Taxonomy bias in metagenome-assembled genome recovery

The recovery of metagenome-assembled genomes (MAGs) from shotgun metagenomic sequencing is rapidly expanding the availability of representative genomes. However, this practice may skew the representation of specific taxa in real-world datasets. This bias is attributed primarily to the known inefficiencies of sequence-by-synthesis platforms in amplifying GC-rich and AT-rich sequence fragments. Here, we recover 216 medium- and high-quality MAGs from an Australian wetland site. Notably, no MAGs were recovered for some dominant cyanobacterial and proteobacterial species known to be present. A new protocol involving read-based classification and alignment to the MAG dataset demonstrated the highly efficient recovery of low-GC organisms in the Actinobacteria and Bacteroidota phyla. Additionally, the recovery of lost taxonomic information was demonstrated through unmatched sample mapping. The findings suggest a bias towards the recovery of smaller, low-GC organisms in MAG recovery, potentially skewing the global representation of microbial diversity. Our pipeline is made publicly available as a tool to help researchers estimate taxonomic losses following MAG recovery efforts.