A structure-prediction run can stall before it starts for a surprisingly ordinary reason: the input does not match the service. You may have a FASTA file, a ligand SMILES string, a CCD code, or a system that needs a restraint—and each prediction workflow accepts a different combination. A little planning before opening the prediction panel can save time and make it easier to understand what the resulting model represents.
SAMSON’s Biomolecular Structure Prediction extension provides access to AlphaFold-2, Boltz-2, and Chai-1 through Home > Predict. The practical first step is not choosing the most familiar name; it is checking what you want to predict and what form your inputs are in.
Match the service to the input
- AlphaFold-2: Use one or more FASTA files. You can then select a model, such as monomer or multimer, and a database for multiple sequence alignment. This is the workflow to consider when your starting material is protein sequence in FASTA format.
- Boltz-2: Add proteins, DNA, RNA, or ligands. The interface accepts corresponding sequences, CCD codes, or ligand SMILES strings. For protein, DNA, or RNA modifications, use Add modification and provide the residue index and modification CCD code.
- Chai-1: Add proteins, DNA, RNA, or ligands using sequences or SMILES strings. If the prediction should include distance information, the interface also offers pocket restraints between a residue and a chain, and contact restraints between two residues.
This input check is useful when a project includes more than a protein sequence. For example, if a ligand is part of the question, confirm that your chosen workflow lets you include it and that you have its input representation ready. If a modification matters, prepare its position and CCD code before submitting. If you have a specific distance relationship to express, check whether the restraint options in Chai-1 fit the question.
A short pre-submission checklist
- Write down the system: protein only, a complex involving DNA or RNA, a ligand-containing system, or a combination.
- Gather the right inputs: FASTA files for AlphaFold-2; sequences, CCD codes, or SMILES strings as supported by Boltz-2; and sequences or SMILES strings for Chai-1.
- Decide whether extra information is needed: AlphaFold-2 asks you to choose a model and an MSA database. Boltz-2 supports sequence modifications. Chai-1 provides pocket and contact restraints.
- Check the practical requirements: predictions run in the cloud. Results can be followed in Interface > Cloud jobs in SAMSON or through SAMSON Connect > Account > Jobs.
Costs are also worth checking before you submit. The documentation notes that 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 between 0.5 and 1 credit. You can buy computing credits or contact SAMSON support about requesting them.
After the job completes
Load the results into SAMSON and inspect the structures rather than treating a prediction as the end of the modeling task. When the result files contain pLDDT values, SAMSON colors structures according to those values, giving you a visual cue to consider while reviewing the model. You can compare candidates when needed and prepare a chosen structure for downstream docking, simulation, or visualization, as appropriate for your project.
The right workflow is the one whose inputs and options fit the scientific question—not simply the one with the shortest form. For the full walkthrough of all three services, see the SAMSON biomolecular structure prediction tutorial.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.