A structure prediction can be delayed before it even starts: the input is in the wrong format, a component is missing, or the selected service does not offer the feature you need. These problems are easier to prevent than to diagnose after submission. A useful first step is to decide what you are trying to model and prepare the inputs around that choice.
SAMSON’s Biomolecular Structure Prediction extension provides access to AlphaFold-2, Boltz-2, and Chai-1 through Home > Predict. The services do not use identical input workflows. Choosing between them is less about finding one universal winner and more about matching your system and available inputs to the service’s supported options.
Start with the components in your system
If your input is one or more FASTA files and you want to select an AlphaFold model and a multiple-sequence-alignment database, the documented AlphaFold-2 workflow is a direct fit. Before opening the prediction panel, check that the FASTA files are ready and that you know which model and database you intend to use. AlphaFold predictions run on cloud machines with different performance and costs, so check the available options and credit requirements before starting.
For a system assembled from different molecular components, Boltz-2 lets you add proteins, DNA, RNA, and ligands. Its ligand input can use a SMILES string, while the documented workflow also accepts CCD codes for the relevant components and modifications. This makes it important to prepare each component separately: have the sequence, ligand representation, or CCD code ready rather than trying to reconstruct it while filling out the form.
Chai-1 also lets you add proteins, DNA, RNA, and ligands, with ligand SMILES strings as an input option. Its documented workflow includes pocket and contact restraints. These are useful when you have a specific spatial relationship you want to include in the prediction: a pocket restraint adds a distance restraint between a residue and a chain, while a contact restraint adds one between two residues.
Use a short pre-submit checklist
- List the components. Write down whether the system includes protein, DNA, RNA, or a ligand. This helps you choose a workflow that accepts the inputs you have.
- Prepare the representations. Gather FASTA files for the AlphaFold-2 workflow, or the appropriate sequences, CCD codes, or SMILES strings for Boltz-2 or Chai-1.
- Check modifications carefully. In Boltz-2, a modification is entered with a residue position and CCD code. Confirm the position against the sequence you are submitting; a position entered for a different sequence can describe the wrong residue.
- Add restraints only when you can specify them clearly. For Chai-1, decide which residues and chains are involved before entering pocket restraints, or which two residues define a contact restraint. Restraints should reflect the relationship you mean to model.
- Consider credits before submission. Boltz-2 and Chai-1 calculations use computing credits; a typical prediction with either is documented as costing 0.5 to 1 credit. AlphaFold-2 uses cloud machines with varying performance and costs, so its cost depends on the selected option.
After the job finishes
Cloud predictions are submitted over a secure connection. You can monitor and access results in SAMSON under Interface > Cloud jobs, or through SAMSON Connect > Account > Jobs. When prediction files include pLDDT values, SAMSON uses them to colorize the loaded structures. Treat that coloring as a cue for inspecting the model, not as a replacement for evaluating whether the structure suits your next step.
Once a prediction is loaded, inspect the structure and compare candidates where appropriate. Depending on your project, the next task may be protein preparation, alignment with a related structure, or preparing a model for docking or simulation. The prediction is a starting point for that work, not a guarantee that every downstream question has been answered.
For the complete workflows and service-specific instructions, see the SAMSON biomolecular structure prediction tutorial.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.