Skip to navigation

Boltz-2 Documentation

View as MarkdownOpen in Claude

The Boltz-2 NIM predicts biomolecular structure and binding affinity for combinations of proteins, RNA, DNA, and other molecules. Based on the Boltz-2 architecture, this NIM predicts complex biomolecular structures and quantifies molecular interactions across a wide range of molecular configurations and binding studies.

Boltz-2 represents a significant advancement in computational biology, offering unprecedented capabilities for predicting:

  • Protein structures: Single and multi-chain protein complexes
  • Nucleic acid structures: DNA and RNA molecules in various configurations
  • Protein-nucleic acid complexes: Interactions between proteins and genetic material
  • Ligand binding and affinity prediction: Small molecule interactions with biomolecules, including predicted binding affinity scores
  • Modified residues: Post-translational modifications and chemical modifications
  • Constraint-guided predictions: Structure predictions conditioned on specified interaction pockets or contacts

The model supports both single molecule predictions and complex multi-molecular assemblies, making it suitable for a wide range of research applications from basic structural biology to drug discovery.

Skip Ahead

Use the following cards to navigate the Boltz-2 documentation.

Use Cases

Explore the common use cases for the Boltz-2 NIM.

For detailed technical information about the Boltz architecture and training methodology, refer to the following technical reports:

  • Boltz-2: Passaro et al. 2025 - “Boltz-2: Towards Accurate and Efficient Binding Affinity Prediction”
  • Boltz-1: Wohlwend et al. 2024 - “Boltz-1: Democratizing Biomolecular Interaction Modeling”