Overview
The project supplies code, weights and command-line inference from YAML inputs. Boltz-1 and Boltz-2 remain one family entry rather than separate version duplicates.
Limitations
The CPU route is slower according to the README. Some Boltz-2 training and evaluation code is still listed as forthcoming. Predictions require experimental assessment.
Key Features
- Biomolecular complex prediction
- Boltz-2 affinity prediction
Use Cases
- Explore structural hypotheses
- Evaluate model predictions for candidate molecules
How to Use
Install in a separate Python environment, prepare YAML inputs and follow the prediction guide for model and MSA settings. Retain inputs and checkpoint versions.