NVIDIA Releases Open Viral Protein Dataset and BioNeMo Pipeline

NVIDIA and Google DeepMind have collaborated to release a dataset of AI-predicted protein structures for over 2,800 viruses, including those known to infect humans. The predictions are available through the AlphaFold Database’s Pandemic Preparedness Portal, which was developed in collaboration with EMBL-EBI. NVIDIA has also released the BioNeMo Structure Prediction Pipeline to help researchers generate 3D protein predictions for other protein complexes.

The AI predictions can provide a starting point for researchers studying unfamiliar viruses, although experiments are still needed to confirm the predictions. Joe Grove, a professor of molecular virology, emphasized the importance of having this knowledge ahead of time in case of future pandemics. The dataset has added about 30% of protein interactions that were not previously documented in the Protein Data Bank.

While the AI model, AlphaFold2, is not capable of confirming the predicted structures, it significantly reduces the time and financial cost associated with predicting 3D protein structures, thereby accelerating the response time for emerging diseases.