Fig. 3: Improvements in sequencing a bacterial community using BOSS-RUNS.
From: Dynamic, adaptive sampling during nanopore sequencing using Bayesian experimental design
Four statistics highlight advantages of BOSS-RUNS (solid lines) compared to control (dashed lines) in an experiment lasting 600 minutes. a, Mean coverage depth over time. Coverage of the most abundant species is traded-off to collect more data from rarer species. As other genomes become resolved, a change in the rate of data accumulation is visible—for example, after ~180 minutes for B. subtilis. b, Reductions in the proportion of sites at <5× reveals that data are redistributed to areas of low coverage. c, Classifying sites as resolved if the posterior probability of one genotype is >0.99, we see that BOSS-RUNS achieves fewer unresolved sites owing to both sampling more data from rarer species and redistributing data within each genome. d, By focusing sequencing on sites with low coverage, BOSS-RUNS gives more even distributions of coverage. Note the different scales on the y axes to allow for sampling statistics of species of widely varying abundances.
