How to Apply Pocket and Contact Restraints for Structure Prediction with Chai-1

When predicting biomolecular structures of complex assemblies, one common challenge is incorporating prior knowledge—such as known binding sites or residue interactions—into structure models. Without these constraints, automated prediction methods may miss critical interactions, especially in flexible systems. This is where the Chai-1 service provided by the Biomolecular Structure Prediction extension in SAMSON can help.

If you’re modeling systems where certain residues need to be close or should interact with particular molecular pockets, the ability to add spatial restraints significantly improves the relevance of the predicted structure. Here’s how you can do that in Chai-1:

Using Restraints in Chai-1

To begin, open SAMSON and go to Home > Predict. Then select the Chai-1 prediction service. This gives you control over both the input molecular sequences and the spatial relationships you want the prediction to honor.

1. Enter Biological Components

  • Use Add protein, Add DNA, Add RNA, or Add ligand to specify molecular sequences or SMILES strings.

2. Add Pocket Restraints

  • Click Add pocket restraint to define a distance-based constraint between a molecular residue (from a protein, DNA, or RNA) and a target chain.
  • This is especially useful when you know that a domain or ligand should bind at a specific region of the structure.

3. Add Contact Restraints

  • Use Add contact restraint to force certain residues to remain close together in space.
  • This is helpful if experiments (e.g., cross-linking data) reveal proximity between residues that should be preserved in the structural model.

4. Start the Prediction

Once you’ve placed your restraints, click Start prediction. Chai-1 will generate a structure that meets your specified constraints using cloud-based GPUs, typically costing 0.5–1 computing credit per prediction run.

Why This Matters

Structure prediction without context can lead to misleading interpretations, especially in systems with known interactions. By adding restraints, you guide the algorithm to explore conformational space more intelligently, helping you reach biochemically plausible conformations faster.

These restraint features provide a lightweight alternative to full molecular dynamics refinement when you’re short on time or computing resources, or simply want to start from a better model before simulation.

Practical Considerations

  • A100 GPUs: All calculations run on high-performance A100 GPUs in the cloud.
  • Computing credits: You can buy computing credits or reach out at contact@samson-connect.net for access.
  • Data access: Once the job is complete, results appear in Interface > Cloud jobs and online via your Jobs page.

If you’re working on a system with known residue-residue interactions or are trying to preserve binding site geometry, the ability to include custom restraints directly in your structure prediction workflow will save time and improve precision.

To learn more about how to use Chai-1 and other structure prediction services in SAMSON, visit the full documentation page: https://documentation.samson-connect.net/tutorials/bsp/bsp/

SAMSON and all SAMSON Extensions are free for non-commercial use. You can download SAMSON at https://www.samson-connect.net.

Comments are closed.