Figure 4. Replicate pairs reveal levels of biological noise associated with repeated sampling.

(A, B) We analyzed the distribution of ΔAF for all SNPs detected across all replicate pairs $(m=62)$ and longitudinal pairs $(n=200)$ for SNPs where ΔAF ≥25%. (B) SNPs were detectable at lower levels of ΔAF for both types of isolate pairs, but SNPs with higher values of ΔAF were only found in longitudinal pairs. (C) To determine a ΔAF threshold for calling SNPs representative of changes in bacterial population composition in-host, we calculated the average number of SNPs per pair of isolates at different ΔAF thresholds for both replicate and longitudinal pairs. At a ΔAF threshold of 70% the number of SNPs between replicate pairs represents 5.27% of the SNPs detected amongst all replicate and longitudinal pairs, weighted by the number of pairs in each group.