Doing More With Less: Predicting Complex Biostructures Without Local Resources

One recurring frustration in molecular modeling is the limitation imposed by local compute resources. Predicting biomolecular structures with high accuracy using methods like AlphaFold-2 can be very demanding, often requiring specialized GPUs and long runtimes. For many researchers, especially in academic or small lab environments, this is a real bottleneck. 

If this sounds familiar, there’s good news. You can now offload this computation-heavy task to the cloud directly within SAMSON using the Biomolecular Structure Prediction extension. In this post, we focus specifically on how to perform structure predictions using AlphaFold-2 without any need for local hardware resources.

Why use AlphaFold-2 in the Cloud?

AlphaFold-2 is known for its accuracy in protein structure prediction, but setting up and running it locally can be a complex and resource-heavy endeavor. The SAMSON ecosystem integrates AlphaFold predictions as an easy-to-access cloud-based feature. This means:

  • No local installation of AlphaFold or its dependencies
  • No requirement for high-end GPUs
  • Configurable parameters based on your needs (e.g., monomer vs. multimer models)

How to access AlphaFold-2 in SAMSON

Here’s how you can submit AlphaFold-2 predictions directly from within SAMSON:

  1. Go to Home > Predict in SAMSON.
  2. Select the AlphaFold-2 prediction service.
  3. Input one or more FASTA files for your target sequences.
  4. Choose the AlphaFold model: monomer, multimer, and the MSA database.
  5. Click Start prediction.

All calculations run in the cloud and are done using dedicated computing hardware (such as A100 GPUs), ensuring performance is not a bottleneck. Once the predictions are complete, results appear under Interface > Cloud jobs in SAMSON or in your SAMSON Connect account.

Understanding the Results

When a prediction is loaded into SAMSON, the resulting structure is automatically colorized based on pLDDT values (if they are included), giving you immediate visual insight into the confidence of the model in different regions.

How Much Does It Cost?

The prediction service uses a credit-based system. You can purchase computing credits or request them for evaluation or academic use. Running a prediction with AlphaFold-2 will deduct credits depending on the cloud computing resources selected.

Reproducibility and Publishing

If you use AlphaFold-generated models in a publication, make sure to cite the original work:

This integration isn’t just about convenience; it opens the door for more flexible, accessible workflows in biomolecular modeling. Even if your hardware isn’t suited for deep structural predictions, you can get publication-ready results — directly from your laptop.

To learn more, visit the full tutorial documentation here: Biomolecular Structure Prediction in SAMSON.

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

Comments are closed.