Rachel Seongeun Kim

BFVD: A Large Repository of Predicted Viral Protein Structures @NAR_Open • The Big Fantastic Virus Database (BFVD) is the largest repository of predicted viral protein structures, containing 351,242 models derived from viral sequence representatives in UniRef30 clusters using ColabFold and AlphaFold2. • BFVD enhances the quality of predictions for 36% of its structures by integrating homologs mined from over two petabases of sequence data in Logan, resulting in deeper alignments and improved structure confidence. • The database reveals unprecedented structural diversity, with 62% of entries showing low or no similarity to existing repositories like AFDB or PDB, highlighting its importance for studying viral protein functions and evolution. • BFVD supports bacteriophage research by enabling structural annotation of proteins previously classified as “hypothetical,” showcasing its utility in decoding unannotated metagenomic data. • The web server (bfvd.foldseek.com) provides tools for structural searches, visualization, and taxonomic exploration, integrating UniProt labels and clustering data for ease of access and enhanced usability. • By focusing on viral proteins excluded from major databases like AFDB, BFVD offers a compact but comprehensive resource tailored for advancing viral research and supporting structural virology. • Future expansions aim to include viral multimer structures and enhance search capabilities with tools like Foldseek-Multimer, ensuring continued contributions to virology and computational biology. @thesteinegger @RayanChikhi @milot_mirdita @eunbelivable 💻Code: github.com/steineggerlab/bfv… 📜Paper: doi.org/10.1093/nar/gkae1119 #ViralProteins #StructuralBiology #Bioinformatics #AlphaFold2 #Virology #OpenScience