[ Co-Folding a Complex | Boltz conformations | Run Confidence Model | Run Affinity Model | Force Weights Update | Run on GPU | Local boltz diffusion | Boltz reference ]
The Boltz tools in ICM provide AI-based prediction of biomolecular complexes using the Boltz-2 diffusion model. Boltz-2 can be used to predict protein structures and biomolecular complexes, including protein-ligand complexes.
For more information about Boltz-2 see:
In ICM, Boltz can be used for co-folding, where one or more protein sequences and an optional ligand are modeled together, or for Local Boltz Diffusion docking, where an existing receptor structure is used as the starting point for locally flexible docking.
The co-folding tools are available from:
Homology -> Co-Folding Boltz Model
At least one protein sequence must first be loaded into ICM. If no sequence is loaded, the Co-Folding Boltz Model menu option will be greyed out.
Local Boltz Diffusion docking is available from:
Docking -> Flexible Receptor -> Local Boltz Diffusion
A previously prepared ICM docking project must be opened before starting Local Boltz Diffusion docking.
Co-folding predicts a molecular complex from the supplied protein sequence information and ligand structure.
One or multiple protein sequences can be included in the same prediction. This enables Boltz to model multi-chain protein complexes as well as protein-ligand complexes.
To perform a co-folding calculation:
- Load the required protein sequence or sequences into ICM.
- Select Homology -> Co-Folding Boltz Model.
- Select the sequence or sequences to include in the prediction.
- Select a ligand if modeling a protein-ligand complex.
- Set N conformations to the number of independent predictions to generate.
- Enable Run Confidence Model and Run Affinity Model if these predictions are required.
- Leave Run on GPU selected to use GPU acceleration, or clear it to run on the CPU.
- If updated model files are required, enable Force Weights Update.
- Click OK to start the prediction.
The co-folding dialog contains the following options:
Sequence: Selects the protein sequence or sequences to include in the prediction.
Ligand: Selects the ligand to include in the co-folding calculation. The ligand can be supplied as a chemical structure, for example from a SMILES representation loaded into ICM.
N conformations: Specifies the number of independent Boltz diffusion predictions to generate. Setting this value to 0 instead stores the intermediate structures from a single diffusion run.
Run Confidence Model: Calculates pLDDT confidence values for the resulting prediction.
Run Affinity Model: Calculates predicted binding free energies in kcal/mol for the resulting conformations.
Force Weights Update: Downloads or refreshes the Boltz model weights and required modules.
Run on GPU: Uses GPU acceleration when selected. This option is selected by default. Clear the checkbox to run on the CPU.
The resulting model contains the predicted protein structure or protein complex together with the predicted ligand pose when a ligand is included. Secondary structure is assigned automatically.
Boltz co-folding is not restricted to a single protein chain.
Multiple protein sequences can be supplied in the same calculation, allowing the relative arrangement of the chains to be predicted simultaneously.
A ligand can also be included so that protein-protein interactions and ligand binding are considered together during the co-folding calculation.
The workflow can therefore be used for single-chain protein-ligand complexes, multi-chain protein complexes, and multi-chain protein-ligand complexes.
The N conformations setting controls the number of independent Boltz diffusion predictions that are generated.
When N conformations is greater than 0, Boltz performs the specified number of independent diffusion runs. The resulting ICM stack contains the final minimized conformation from each independent run.
For example:
N conformations = 5
produces five independently generated and minimized predictions.
The conformations in this stack are independent predictions and should not be played or interpreted as a trajectory. Each frame represents the final result of a separate diffusion run rather than a successive stage of the same calculation.
Differences in protein orientation, ligand orientation, or overall structure between frames are therefore expected.
The diffusion and denoising process can be visualized by setting:
N conformations = 0
In this mode, instead of storing independent final predictions, ICM stores the intermediate structures from a single diffusion run.
The resulting stack contains approximately 200 diffusion steps showing the structure progressively changing from the initial noisy state toward the final denoised prediction.
This stack can be played to visualize the Boltz diffusion and denoising process. It represents the stages of a prediction, not a molecular dynamics trajectory.
Enable Run Confidence Model to calculate pLDDT (Predicted Local Distance Difference Test) confidence values for the predicted structure.
Confidence values are calculated per residue for proteins and per atom for ligands.
The pLDDT values are scaled from 0 to 1. Higher values indicate greater model confidence in the predicted local structure and atomic placement.
The values are stored in the atom B-factor field of the resulting structure.
To visualize the confidence values:
- Open the Display tab.
