Abstract
Nucleocytoplasmic large DNA and Mirusviricota viruses exhibit taxonomic richness which continually expands due to metagenomic sequencing. Here we curate a database of giant virus metagenome-assembled genomes (GVMAGs V2), comprising 8,508 species-level representatives from 18,727 GVMAGs, a sixfold increase from the previous giant virus phylogenetic frameworks. Phylogenomics revealed 712 previously undescribed genera, 13 previously unknown viral families and an order we propose named Mycodnavirales. By accounting for alternative and custom genetic codes, we improved gene calling in over 1,300 GVMAGs, enabling more accurate identification of protein-coding genes. Database mining uncovered putative endogenous viral elements in hosts spanning algae, fungi and parasitic protists, highlighting that giant virus integration is widespread and evolutionarily persistent. Protein-level analysis revealed enriched genes for pollutant degradation in Algavirales and widespread biosynthetic gene clusters linked to antimicrobial-like and antibiotic resistance gene-like activity. This public resource will serve as a foundation for expanding giant virus diversity, uncovering virus-host interactions and exploring viral evolution.