Fig. 3

Genome properties and filter size fraction distributions of MAGs. a Principal components analysis (PCA) on our 30 MAG clusters and 135 genomes from marine isolates [4] based on log-transformed percentages of non-coding DNA, GC content, COG categories "transcription" (K), "signal transduction" (T), "defense mechanism" (V), "secondary metabolites biosynthesis" (Q) and "lipid transport and metabolism" (I). Only isolates belonging to the phyla (classes for proteobacteria) represented by MAGs were included. b, c Genome size versus percentage of non-coding DNA plotted for the same set of genomes (b), with a zoom-in on the smaller and denser genomes (c). d, e Number of sequence reads matching to the MAG clusters (in log scale) per 10,000 reads compared from each filter fraction in Dupont et al. [13]. f The ratio of matches between the 0.8–3.0 and the 0.1–0.8 μm fraction versus genome size, both in log scale