How to Use the Lemna Platform: A Complete Guide

The Lemna Platform is a powerful web-based interface for biomolecular design, offering access to state-of-the-art AI models for protein structure analysis and sequence design. This guide will walk you through everything you need to know to get started and make the most of the platform.

What is the Lemna Platform?

Lemna provides physics-driven AI tools for biomolecular design, helping researchers understand molecular interactions at the atomic level. The platform hosts two main models:

• PeSTo (Protein Structure Transformer): Predicts protein interaction interfaces from protein structures
• CARBonAra: Designs protein sequences based on atomic coordinates and molecular context

Both models are freely available for academic research use, with results typically available in under a minute.

Getting Started

1. Access the Platform

Visit the Lemna Platform at www.lemna.bio and navigate to the dashboard. You’ll need to sign in with your Google account to access the models.

2. Choose Your Model

Once logged in, you’ll see the available models on the main dashboard:

• Click on PeSTo for protein interaction interface prediction
• Click on CARBonAra for protein sequence design

Each model has its own dedicated page with a submission form and job history.

Using PeSTo: Protein Interaction Interface Prediction

PeSTo predicts interaction interfaces for proteins, identifying binding hot spots for proteins, nucleic acids, lipids, small molecules, and ions.

Step 1: Prepare Your Input

You can provide input in three ways:

1. PDB ID: Enter a 4-character PDB ID (e.g., 2OOB) from the RCSB Protein Data Bank
2. UniProt ID: Enter a 6-character UniProt ID (e.g., P12345) to fetch AlphaFold structures
3. File Upload: Upload your own PDB file directly from your device

Step 2: Configure Your Job

• Model Version: Select the version you want to use (defaults to the latest)
• Job Title: Give your job a descriptive name (optional - will be auto-generated if left blank)
• Selected Chains: Specify which chains to analyze (e.g., A,B for chains A and B, or leave empty for all chains)

Step 3: Submit and Monitor

Click “Submit Job” to start the prediction. Your job will be queued and processed. You can:

• View job status in real-time
• See progress updates as the job runs
• Access results automatically when complete

Understanding PeSTo Results

PeSTo results are organized by interaction type:

• Protein: Protein-protein interaction predictions
• Nucleic Acid: Protein-nucleic acid interaction predictions
• Ion: Ion binding site predictions
• Ligand: Small molecule binding site predictions
• Lipid: Lipid interaction predictions

Each result includes:

• Per-residue predictions: A table showing predictions for each residue
• Prediction scores: Values from 0 (no interface) to 1 (strong interface)
• Interactive 3D viewer: Visualize predictions directly on your structure
• Downloadable files: PDB files with prediction annotations for each interaction type

Interpreting Prediction Scores

• 0.7 - 1.0: High confidence interaction site (shown in green)
• 0.4 - 0.7: Moderate confidence (shown in yellow)
• 0.0 - 0.4: Low confidence (shown in red)

Using CARBonAra: Protein Sequence Design

CARBonAra designs protein sequences tailored to specific structural environments, including interactions with ligands, ions, or other molecules.

Step 1: Prepare Your Input

Similar to PeSTo, you can use:

• PDB ID: Enter a 4-character PDB ID
• UniProt ID: Enter a 6-character UniProt ID
• File Upload: Upload your own PDB file

Step 2: Configure Design Parameters

CARBonAra offers several customization options:

• Number of Sequences: How many sequences to generate (default: 10, max: 100)

• Imprint Ratio: Controls the balance between sequence diversity and confidence

  • Higher values (0.7-1.0): More confident but less diverse sequences
  • Lower values (0.0-0.3): More diverse sequences with reduced confidence
  • Default: 0.5: Balanced approach

• Molecular Context Options:

  • Ignore Heteroatoms: Exclude non-standard atoms from design context
  • Ignore Water: Exclude water molecules from design context

• Chain and Position Selection:

  • Known Chains: Chains with fixed sequences (e.g., A,B)
  • Known Positions: Specific residue positions to keep fixed (e.g., 37,38,39,40)
  • Unknown Positions: Positions to design (leave empty for all positions)
  • Ignored Amino Acids: Amino acids to exclude from design (e.g., C for cysteine)

Step 3: Submit and Monitor

Submit your job and monitor its progress. CARBonAra jobs typically complete in under a minute.

