NVIDIA and Google DeepMind Release 3D Protein Structures for 2,800+ Viruses

Open Access to Viral Protein Structures

NVIDIA, in collaboration with Google DeepMind and the European Molecular Biology Laboratory’s European Bioinformatics Institute (EMBL-EBI), has released predicted 3D structures for protein complexes from over 2,800 viruses. These structures are now openly accessible to scientists globally through the AlphaFold Database.

AI-Powered Prediction and Optimization

The dataset was generated using AlphaFold2, Google DeepMind’s AI model for predicting protein 3D shapes. NVIDIA BioNeMo Inference Runtime optimized this process, enabling the team to scale inference across thousands of viral proteomes and predict interacting protein groups within each virus.

Democratizing Biological Insights

This collaboration expands the AlphaFold Database, providing researchers with insights crucial for preparing for future outbreaks. Risha Patel, life sciences partnerships manager at Google DeepMind, stated that the goal is to democratize access to foundational biology at scale.

New Discoveries and Open-Sourced Tools

Approximately 30% of the newly added protein interactions are novel, displaying interaction shapes not previously documented in the Protein Data Bank. This offers new avenues for biological research. Additionally, NVIDIA has open-sourced its BioNeMo Structure Prediction Pipeline, the GPU-accelerated workflow used to create the dataset, allowing researchers to generate their own 3D protein structure predictions.

Proactive Pandemic Preparedness

This effort is a proactive measure against future pandemics. Joe Grove, professor of molecular virology at the Medical Research Council-University of Glasgow Centre for Virus Research, highlighted the importance of stockpiling knowledge ahead of time, especially for unknown pathogens that may emerge without prior research.