- Click and hold the Ribbon representation button.
- Select Color by B-Factor.
This provides a convenient way to identify regions of high and low confidence in the predicted complex.
Select Run Affinity Model in the co-folding dialog to calculate Boltz-predicted binding free energies.
The predicted values are reported in kcal/mol for each conformation. They are printed in the terminal and stored in the resulting object's _AFFINITY_PRED field.
For example, the first value can be accessed using:
Field( a_ "_AFFINITY_PRED" )[1]
When updating from an installation without the affinity modules, enable Force Weights Update to download the additional required model files.
For Local Boltz Diffusion docking, affinity predictions are stored in the dG_bind_pred results column.
The Force Weights Update option downloads or refreshes the Boltz model weights and required modules used for the calculation.
Use this option when updated weights or additional model modules become available, or when the locally installed files need to be refreshed.
In particular, updating an installation that does not yet contain the affinity modules requires a weights update before using affinity prediction.
Under normal circumstances, it is not necessary to enable this option for every prediction.
Run on GPU is selected by default in both:
Homology -> Co-Folding Boltz Model
Docking -> Flexible Receptor -> Local Boltz Diffusion
Leave this option selected to use GPU acceleration. Clear the checkbox to perform the calculation on the CPU.
[ Local boltz diffusion gui | Results and Scores | Local Boltz Diffusion Command ]
Local Boltz Diffusion docking (_LBDdock) uses the Boltz diffusion model within the ICM docking framework.
The calculation starts from an existing receptor structure and applies diffusion at a lower noise magnitude than full co-folding. Both the ligand and the surrounding receptor region can adapt during docking while retaining information from the starting receptor structure.
ICM crops the receptor to the region surrounding the docking box, reducing memory requirements and accelerating the calculation. The ligand remains localized within the region of interest defined by the docking project.
Local Boltz Diffusion uses approximately half the number of diffusion steps used for full co-folding.
This method is useful when binding-site adaptation or induced fit may be important and an experimentally determined or previously modeled receptor structure is available.
- Open (Docking -> Open Project) a previously prepared ICM docking project or prepare it in the same session..
- Load the molecular table (e.g. .sdf file) containing the ligands to dock.
- Select Docking -> Flexible Receptor -> Local Boltz Diffusion.
- In the Run Local Diffusion dialog, select the ligand table under Mol Table.
- Set Number of Conf to the number of conformations to generate for each ligand.
- Leave Run on GPU selected to use GPU acceleration, or clear it to run on the CPU.
- Click OK to start the calculation.
The dialog contains the following settings:
Project name: Displays the currently opened docking project, which defines the receptor and docking region.
Mol Table: Selects the molecular table containing the input ligands.
Number of Conf: Specifies the number of conformations generated for each ligand. The dialog initially displays 20.
Run on GPU: Uses GPU acceleration when selected. This option is selected by default. Clear the checkbox to run on the CPU.
Options: Accepts additional script options that are passed directly to _LBDdock. This provides access to options that are not exposed as separate controls in the dialog. Leave this field empty unless additional script options are required.
The resulting structures can be browsed and analyzed similarly to a conventional ICM docking hitlist.
Chiral_Good: Indicates whether the chirality of the resulting ligand conformation matches the original input ligand. A value of 1 indicates a match; 0 indicates that the chirality check failed. Chirality is enforced during diffusion and checked again in the final structure.
RTCNNorig: The original RTCNN score, including free strain.
RTCNN: The RTCNN score recalculated using the -S option, without strain. The VLS score is also calculated.
dG_bind_pred: Contains the Boltz-predicted binding free energy in kcal/mol for each result row when affinity predictions are generated.
Local Boltz Diffusion can also be run from the command line:
icm _LBDdock <dockingproj> input=<ligands.sdf> nconf=<confs_per_lig> output=<out.icb>where:
dockingproj: An existing ICM docking project defining the receptor and docking region.
input: Specifies the input ligand file in SDF format.
nconf: Specifies the number of conformations generated for each ligand.
output: Optionally specifies the output ICB filename.
Passaro, S., Corso, G., Wohlwend, J., Reveiz, M., Thaler, S., Somnath, V. R., Getz, N., Portnoi, T., Roy, J., Stark, H., Kwabi-Addo, D., Beaini, D., Jaakkola, T., and Barzilay, R. (2025). Boltz-2: Towards Accurate and Efficient Binding Affinity Prediction. bioRxiv. doi:10.1101/2025.06.14.659707