A structure-prediction job can be easy to launch and still be frustrating to set up. The input may be a protein sequence, a mixture of biomolecules, or a ligand described by a SMILES string. The available services also ask for different kinds of information. Choosing a workflow before opening the prediction panel can save time and help you avoid preparing the wrong input.
In SAMSON, AlphaFold-2, Boltz-2, and Chai-1 are available through the Biomolecular Structure Prediction extension. They are not interchangeable forms: each has its own input workflow and options. A useful first step is to write down what you want to model and what information you already have.
Start with the input, not the button
- Have one or more FASTA files? The documented AlphaFold-2 workflow starts from FASTA files. You can select a model, such as monomer or multimer, and choose a database for multiple sequence alignment.
- Need to enter protein, DNA, RNA, or ligand components? Boltz-2 lets you add those components and provide the corresponding sequences, CCD codes, or ligand SMILES strings. It also offers an option to add sequence modifications using a residue position and a modification CCD code.
- Want to specify distance restraints? Chai-1 accepts protein, DNA, RNA, or ligand inputs and provides options for pocket and contact restraints. A pocket restraint defines a distance restraint between a residue and a chain; a contact restraint defines one between two residues.
This is a practical way to narrow down the interface you need. For example, a FASTA-based workflow with model and alignment-database choices points you toward AlphaFold-2. A job that needs explicitly entered components or modifications fits the documented Boltz-2 workflow. If your setup calls for the described distance restraints, review Chai-1’s restraint options before starting.
Make a small preflight checklist
- List every component. Note whether the system includes protein, DNA, RNA, or a ligand.
- Check the format you have. Prepare FASTA files, sequences, CCD codes, or ligand SMILES as appropriate for the selected service.
- Record special requirements. For Boltz-2, note any sequence modification and its residue index and CCD code. For Chai-1, decide whether to enter pocket or contact restraints.
- Review the compute arrangement. Predictions run in the cloud, and data are sent over a secure connection. AlphaFold-2 offers cloud machines with different performance and costs, including A100 GPU instances. Boltz-2 and Chai-1 use a dedicated cloud A100 GPU and computing credits; the documentation gives a typical cost of 0.5 to 1 credit for each of those services.
When you are ready, open Home > Predict, choose the service, enter the inputs and options, then select Start prediction. Results can be viewed in SAMSON under Interface > Cloud jobs or on SAMSON Connect under Account > Jobs. If the result files contain pLDDT values, SAMSON colorizes the loaded structures based on those values. Treat that coloring as information to inspect alongside the model, rather than as a substitute for evaluating whether the structure suits your next step.
Plan for what happens after prediction
A prediction is often one stage in a larger modeling task. Once results are loaded, inspect the structures and compare candidates when needed. Then decide whether the chosen model is ready for docking, simulation, or visualization. The documentation also links to tutorials on protein preparation and validation, protein alignment, and GROMACS workflows, which can help with those follow-up tasks.
For service-specific steps and further details, read the original structure-prediction tutorial.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.