NVIDIA has partnered with Google DeepMind and the European Molecular Biology Laboratory’s European Bioinformatics Institute (EMBL-EBI) to publish 3D structure predictions for protein complexes across more than 2,800 viruses. The dataset is freely available to the public through the AlphaFold Database and aims to accelerate global pandemic preparedness and therapeutic discovery.
The viral protein complexes were predicted using DeepMind’s AlphaFold2 model running on the NVIDIA BioNeMo Inference Runtime, which optimized GPU inference across thousands of viral proteomes. Approximately 30% of the newly added protein interactions are entirely new to scientific literature, offering researchers structural data on multi-protein interactions that previously required years and significant capital to determine through laboratory crystallization.
In tandem with the dataset, NVIDIA openly released its BioNeMo Structure Prediction Pipeline. The GPU-accelerated workflow allows researchers to run inference on custom protein sequences, enabling computational biology teams worldwide to scale their own molecular pipeline research.
Why it matters
Provides drug discovery teams open access to thousands of novel viral targets without high experimental costs.
Demonstrates how GPU-optimized inference pipelines can drastically accelerate scientific workflows and foundational biological research.
Offers open-source infrastructure tools for biotech startups to integrate structural predictions into proprietary pipelines.
Source: blogs.nvidia.com



