FASTA in, Structure out: Using AlphaFold-2 in SAMSON

Many molecular modelers face a common challenge: how to obtain reliable 3D structures for proteins when experimental data isn’t available. Whether you’re trying to model a newly discovered sequence or understand domain interactions, structure prediction is your go-to route. Fortunately, SAMSON offers a streamlined way to do this using AlphaFold-2.

Developed by DeepMind, AlphaFold-2 is well-known for its high-accuracy predictions, but setting it up manually, allocating GPUs, installing dependencies, and managing input/output files can be overwhelming. SAMSON helps solve this practical bottleneck by providing a cloud-based, graphical interface for AlphaFold-2 right in its platform — no setup pains.

Turning a Sequence into a Structure

With SAMSON, you can use AlphaFold-2 in just a few clicks:

  1. Go to Home > Predict in the SAMSON workspace.
  2. Select the AlphaFold-2 prediction service.
  3. Upload one or more .fasta sequence files.
  4. Choose the AlphaFold model variant (e.g., monomer or multimer).
  5. Select a database for multiple sequence alignment (MSA).
  6. Click Start prediction.

The predictions are executed in the cloud, on high-performance machines with NVIDIA A100 GPUs. Cloud usage is based on computing credits, which can be purchased or requested via contact if needed.

Where to View Results

You don’t need to wait around. Once computation is done, results show up directly in SAMSON under Interface > Cloud jobs, or on your online account page at SAMSON Connect > Account > Jobs.

In addition to viewing the structural model, SAMSON automatically colorizes the result based on pLDDT values, giving quick visual insight into prediction confidence on a per-residue level.

FAST Tips for Better Outcome

  • Use well-curated and clean FASTA sequences.
  • Choose “multimer” models if modeling complexes or domain interactions.
  • Analyze pLDDT-based coloring to identify potentially unreliable regions.
  • Combine AlphaFold predictions with SAMSON’s analysis tools (e.g., surface mapping or interaction diagrams) for downstream tasks.

When Should You Use AlphaFold-2?

This service is particularly helpful when you:

  • Need missing structure data for docking or simulation.
  • Want to compare structural differences among sequence variants.
  • Are designing multifunctional proteins or complexes and need a starting model.

A Note on Publications

If you use AlphaFold-2 predictions in your research, make sure to cite the original AlphaFold publication. If you used the multimer model, also cite the AlphaFold-Multimer paper. This ensures proper attribution to the developers.

To read more and explore other prediction options like Boltz-2 and Chai-1, visit the full documentation here: 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.