A structure-prediction job can be easy to launch and still be frustrating if the inputs are not ready. Which service accepts your system? Do you need a FASTA file, a ligand SMILES string, or a residue modification? And what should you check after the results arrive? A short decision process can help you choose a suitable workflow before spending time or computing credits.
Start with the system you want to model
In SAMSON, the Biomolecular Structure Prediction extension provides access to AlphaFold-2, Boltz-2, and Chai-1 through Home > Predict. Before opening the prediction interface, identify the molecules in your system and prepare the corresponding inputs. Depending on the service and system, these may include FASTA files, protein, DNA, or RNA sequences, CCD codes, or ligand SMILES strings.
This preparation is more than a clerical step: it helps you avoid choosing a workflow that does not match the information you have. For example, AlphaFold-2 takes one or more FASTA files. Boltz-2 and Chai-1 let you build an input by adding molecular components in the interface.
Choose the workflow by its input options
- AlphaFold-2: Use one or more FASTA files, then select an AlphaFold model, such as monomer or multimer, and a database for multiple sequence alignment. This is the workflow in the documentation for predictions based on FASTA input.
- Boltz-2: Add proteins, DNA, RNA, or ligands. Enter the corresponding sequences, CCD codes, or ligand SMILES strings. If you need to represent a sequence modification, use Add modification and specify the residue position and modification CCD code.
- Chai-1: Add proteins, DNA, RNA, or ligands using sequences or SMILES strings. Its listed restraint options include a pocket restraint between a residue and a chain, and a contact restraint between two residues.
These options provide a practical way to narrow your choice: look first at the molecules and input details your task requires, then select a service whose interface supports them. The documentation does not establish that one service is best for every system, so treat this as a workflow-selection guide rather than a ranking.
Check cost and job handling before you start
Predictions run in the cloud, and SAMSON states that data are sent through a secure connection. Results can be viewed in SAMSON under Interface > Cloud jobs, or through SAMSON Connect > Account > Jobs.
Costs depend on the selected service. AlphaFold-2 predictions use a choice of cloud machines with varying performance and costs, including machines with A100 GPUs. Boltz-2 and Chai-1 use a dedicated A100 GPU and computing credits; the documentation gives a typical cost of 0.5 to 1 computing credit for each of those predictions. If you need credits, you can contact the team or buy computing credits.
Plan what you will do with the result
When prediction files are loaded into SAMSON, structures are colored according to pLDDT values when those values are available in the files. Use that display as one piece of information while inspecting the model. The next step is to review the predicted structure, compare candidates when appropriate, and decide whether it is ready for downstream work such as docking, simulation, or visualization. For protein-focused follow-up, the documentation also links to tutorials on protein preparation and validation and protein alignment.
For a short walkthrough of Boltz-2, including sharing job results and generating protein–ligand interaction diagrams, see the video tutorial.
For service-specific steps and references, see the original biomolecular structure prediction tutorial.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON at samson-connect.net.