Figure 2: Automated binning of contigs and population-level genome characteristics.

a, Example (sample M01.2-V1) of a two-dimensional visualization of pentamer frequency-based signatures of assembled contigs; distinct bins are highlighted based on completeness (only bins with at least 29% of essential genes with fewer than 20% multiple copies are labelled). b, Metagenomic coverage of genes in the cluster encircled in a, which was subdivided in the clustering process based on metagenomic depths of coverage of essential genes. The red dashed line indicates the determined cutoff between bins. c, Colour key for subpanels a and d, with a histogram (black lines) of completeness of all binned population-level genomes with at least one essential gene. d, Metagenomic mOTU abundances calculated from mapping reads against a collection of phylogenetic marker genes compared to a calculation based on metagenomic depth of coverage of binned population-level genomes annotated with the taxonomy of the most similar phylogenetic marker genes. Dots representing mOTUs with reconstructed genomes are coloured based on the genomes' completenesses; others are indicated as blue, open circles. e, Example of genes of a common functional category (here K03149, thiazole synthase) linked to population-level genomes with taxonomic annotations. The taxonomic annotation of the population-level genomes recruiting >10% of the overall metagenomic reads mapping to genes with the function of interest are indicated; all other genes are gathered in ‘others/unknown’. Groups of two or three bars represent one sample (sample names are indicated in the colour scheme defined in Fig. 1a; visit (V)1, first sample; V2, second sample; V3, third sample), with the first bar representing the origins of the metagenomic reads, the middle bar representing the origins of the metatranscriptomic reads, and the (optional) last bar representing the origins of the uniquely identified proteins.