A structure-prediction job can go wrong before the computation even begins: the input may not fit the service, a sequence may be missing a modification, or the result may be difficult to find after submission. A little planning helps avoid those workflow snags. This guide focuses on choosing an input path and preparing a prediction in SAMSON—not on claiming that one service is best for every system.
Start with the system you want to model
Before opening the prediction interface, identify what you are submitting and gather the corresponding inputs. Are you predicting from protein sequences in FASTA files, or building a system that includes DNA, RNA, or a ligand? Do you need to represent a chemical modification or specify a spatial restraint? These details help determine which service’s input options fit your task.
In SAMSON, the services are available through Home > Predict, provided you have added the Biomolecular Structure Prediction extension.
Choose by input and setup
- AlphaFold-2: Use one or more FASTA files. You can select a model, such as monomer or multimer, and choose a database for multiple sequence alignment. This is the documented route when your starting point is FASTA input and you need those model and database choices.
- Boltz-2: Add protein, DNA, RNA, or ligand components. Enter the corresponding sequences, CCD codes, or ligand SMILES strings. If you need to represent a sequence modification, use Add modification and provide its residue index and CCD code.
- Chai-1: Add protein, DNA, RNA, or ligand components using sequences or ligand SMILES strings. The interface also offers pocket restraints, which specify a distance restraint between a residue and a chain, and contact restraints, which specify one between two residues.
This is a practical way to narrow the choice: match the service to the input and setup you need. The documentation does not establish that one service will produce better results for every target, so avoid treating the service names as a quality ranking.
Check details before starting
Before submitting, confirm that your sequences and files are ready, that each component is represented in the intended way, and that any modifications or restraints are entered deliberately. In particular, a residue index is part of the modification input for Boltz-2; check it against the sequence you are submitting. For restraints in Chai-1, make sure you understand which residues and chains the distance applies to.
Predictions run in the cloud, with data sent over a secure connection. Results can be viewed in SAMSON under Interface > Cloud jobs or through SAMSON Connect > Account > Jobs. It is useful to note where to find the job before you submit it, especially when you are running multiple candidates.
Plan for credits and inspect the result
Boltz-2 and Chai-1 calculations use computing credits on a dedicated cloud A100 GPU. The documentation gives a typical cost of 0.5 to 1 computing credit for a structure prediction with either service. AlphaFold-2 uses a choice of cloud machines with different performance and costs, including A100 instances; no single cost is specified there. Check your available credits and the current options before launching a job. You can buy computing credits or contact SAMSON support.
When prediction results are loaded, SAMSON colorizes structures by pLDDT values if those values are present in the files. Treat that coloring as information to inspect, not as a substitute for reviewing the model and deciding whether it is suitable for your next step. Compare candidates when needed, and prepare a chosen structure appropriately before docking or simulation.
For a complete walkthrough of the services and their controls, see the biomolecular structure prediction tutorial.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.
