Fig. 3

Comparison of computation times of the FLB MILP and the NB MILP when used either with the condensed (“no split”) MILP variant [Eqs. (14) and (16), respectively] or the (“rev. split”) MILP variant with separated and split reversible reactions [Eqs. (14a) and (16a), respectively]. a, d “no split” versus “rev. split” in FLB for MCS for growth coupling (a) and for synthetic lethals (d); b, e “no split” versus “rev. split” in NB for MCS for growth coupling (b) and for synthetic lethals (e); c, f: comparison of “rev. split” variants for FLB and NB for MCS for growth coupling (c) and for synthetic lethals (f). Each dot represents one particular MCS enumeration scenario (product/organism/seed combination in a–c and growth rate threshold/seed combination in d, e and the dot color marks the respective model (organism) in which the computation has been conducted: green iJM658; red: iMM904; black: iJO1366. The total number of comparable enumeration scenarios as well as the cumulative sum of the computation times over all scenarios for the MILP variants are given below the diagram. The percentage at the axes quantifies the relative frequency with which the respective MILP variant was faster than the other MILP variant