How Open Science Can Help Researchers Prepare for the Next Pandemic
NVIDIA and DeepMind have released predicted 3D structures for protein complexes of over 2,800 viruses via the AlphaFold Database, using AlphaFold 2 optimized with NVIDIA BioNeMo Inference Runtime to scale inference to thousands of viral proteomes. The BioNeMo Structure Prediction Pipeline, a GPU‑accelerated workflow, is also open‑sourced, enabling researchers to generate structures from protein sequences themselves. About 30 % of the added protein interactions are novel, providing new hypotheses for vaccine and drug target discovery. This open dataset aims to democratize structural biology and improve preparedness for future pandemics.
⚡ Key Takeaways
- Predicted structures for protein complexes of more than 2,800 viruses are now publicly available.
- Inference was performed with AlphaFold 2 plus NVIDIA BioNeMo Inference Runtime for large‑scale protein‑complex prediction.
- Approximately 30 % of the added protein interactions are completely new to science.
- Researchers can use the open‑source BioNeMo Structure Prediction Pipeline to go from sequence to 3D structure on GPUs.
- The dataset is hosted in the AlphaFold Database, requiring no proprietary access.
🔧 Tools & Libraries
Engineers can leverage the released GPU‑accelerated pipeline and dataset to accelerate hypothesis generation for antiviral therapeutics, reducing reliance on costly experimental methods and shortening the time to target identification in emerging outbreaks.
✅ Practical Steps
- Clone the BioNeMo Structure Prediction Pipeline repository and run the provided GPU‑accelerated inference scripts on your protein sequences.
- Query the AlphaFold Database for viral complex structures to inform target selection in vaccine or drug design projects.
- Integrate the BioNeMo pipeline into your own HPC or cloud GPU workflow to predict structures for novel viral proteomes.
Want the full story? Read the original article.
Read on NVIDIA Blog ↗