Fig. 4: Estimates of engraftment in FMT triads.

a, The inferred number of plausible transmission pairs (shared strains), inferred by each algorithm, between samples taken as part of study investigating the impact of FMTs. Transmission between samples from unrelated participants (shown in red) are highly likely to be false positives. By contrast, samples from the same participant following FMT (shown in black) are more likely to be true positives, with an error rate proportional to that observed in the unrelated participant results. Species-specific SNP thresholds for identifying recent transmission were inferred using the mixture distribution approach (Methods). An analogous plot using the Youden method, as described by Valles-Colomer et al.⁴, is given in Extended Data Fig. 6. b, Histograms indicating the distribution of pairwise SNP distances (truncated at 500 SNPs) between donor and recipient (post-transplant) samples for major bifidobacterial species. Vertical black lines indicate the SNP thresholds inferred using the TRACS mixture distribution method. InStrain identified the transmission of B. infantis, which is probably B. longum but mislabelled in the UHGG genome collection database, whereas StrainGE identified no transmission of B. longum. c, An example of multiple strains of B. longum being transmitted between a single donor and multiple recipients was detected exclusively by TRACS. The allele frequencies at polymorphic loci within a segment of the B. longum reference genome are shown for each recipient sample, with colours indicating the two distinct strains. The red strain is dominant in recipient sample SFMT_03_t15 but in the minority in a separate recipient (SFMT_27_t15) who received the same donor stool.