Efficient Biomolecular Structure Predictions in SAMSON: A Quick Guide

Efficient Biomolecular Structure Predictions in SAMSON: A Quick Guide

Biomolecular modeling often demands accurate and efficient structure predictions. For molecular modelers, balancing computational cost, time, and accuracy can be a persistent challenge. SAMSON’s Biomolecular Structure Prediction extension offers a streamlined solution, enabling users to take full advantage of cloud-based prediction services, including AlphaFold-2, Boltz-2, and Chai-1. Here’s what you need to know to get started and make the most of these tools!

Why Use Cloud-based Structure Prediction?

Local simulations are resource-intensive and often require advanced computing infrastructure. By leveraging cloud-based predictions in SAMSON, users can access high-performance GPU resources, such as A100 GPUs, to quickly and efficiently predict biomolecular structures without needing expensive hardware. The results are securely delivered to your workspace, ensuring both convenience and data integrity.

Accessing the Biomolecular Structure Prediction Extension

First things first, make sure you have the Biomolecular Structure Prediction extension installed in SAMSON. Once installed, navigate to Home > Predict to access the prediction services.

Choosing the Right Service

SAMSON provides access to three powerful prediction tools, each designed for specific needs:

  • AlphaFold-2: Ideal for highly accurate predictions of monomeric or multimeric protein structures, utilizing state-of-the-art machine learning models.
  • Boltz-2: Supports predictions for proteins, DNA, RNA, and small molecule ligands, with options for adding sequence modifications.
  • Chai-1: Specializes in incorporating restraints, such as distance constraints, for proteins, DNA, RNA, and ligands.

Step-by-step: Predicting with AlphaFold-2

Let’s walk through the steps to launch a prediction using AlphaFold-2, SAMSON’s preferred service for protein-centric modeling:

  1. Open Home > Predict and select AlphaFold-2.
  2. Upload your FASTA file(s).
  3. Select an AlphaFold model (e.g., monomer, multimer) and specify the appropriate database for sequence alignment.
  4. Click the Start prediction button.

Once the cloud service completes your prediction, SAMSON automatically colorizes structures based on pLDDT values (confidence scores), assisting you in assessing the reliability of the predictions.

Efficient Predictions with Boltz-2

Boltz-2 adds versatility for projects requiring multi-component predictions. You can input protein/DNA/RNA sequences, CCD codes for small molecules, or SMILES strings for ligands. It also supports sequence modifications, allowing you to modify residues and add chemical modifications with ease. Typical predictions cost between 0.5 and 1 cloud computing credit.

Streamlining Task Complexity with Chai-1

For tasks needing advanced restraints or fine-tuned control, Chai-1 empowers users to incorporate pocket or contact restraints between residues and chains. Whether dealing with complex protein-ligand systems or nucleic acid interactions, Chai-1 offers a customizable approach to molecular modeling.

Manage Results Seamlessly

All prediction outputs are stored securely in the cloud. You can track ongoing jobs in SAMSON via Interface > Cloud Jobs or through your account on the SAMSON Connect platform. Results can also be shared with collaborators, enabling cross-disciplinary workflow enhancements.

Learn More

For a comprehensive understanding and additional tips, visit the full documentation at SAMSON Biomolecular Structure Prediction Tutorial.

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

Comments are closed.