@Midnighter Thanks. I wanted a method to obtain a reproducible, unique, FBA solution. As you probably know, FBA problems can have multiple optima, and the optima returned depends on the specific solver used. This can also lead to nasty discontinuities if you try to plot the variation of the solution with respect to a parameter, because the solution returned jumps between the multiple admissible optima. The geometricFBA paper (10.1016/j.jtbi.2009.01.027) offers a solution to this problem. But I am open to suggestions. Are there other methods tackling this issue?