The original publication (Atarashi et al., 2013) reported on 17 immune-modulating bacteria. In Bucci et al. (2016), a dynamical ecological model for 13 of the original 17 strains was presented. These 13 strains were selected because they do not harbor antibiotic resistance genes. For our study, we used a 12-strain subset of these 13 strains based on ecological stability considerations. Colonization and CD4+FOXP3+Treg data for the three 11-strain sets, measured by qPCR and FACS, were used along with the simulated mono-colonization concentrations (Figure 2) and included into the microbiome–immune system model (Figure 1). The resulting parameters were employed to predict ecologically stable subsets with different Treg induction potentials. 4-strain subsets were selected to validate the predictions of the mathematical model (see main text and Figure 3). Strain 1: Clostridium saccharogumia/Clostridium ramosum JCM1298, Strain 3: Flavonibacter plautii/Pseudoflavonifractor capillosus, Strain 4: Clostridium hathewayi/Clostridium saccharolyticum WM1, Strain 6: Blautia coccoides/Lachnospiraceae bacterium, Strain 7: Clostridium bolteae, Strain 8: Clostridium sp. MLG055/Erysipelotrichaceae bacterium 2 44A, Strain 9: Clostridium indolis/Anaerostipes caccae DSM 14662, Strain 13: Anaerotruncus colihominis, Strain 14: Ruminococcus sp. ID8/Lachnospiraceae bacterium 2 1 46FAA, Strain 15: Clostridium asparagiforme/Clostridium lavalense, Strain 16: Clostridium symbiosum, Strain 18: Clostridium ramosum, Strain 21: Eubacterium fissicatena/Eubacterium contortum/Clostridium sp. D5, Strain 26: Clostridium scindens/Lachnospiraceae bacterium 5 1 57FAA, Strain 27: Lacnospiraceae bacterium A4/Lachno bacterium 3 1 57FAA CT1, Strain 28: Clostridium sp. 316002/08, Strain 29: Lacnospiraceae bacterium A4/Lachno bacterium 3 1 57FAA CT1.