we call proportional flux forcing (PFF) to model artificially induced enzymatic genes. The model modifications introduced by PFF can be transformed into a set of simple mass balance constraints, which allows computational methods for strain optimization based on flux balance analysis (FBA) to be utilized. We applied PFF to the metabolic engineering of Escherichia coli (E. coli) for free fatty acid (FFA) production-a metabolic engineering problem that has attracted significant attention because FFAs are a precursor to liquid transportation fuels such as biodiesel and biogasoline. We show that PFF used in conjunction with FBA-based computational strain optimization methods can yield non-obvious genetic manipulation strategies that significantly increase FFA production in E. coli. The two mutant strains constructed and successfully tested in this work had peak fatty acid (FA) yields of 0.050 g FA/g carbon source (17.4% theoretical yield) and 0.035 g FA/g carbon source (12.3% theoretical yield) when they were grown using a mixed carbon source of glucose and casamino acids in a ratio of 2-to-1. These yields represent increases of 5.4- and 3.8-fold, respectively, over the baseline strain. (C) 2014 Wiley Periodicals, Inc.