A structure-prediction job can be delayed before it even starts—not because the biology is unusually difficult, but because the input is not ready or the selected service does not match it. A sequence, a ligand SMILES string, and a request to include a residue modification are different kinds of input. Sorting that out first makes the workflow easier to plan and the results easier to interpret.
SAMSON’s Biomolecular Structure Prediction extension provides access to AlphaFold-2, Boltz-2, and Chai-1 through Home > Predict. A useful first question is not “Which service is best?” but “What am I trying to submit?” The services have different input options and controls, so choose based on the system and information you have.
Match the input to the workflow
- AlphaFold-2: Prepare one or more FASTA files. In the interface, select an AlphaFold model—such as monomer or multimer—and a database for multiple sequence alignment before starting the prediction. This makes it a natural workflow when your starting material is protein sequence data and you need to choose those model and database settings.
- Boltz-2: Add protein, DNA, RNA, or ligand inputs. Ligands can be entered using SMILES, and the workflow also supports modifications to protein, DNA, or RNA sequences using a residue position and a CCD code. If you plan to use a modification, have both pieces of information ready before submitting.
- Chai-1: Add protein, DNA, RNA, or ligand inputs, including ligand SMILES. You can also add pocket restraints, which specify a distance restraint between a residue and a chain, or contact restraints, which specify a distance restraint between two residues. These controls are useful when you want to include those kinds of distance constraints in the prediction setup.
A quick pre-submission check
- Identify your components. List the protein, nucleic-acid, and ligand inputs you intend to include. Check that you have the corresponding sequence, FASTA file, CCD code, or SMILES string required by your chosen workflow.
- Decide whether you need extra controls. For AlphaFold-2, choose the model and multiple-sequence-alignment database. For Boltz-2, prepare modification positions and CCD codes if needed. For Chai-1, decide whether pocket or contact restraints belong in the setup.
- Check the resource and credit implications. AlphaFold-2 predictions use a choice of cloud machines with different performance and costs. Boltz-2 and Chai-1 calculations use a dedicated cloud A100 GPU and computing credits; a typical prediction with either is listed as costing 0.5–1 credit. Confirm your available credits or arrange them before starting.
- Know where to find the result. Predictions run in the cloud over a secure connection. Results can be viewed in SAMSON under Interface > Cloud jobs or on SAMSON Connect > Account > Jobs.
After the job finishes
Do not treat a predicted structure as the end of the modeling process. Load it into SAMSON, inspect the structure, and compare candidates when that is useful for your question. When pLDDT values are present in the result files, SAMSON colorizes the structures based on those values; check whether they are available rather than assuming every output includes them. Then prepare the selected model for an appropriate next step, such as docking, simulation, or visualization.
Planning the input first helps avoid a common source of friction: discovering halfway through setup that a needed sequence, identifier, restraint, or credit is missing. For the complete service instructions and related tutorials, read the SAMSON structure-prediction tutorial.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.