A comprehensive example demonstrating how to use the BioLM SDK within Nextflow pipelines for protein structure prediction and antibody engineering. This project provides two main workflows:

1. intro.nf: Protein structure prediction using ESMFold
2. antibody_engineering.nf: Antibody variant generation using AntiFold

# Install BioLM SDK
pip install biolmai

# Install Nextflow (if needed)
curl -s https://get.nextflow.io | bash

1. Visit BioLM

2. Sign up and get your API token

3. Set it as an environment variable:

   export BIOLMAI_TOKEN="your_token_here"

   Note: The workflows include built-in token validation and will provide clear error messages if the token is missing or invalid.

Option A: Quick Demo (Recommended for first-time users)

nextflow run intro.nf --demo
# Output: results/DEMO.pdb

Option B: Default Example

nextflow run intro.nf
# Output: results/GFP.pdb

Option C: Your Own Sequences

nextflow run intro.nf --input your_proteins.fasta
# Output: results/sequence1.pdb, results/sequence2.pdb, etc.

Quick Start:

nextflow run antibody_engineering.nf --num_variants 5
# Output: results/antibody_engineering_summary.html + analysis files

With Custom Parameters:

nextflow run antibody_engineering.nf \
  --num_variants 10 \
  --sampling_temp 0.8

# List output files
ls -la results/

# View PDB structure (first few lines)
head -20 results/*.pdb

You can run this workflow directly on Seqera Platform without any local setup:

1. Click the badge:
2. Sign in to your Seqera Platform account
3. Configure parameters:
   • Set your BIOLMAI_TOKEN as an environment variable
   • Choose your input mode (demo, default, or custom FASTA)
   • Configure compute resources
4. Launch the workflow

Benefits of Seqera Platform:

• No local installation required
• Scalable cloud compute resources
• Built-in monitoring and visualization
• Easy parameter configuration
• Automatic result management

This project demonstrates two main use cases:

1. Unified Workflow: Single workflow that handles both demo and production use cases
2. ESMFold Integration: Using BioLM's ESMFold for protein structure prediction
3. Nextflow Orchestration: Parallel processing and workflow management
4. PDB Output: Direct extraction of protein structure files

1. AntiFold Integration: Using BioLM's AntiFold for antibody variant generation
2. Multi-Target Processing: Handles EGFR, PDL1, MBP, and IL-7RALPHA targets
3. CDR Analysis: Comprehensive analysis of Complementarity-Determining Regions
4. Automated PDB Download: Downloads PDB files directly from RCSB
5. Rich Reporting: HTML summaries with diversity analysis and statistics

1. BioLM API Token: Get your token from BioLM
2. Python 3.7+: With the BioLM SDK installed
3. Nextflow: Version 20.0 or later

• Purpose: Quick test with a hardcoded protein sequence
• Input: Built-in demo sequence (MKTVRQERLKSIVRILERSKEPVSGAQLAEELSVSRQVIVQDIAYLRSLGYNIVATPRGYVLAGG)
• Output: DEMO.pdb

• Purpose: Standard example using GFP protein
• Input: Built-in GFP sequence
• Output: GFP.pdb

• Purpose: Process your own protein sequences
• Input: Your FASTA file
• Output: One .pdb file per sequence in the FASTA

• Downloads PDB files from RCSB (3c09, 5x8m, 5bjz, 6p67)
• Extracts sequences from heavy, light, and antigen chains
• Generates variants using AntiFold for CDR regions
• Analyzes diversity of generated antibody variants
• Creates reports with statistics and visualizations

• PDB files: Downloaded antibody structures
• Sequence files: Extracted chain sequences
• Variant files: Generated antibody variants
• Analysis files: CDR diversity analysis
• HTML report: Comprehensive summary with statistics

nf-biolm/
├── intro.nf                    # Protein structure prediction workflow
├── antibody_engineering.nf     # Antibody engineering workflow
├── antibody_engineering_test.nf # Test version (mock data)
├── nextflow.config            # