Abstract
Although plasmids are important for bacterial survival and adaptation, plasmid detection and assembly from genomic, let alone metagenomic, samples remain challenging. The recently developed plasmidSPAdes assembler addressed some of these challenges in the case of isolate genomes but stopped short of detecting plasmids in metagenomic assemblies, an untapped source of yet to be discovered plasmids. We present the metaplasmidSPAdes tool for plasmid assembly in metagenomic datasets that reduced the false positive rate of plasmid detection as compared to the state-of-the-art approaches. We assembled plasmids in diverse datasets and have demonstrated that thousands of plasmids remained below the radar in already completed genomic and metagenomic studies. Our analysis revealed the extreme variability of plasmids and has led to the discovery of a high number of novel plasmids (including many plasmids carrying antibiotic-resistance genes) without significant similarities to currently known ones.