(A) The tri-culture dynamics is simulated with different lag-phase values for Faecalibacterium prausnitzii A2-165 (FP) and Roseburia intestinalis L1-82 (RI) and the resulting end point abundance ratio of FP and RI is plotted in a heat map that is colored in blue for FP dominance and in red for RI dominance. The observed tri-culture data (black circles) are plotted according to the estimated experimental lag phases for RI and FP. The predicted RI or FP dominance agrees with the observed dominance in all six cases. (B, C) Simulations illustrate the dependency of the end point abundances (Xi(te)) of the three strains on the lag phase of RI and FP. (D) The tri-culture dynamics is simulated for varying initial abundances (init. abund.) of FP and RI and their resulting end point abundance ratio is visualized in a heat map. Three of the four FP-dominated experiments (13–15) and both RI-dominated experiments (10 and 11) are situated within their predicted region of dominance. (E, F) The end point abundance Xi(te) of the three strains is non-linearly dependent on the initial abundance of RI and FP in simulations, illustrating that dominance in batch is sensitive to initial conditions. All simulations were carried out with the model parameterized on mono- and bi-culture data (parameterization 2). Initial abundances are plotted in logarithmic scale. For the simulations in (A–C), the initial abundances of RI, FP and Blautia hydrogenotrophica S5a33 (BH) were set to 0.58, 0.04 and 0.21 108 counts/mL, respectively, whereas for the simulations in (D–F), the lag phase for RI, FP and BH were set to 0.33, 0.08 and 0.1 h, respectively. These initial abundance and lag-phase values represent the averages of observed initial abundances and estimated lag phases across all tri-culture experiments.