Wohlwend et al. |
Boltz-2
- Run
- About
- API Example
More
Step 1: Upload your data
Input MSA file - optional
Drag your file(s) or upload
- Your file can be in the following formats:a3m
The input MSA file is a file format used to store information about the multiple sequence alignment of a protein. It is OPTIONAL, if you do not provide it, the Use Msa Server checkbox will be set to True automatically. Please READ the documentation for proper usage.
or
Don’t have a file?
Use our demo data to run
Use Demo Data
Input batch file - optional
Drag your file(s) or upload
- Your file can be in the following formats:csv
A batch file allows you to upload multiple parameters at once using CSV format. Download the template file and modify it for your specific requirements. Keep the first column (Sample) consistent with entries such as Sample1, Sample2, Sample3... to identify different experiments. IMPORTANT: If you upload a batch file, you don't need to fill in the parameter sections below.
or
Don’t have a file?
Use our demo data to run
Use Demo Data
Templates - optional
Drag your file(s) or upload
- Your file can be in the following formats:cif
A flexible, text-based format used to store crystallographic and structural biology data. The mmCIF (macromolecular CIF) extension is the modern standard for representing 3D macromolecular structures in the Protein Data Bank. It uses a self-describing, table-like syntax that supports high precision coordinates, unlimited atoms/chains, and rich metadata about experimental conditions, symmetry, and validation — making it more robust and machine-readable than the older PDB format.
or
Don’t have a file?
Use our demo data to run
Use Demo Data
Step 2: Set parameters
Your structures to fold will appear here
Optional parameters
- Pocket restraints
- Covalent bond
Your constraints will appear here
Note - constraints are optional
Your templates will appear here
Note - templates are optional
Step 3: Complete run profile
Short Description
Boltz-2 is an interactive web application that lets researchers simulate, visualize, and score biomolecular interactions at proteome scale in minutes. It integrates a next-generation Boltz diffusion model with physics-aware scoring to prioritize high-affinity complexes. The interface streamlines hypothesis testing, from target selection through in-silico validation.
Example use case
Predicting affinities (Kd) of new small molecule drug candidates to their cognate protein or protein-NA complex.
Technology
- Boltz diffusion backbone fine-tuned on protein–ligand pairs
- Hybrid SE(3)-equivariant graph neural networks for pose refinement
Limitations
- Boltz diffusion backbone fine-tuned on protein–ligand pairs
- Hybrid SE(3)-equivariant graph neural networks for pose refinement
Metrics
- Pre-trained model adapted from Boltz-1 by fine-tuning on 30M protein-ligand pairs.
Citation:
Saro Passaro, Gabriele Corso, Jeremy Wohlwend, Mateo Reveiz, Stephan Thaler, Vignesh Ram Somnath, Noah Getz, Tally Portnoi, Julien Roy, Hannes Stark, David Kwabi-Addo, Dominique Beaini, Tommi Jaakkola, Regina Barzilay. BioRxiv 2025.06.14.659707; doi: https://doi.org/10.1101/2025.06.14.659707
Released:
Aug-01-2025
Aug-01-2025