Overcoming circularity bias reveals a hidden linear DNA and RNA virosphere

Abstract Metagenomic sequencing has transformed our view of the global viral diversity, yet a fundamental bias limits our understanding of the diversity of complete viral genomes. Current quality assessment tools rely on identifying circular contigs, systematically underestimating the completeness of linear DNA and RNA viruses, thus leaving a large fraction of the virosphere uncharacterized. Here, we overcome this bias, by leveraging within-species genomic termini conservation across diverse metagenomic datasets. Applying this approach to the metaVR database, we recover and validate 1,585 conserved, complete linear viral genomes, representing a 21-fold increase in recognized linear completeness. Our results reveal previously uncovered complete monopartite RNA genomes and individual segments of multipartite viruses across 13 classes, alongside deeply branching viral dark matter. Furthermore, we expand the known physical boundaries of linear DNA viruses, by recovering intact megaphages up to 220 kb and distinguishing true complete genomes from highly transcribed operons. Ecological profiling reveals a strong enrichment of these reassigned as complete linear viruses in specific ecosystems, highlighting their ecological relevance. By enabling systematic recovery of complete linear genomes, this framework unlocks a previously hidden fraction of the virosphere and reshapes our understanding of viral diversity and evolution.