Understanding CARBonAra Results

Results include:

• Designed Sequences: A list of generated sequences with confidence scores
• Position-Specific Scoring Matrix (PSSM): For each position, shows the likelihood of each amino acid
• Sequence Statistics: Summary statistics for all generated sequences
• Download Options:
  • Download sequences as FASTA files
  • Download PSSM data as CSV
  • Download complete results as JSON

Best Practices for CARBonAra

1. Start with default settings: Begin with imprint ratio 0.5 and adjust based on your needs
2. Use known positions: If you know certain residues must remain fixed, specify them in “Known Positions”
3. Consider molecular context: Include ligands or ions if they’re important for binding
4. Review multiple sequences: The top-scoring sequences may not always be the best for your specific application

Managing Your Jobs

Job History

Both PeSTo and CARBonAra pages display your job history in a table showing:

• Job title and status
• Creation date
• Model version used
• Quick access to view results

Job Details Page

Click on any completed job to view detailed results:

• Overview Tab: Job metadata, input parameters, and status information
• Results Tab: Interactive results viewer with tables, statistics, and visualization
• Viewer Tab: 3D molecular viewer powered by Mol* to visualize structures and predictions

Downloading Results

You can download results in multiple formats:

• JSON: Complete results data for programmatic access
• CSV: Tabular data for spreadsheet analysis
• FASTA: Sequence files (for CARBonAra)
• PDB Files: Structure files with annotations (for PeSTo)

Tips and Best Practices

General Tips

1. Validate your input: Ensure PDB files are properly formatted and not corrupted
2. Use descriptive job titles: Help yourself find jobs later in your history
3. Check job status: Jobs typically complete in under a minute, but check the status if it seems stuck
4. Review the input summary: Always verify that the correct chains or parameters were used

For PeSTo

1. Chain selection: If your protein has multiple chains, consider analyzing them separately first
2. Compare interaction types: Look at multiple interaction types to get a complete picture
3. Use the 3D viewer: Visual inspection can reveal patterns that tables might miss

For CARBonAra

1. Experiment with imprint ratio: Try different values to balance diversity and confidence
2. Use constraints wisely: Fixed positions and ignored amino acids can guide design toward functional sequences
3. Generate multiple sequences: Don’t rely on a single sequence; review several top candidates
4. Consider post-processing: Generated sequences may need further optimization or validation

Troubleshooting

Common Issues

• Job fails: Check that your input file is valid and properly formatted
• Results not loading: Refresh the page or check your internet connection
• File upload fails: Ensure your file is a valid PDB file and under size limits
• Invalid PDB/UniProt ID: Verify the ID exists in the respective databases

Getting Help

If you encounter issues:

• Check the job details page for error messages
• Review your input parameters
• Contact support at [email protected]

Usage Terms

Remember that:

• Academic use is free: Free access for academic research purposes
• Commercial use requires permission: Contact us for commercial licensing
• Data privacy: Your data remains your property
• Citation: Cite the relevant papers when publishing results:
  • PeSTo Paper
  • CARBonAra Paper

Next Steps

Now that you understand how to use the platform:

1. Try PeSTo: Start with a well-known protein structure to familiarize yourself with the interface
2. Experiment with CARBonAra: Design sequences for a structure you’re working with
3. Explore the results: Use the interactive viewers and download options to analyze your results
4. Build your workflow: Integrate Lemna predictions into your research pipeline

The Lemna Platform makes advanced biomolecular design accessible and straightforward. Whether you’re predicting interaction interfaces or designing new sequences, the platform provides the tools you need to accelerate your research.

For questions or feedback, visit www.lemna.bio or reach out to our team at [email protected].