Fig. 10

Contaminant features in a set of BALF samples. a Taxa representing more than 1% of the total set of BALF samples, coloured by whether they are negatively correlated (red), positively correlated (green), or not correlated (blue) with input sample bacterial density. The shaded area represents the locally weighted scatterplot smoothing (LOWESS). b. The heatmap of relative abundance z-score for OTUs which were correlated with samples input density. Positively correlated OTUs (putative true positives) are represented with a gradient of green while negatively correlated OTUs (suspected contaminants) are represented with a gradient of red. c Reproducibility of OTUs in BALF samples plotted by sample density and log relative abundance. d Effect of post-sequencing filtering on sequencing accuracy and precision of a dilution series of a single BALF sample, comparing dilutions to the composition of the highest density sample