Fig. 3

PSOMCS small example. Running the PSOMCS on a toy network. This network has three input reactions, which can be assumed to be substrates and three secretion reactions, which can be assumed to be three different products. We want to maximise the yield of R4, that is maximize (R4/(R1 + R2 + R3)). Note that the particles operate in a single dimensional search space and x represents the yield for R4. After performing FBA to determine the maximum and minimum yields for R4 given unit substrate uptake, four particles are initialised within this range. Initial velocities are also assigned. cMCSs are calculated after creating and solving the dual system. Fitness is a function of x and the cardinality of the cMCS. g corresponds to x with the highest fitness which is particle 4 after both the first and second iterations. After the first iteration, every particle except the first has a value for p. Note that for particle 4 a yield higher than 0.98 is guaranteed. In reality, the minimal yield with the corresponding cMCS is 1, which is also the case for particle 1. This is the value the algorithm will return if allowed to run for a few more iterations