Traditional flux models of metabolic processes, including flux balance analysis and metabolic flux analysis, have been an important computational element of metabolic engineering since their development, nearly 25 years ago. Central to model development is the pseudo steady-state approximation argument, which allows all differential equations describing a metabolic process to be written as linear equations (forming a stoichiometric matrix and a vector of reaction fluxes). The resulting system of equations is then traditionally solved using linear programming. We will present developments in which non-linear metabolic processes are incorporated into a genome-scale model. For example, the selectivity of weak acid (acetate and butyrate) re-uptake by Clostridium acetobutylicum for the production of solvents (acetone and butanol) is known to be a non-linear function dependent on extracellular medium conditions. Real-coded genetic algorithms were applied to this problem to replace linear programming in solving the system of equations. More generally, we will show how genetic algorithms can be used with traditional stoichiometric matrices to locate singularities, which may then be identified as targets for further experimental study. Many singularities exist in the recently developed genome-scale model for C. acetobutylicum. Here, we also focus on a selected set of singularities of the stoichiometric matrix and the roles of thermodynamics calculations, available transcriptional data, and design of further experiments for developing resolutions.