Predicted structures for 2,800+ viral protein complexes go open, a third never seen before
Viral structural data now covers thousands of pathogens instead of a few well-studied ones; predictions await experimental validation, so watch how biologists actually use them.
Original event 2026-09-24
This is an archived article. Forecasts, prices and upcoming events refer to its original time period.
NVIDIA, together with Google DeepMind, EMBL-EBI and other partners, has added predicted 3D structures for the protein complexes of more than 2,800 viruses to the openly accessible AlphaFold Database, free for any researcher.
The structures were generated in bulk by AlphaFold2, optimized with NVIDIA BioNeMo, covering viral families known to infect humans from common-cold viruses to Mpox. About 30% of the protein interactions being added have never been documented in the Protein Data Bank, the main repository of experimentally determined structures.
Previously, outside a handful of well-studied viruses like the one behind COVID-19, most viruses had no public complex-structure data, yet such structures are often what vaccines and drugs must target. Note that all of the data is computational prediction labeled by confidence, not experimentally determined, and its real value depends on how researchers validate and use it.