Fig. 2

Effects of sequential filtration on hospital-associated surface samples. a Bacteria proportion was not significantly increased after filtration according to paired t-tests. b Biomass of samples with and without filtration as well as retained by filters according to 16S rRNA gene copy number. In a and b, error bars represent the mean standard error of triplicates. Filter retention includes all biomass captured by 100, 80, 41, and 5 μm filters. Ns and ** are significance codes, representing p > 0.05 and 0.001 < p ≤ 0.01, respectively. A linear scale was used for both a and b because for a, a linear scale is more conservative than a log scale when no significant difference was concluded; for b, linear-scale biomass loss is more informative for metagenomic sequencing. c Principal coordinate analysis using Jaccard distance metric among samples with and without filtration. d Principal coordinate analysis based on Jaccard distance metric revealed that bacterial profiles retained on 5 μm filters clustered together with liquid samples, while those on 100, 80, and 41 μm filters were away from the major group