A structure prediction can stall before it begins—not because the science is difficult, but because the inputs are scattered across FASTA files, sequence strings, ligand identifiers, and restraint details. Choosing a service without first checking what your system contains can mean repeating setup work or leaving out information the prediction needs.
SAMSON’s Biomolecular Structure Prediction extension brings three cloud services into one workflow: AlphaFold-2, Boltz-2, and Chai-1. The practical first step is to match your input and modeling question to the service’s available setup options.
Start with the inputs you have
- AlphaFold-2: Choose this route when your input is one or more FASTA files. You can select a model, such as monomer or multimer, and choose a database for multiple sequence alignment. Have the FASTA files ready before opening the prediction tool.
- Boltz-2: Use its input panel to add proteins, DNA, RNA, or ligands. Enter the corresponding sequences, CCD codes, or ligand SMILES strings. If you need a sequence modification, add it with its residue index and CCD code.
- Chai-1: Add protein, DNA, RNA, or ligand inputs using sequences or SMILES strings. If your modeling setup calls for distance restraints, you can add a pocket restraint between a residue and a chain, or a contact restraint between two residues.
This quick inventory is useful when a system includes more than a protein sequence. For example, if your planned input includes a ligand represented as SMILES, Boltz-2 and Chai-1 both provide an “Add ligand” option. If you need to specify modifications or restraints, check which service offers the particular control you need before you enter the rest of the job.
A simple preflight checklist
- List every component. Note whether the system includes protein, DNA, RNA, or ligand input.
- Check the input format. Prepare FASTA files for AlphaFold-2, or the appropriate sequences, CCD codes, and SMILES strings for Boltz-2 or Chai-1.
- Write down special details first. For Boltz-2, record modification positions and CCD codes. For Chai-1, identify the residues and chains involved in any distance restraints.
- Consider the model setup. For AlphaFold-2, choose the model type and multiple-sequence-alignment database before starting.
- Check computing credits. Boltz-2 and Chai-1 calculations use credits; a typical prediction with either is listed at 0.5–1 credit. AlphaFold-2 runs use cloud machines with different performance and costs, so check your available credits and the selected machine options.
Launch and review
Open Home > Predict, select the service, enter the prepared inputs, and click Start prediction. Predictions run in the cloud, with data sent via a secure connection. You can track results in Interface > Cloud jobs in SAMSON or through SAMSON Connect > Account > Jobs.
Once results are loaded, inspect the structures rather than treating a prediction as a final validation. SAMSON colorizes structures based on pLDDT values when those values are present in the files. You can then compare candidates as needed and decide whether the chosen model is ready for a relevant next step, such as docking, simulation, or visualization.
For a short Boltz-2 walkthrough, the documentation also includes a video tutorial covering prediction, sharing job results, and generating protein–ligand interaction diagrams.
For the full instructions and service details, read the original SAMSON biomolecular structure prediction tutorial.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.