A structure-prediction run can stall before it starts for a surprisingly practical reason: the input does not match the service, or the service does not offer the kind of control your system needs. A protein sequence, a ligand SMILES string, a chemical component code, and a residue-level restraint are not interchangeable inputs. Matching the job to the information you already have makes setup clearer and helps you know what to inspect when results arrive.
SAMSON’s Biomolecular Structure Prediction extension provides access to AlphaFold-2, Boltz-2, and Chai-1 through Home > Predict. Rather than treating them as three identical buttons, use the details of your system and input to guide your choice.
Start with the information in hand
- Have FASTA files and need to choose an AlphaFold model? AlphaFold-2 accepts one or more FASTA files. During setup, choose the model, such as monomer or multimer, and the database used for multiple sequence alignment. The available cloud machines have different performance and costs, so check the job options and credit requirements before starting.
- Need to enter proteins, nucleic acids, or a ligand? Boltz-2 lets you add protein, DNA, RNA, or ligand inputs. For a ligand, the documentation lists CCD codes or SMILES strings as input options. Boltz-2 also supports sequence modifications specified by a residue position and a CCD code.
- Need to express distance restraints? Chai-1 supports pocket restraints between a residue and a chain, and contact restraints between two residues. These controls can be useful when you have specific interaction information you want to include in the prediction setup.
This is a practical way to narrow the choice: begin with the molecules and evidence you can actually provide. If your key input is a FASTA file and model choice, look at the AlphaFold-2 workflow. If you need to assemble a system from different molecule types or enter a modification, review Boltz-2’s input options. If your setup calls for the documented distance-restraint controls, review Chai-1.
Prepare before opening the prediction panel
A small amount of preparation can prevent interruptions. Gather the sequences, FASTA files, CCD codes, or SMILES strings you plan to enter. For a modified protein, DNA, or RNA in Boltz-2, note the residue index and the modification’s CCD code. For Chai-1 restraints, decide which residue and chain define a pocket restraint, or which two residues define a contact restraint. This turns the form-filling step into a deliberate description of the system rather than a search for missing input details.
Plan for cloud jobs and credits
Predictions run in the cloud, and data are sent via a secure connection. You can follow results in SAMSON under Interface > Cloud jobs or in your SAMSON Connect account under Account > Jobs. The documented typical cost for a Boltz-2 or Chai-1 prediction is between 0.5 and 1 computing credit. AlphaFold-2 uses a choice of cloud machines with varying performance and costs. If you need credits, the documentation provides options to buy computing credits or contact the team.
Make the result useful
Once a prediction is loaded, inspect the structure before using it in downstream work. SAMSON colorizes structures based on pLDDT values when those values are available in the result files. Treat that visualization as one piece of information to review, then compare candidates when appropriate and prepare the chosen model for docking, simulation, or visualization. For publication, follow the relevant citation guidance: the documentation asks users of the AlphaFold service to cite the AlphaFold paper and, when applicable, the AlphaFold-Multimer paper.
The key habit is simple: match the prediction setup to your input, record any modifications or restraints you intend to use, and check the returned structure before building the next step of your workflow around it.
Read the original biomolecular structure prediction tutorial for the complete service instructions and references.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.