NVIDIA and DeepMind Open Predicted Protein Structures for 2,800+ Viruses

On September 24, NVIDIA announced a collaboration with Google DeepMind, EMBL-EBI, and multiple research institutions to release a new dataset into the AlphaFold database, offering free access to predicted 3D structures of protein complexes from more than 2800 viruses, aimed at helping scientists develop vaccines and therapies ahead of potential outbreaks.

Confirmed

• Partners include NVIDIA Health, Google DeepMind, EMBL-EBI, CEPI, Seoul National University, and the University of Glasgow, among others.
• The structures were predicted by AlphaFold2 and generated with optimization via the NVIDIA BioNeMo inference runtime.
• The dataset covers predicted protein complex structures from over 2800 viruses and is now publicly available, free of charge, in the AlphaFold database.
• The data is intended to support vaccine and therapy development as well as pandemic preparedness.

Why it matters

• Viral protein structures are foundational information for vaccine and drug design, yet most viruses previously lacked experimentally determined structures; making large-scale AI predictions publicly available can significantly shorten researchers' response time to emerging viruses.
• The project also marks a landmark partnership combining NVIDIA BioNeMo's inference optimization with the DeepMind AlphaFold ecosystem, exemplifying coordination between industry and public research institutions in pandemic defense.

2026-09-24 ~ 2026-09-24 · 6 related posts

• [source] NVIDIA, DeepMind and EMBL-EBI Open 3D Structures for 2,800+ Viral Proteomes — nordicinst · 2026-09-24
• [source] NVIDIA and DeepMind open up AI-predicted protein structures for 2,800+ viruses — nvidia · 2026-09-24
• AlphaFold Database expands to 2,800+ virus protein structures for vaccine research — AllThingsApx · 2026-09-24

3 near-duplicate retellings: nvidia · AllThingsApx · AllThingsApx