A method for identifying metagenomic species and variable genetic elements by exhaustive co-abundance binning

The microbial diversity of environments like the human gut extends far beyond what is covered by reference genomes. Here we present a method for exhaustive and unsupervised co-abundance gene binning across a series of highly complex metagenomic samples. As the method does not rely on previously sequenced reference genomes it allows for discovery of new species, viruses and clonal heterogeneity. We demonstrate the method on human gut microbiome data and identify 7,381 co-abundance gene groups (CAGs) ranging in size from 3 to 6,319 genes. The CAGs represent a wide variety of biological entities including microbial genomes, phages and clonal differences. We name the 741 largest of these metagenomic species (MGS), because they correspond to microbial species. In addition, we establish microbial host affiliations by dependency-associations for many small CAGs. Longitudinal sampling in the same individuals indicates that some of these dependency-associations are important for the persistence of their microbial host