As the global community confronts a roughly 50% chance of facing a pandemic as severe as COVID-19 by 2050 according to Center for Global Development estimates, NVIDIA, Google DeepMind, and EMBL-EBI have released predicted 3D structures for protein complexes of over 2,800 viruses on the AlphaFold Database to accelerate outbreak preparation worldwide.
When COVID-19 surfaced, researchers possessed a vital head start because decades of prior coronavirus studies mapped key viral proteins efficiently. Future pathogens may not offer similar foundational insights. To bridge this knowledge gap before an emergency strikes, major technology companies and international research institutions have united to stockpile structural biology data openly.
Scaling Viral Proteomes with AI and Accelerated Compute
The newly published dataset maps the complexes and interacting proteins encoded within thousands of viral proteomes. Generating this scale of structural data required heavy computational muscle. The project team inferred these 3D structures using AlphaFold2, Google DeepMind’s protein-folding AI model, optimized specifically with the NVIDIA BioNeMo Inference Runtime.
Traditional experimental methods—such as crystallizing proteins and shooting X-rays at them—demand years of painstaking lab work and thousands of dollars per single structure. By contrast, running AlphaFold2 on NVIDIA GPUs yields complex predictions in minutes. This leap in throughput allows researchers to evaluate entire viral families at once, scanning organisms ranging from common-cold viruses to emerging threats like Mpox.
To support independent research, NVIDIA is also openly releasing the BioNeMo Structure Prediction Pipeline. This GPU-accelerated workflow enables biologists to move directly from a raw protein sequence to a predicted 3D structure for their own targets.
Uncharted Biological Territory and Hypothesis Generation
Proteins rarely operate in isolation. They form intricate multi-molecule complexes that execute sophisticated cellular functions, and those precise configurations typically serve as the primary targets for vaccines and antiviral drugs. By mapping these structures ahead of time, the global scientific community gains an immediate arsenal for diagnostic and therapeutic development.
Approximately 30% of the protein interactions added to the database are entirely new to science. These interaction shapes have never appeared in the Protein Data Bank, the primary repository for experimentally determined protein structures.
“This database is an engine for hypothesis generation. We’re enabling biologists and the AI community to investigate protein interactions, not just as single molecules but as complexes, so the whole field can move forward,” said Chris Dallago, applied research science team lead in digital biology at NVIDIA.
This structural injection acts as a vital tool for the next generation of scientists. As Joe Grove, professor of molecular virology at the Medical Research Council-University of Glasgow Centre for Virus Research and a collaborator on the project, noted of his own doctoral work: “When I did my Ph.D., there were no structures for any of the proteins we were investigating. It was like working in the dark — we had to guess what was going on.” He added that the new dataset provides current doctoral researchers with high-quality structural data to accelerate foundational science.
A Global Collaborative Framework
Unveiled during a United Nations General Assembly meeting convened by the World Economic Forum in New York City focusing on pandemic prevention, preparedness, and response, the dataset expands the AlphaFold Database to more than 260 million protein and protein complex predictions covering nearly every cataloged protein known to science.
The collaborative network behind the release includes:
- Coalition for Epidemic Preparedness Innovations
- European Molecular Biology Laboratory’s European Bioinformatics Institute (EMBL-EBI)
- Google DeepMind
- NVIDIA
- Seoul National University
- Sungkyunkwan University
- Swiss Institute of Bioinformatics
- University of Glasgow
Open access remains a core pillar of the initiative. By removing technical and financial barriers, the project assists researchers operating in low-resource settings who frequently confront localized outbreaks on the front lines.
“Making this data open is critical for understanding viral diagnostics and developing treatments and vaccines. The dataset also covers lesser-studied viruses and lowers the barriers for scientists in low-resource settings who are confronting outbreaks firsthand,” said Jo McEntyre, interim director of EMBL-EBI.
As Risha Patel, life sciences partnerships manager at Google DeepMind, emphasized: “Our ambition with the AlphaFold Database has always been to democratize access to foundational biology at scale. This collaboration to bring thousands of viral complexes into the database will equip scientists around the world with insights they need to help prepare for future outbreaks.”
Researchers can access the viral protein complex dataset directly through the AlphaFold Database Pandemic Preparedness Portal and utilize the BioNeMo Structure Prediction Pipeline to generate structural forecasts for custom targets.
