Extended Data Fig. 7: Bacterial loss accross independent datasets and in vivo proof-of-concept.

Metagenomics analysis using the MetaPhlAn4 pipeline of absolute lost SGBs in non-responders compared to responders in A. MIMIC (n = 32 samples from n = 12 patients), B. Baruch et al., (n = 30 samples from n = 10 patients) C. Davar et al. (n = 168 samples from n = 15 patients), p-values for panel A-C computed by Wilcoxon test (two-sided). D. Metagenomics analysis using the BiomScope pipeline, for each patient, species were categorized according to their presence status in the patient at baseline and their donor: “only patient” category is shown, which represent the species present in patient at baseline and not in the donor. For each species, the retention was estimated as the proportion of patients where the species is still present at the considered timepoint. Retention across species is shown over time. Pooled data from MIMIC, FMT-LUMINate, Baruch et al., and Davar et al., 95% confidence intervals for median shown; p-values were calculated using Mann-Whitney-Wilcoxon test (two-sided), data from n = 250 samples from n = 64 patients. E, F. Tumor kinetics for n = 10 mice per group, results from three independent experiments, p-value calculated by Non-parametric Mann–Whitney Wilcoxon test (two-tailed), data are presented as mean values ± SEM. For box plots (A–C), the center line represents the median, box bounds represent the interquartile range (IQR), whiskers extend to 1.5 x IQR, and each dot represents an individual patient sample. R; responder. NR; non-responder. C; cocktail.