From molecular components to a prediction: a practical Boltz-2 checklist

A structure-prediction run can stall before it starts for a simple reason: the molecular system has not yet been translated into the format the prediction service expects. Is the ligand a sequence or a SMILES string? Does a modified residue need to be specified separately? Where will the result appear, and what will the run cost?

This guide focuses on setting up a Boltz-2 prediction in SAMSON, especially when your system includes more than a protein. The goal is to make the input decisions clear before you submit the job.

First, map your system to the input fields

Boltz-2 in SAMSON lets you add proteins, DNA, RNA, and ligands. For proteins and nucleic acids, prepare the relevant sequence. For a ligand, prepare its SMILES string or use a CCD code where applicable. Having these ready before opening the prediction panel makes it easier to check that every component of the system is represented.

Think of this as an inventory, not just a form to fill in:

  • Protein: Which protein sequence should be included?
  • Nucleic acid: Does the system include DNA or RNA that should be modeled alongside the protein?
  • Ligand: Do you have the ligand’s SMILES string or CCD code ready?
  • Modifications: Are there sequence modifications that need to be specified?

This quick inventory helps avoid a common modeling mismatch: preparing a prediction for only one component when the scientific question concerns a larger molecular system.

Add modifications explicitly

If you need to alter a protein, DNA, or RNA sequence, use Add modification. The input requires the residue index and the CCD code of the modification. Before submitting, check that the index refers to the position you intend to modify and that you have the correct CCD code. These details define what you are asking the prediction service to model.

If the system does not require modifications, you can leave this step out. Keeping the input focused on the components and changes relevant to your question makes the setup easier to review.

Submit the job and know where to find it

In SAMSON, open Home > Predict, select Boltz-2, and add the needed components. Add any modifications, then select Start prediction. Predictions run in the cloud over a secure connection. You can follow results in SAMSON under Interface > Cloud jobs, or on SAMSON Connect > Account > Jobs.

Planning for compute credits is part of planning the run. Boltz-2 calculations use a dedicated A100 GPU in the cloud, and a typical structure prediction costs between 0.5 and 1 computing credit. If you need credits, you can buy computing credits or contact SAMSON support.

After the prediction: inspect before moving on

When a result is loaded into SAMSON, the structure is colorized using pLDDT values if those values are available in the files. Treat this as a useful visual aid for examining the returned structure—not as a substitute for checking whether the model fits your research question. Inspect the result, compare candidates if needed, and decide whether the chosen model is suitable for the next stage of your workflow.

For example, a predicted structure may be a starting point for visualization, docking, or simulation preparation. SAMSON’s related tutorials cover protein preparation and validation, protein alignment, and GROMACS Wizard workflows.

The key habit is simple: define the complete molecular system, verify sequence changes and ligand representation, account for compute credits, and inspect the result before using it downstream. For the full walkthrough of AlphaFold-2, Boltz-2, and Chai-1 in SAMSON, learn more in the original documentation.

Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.

Comments are closed.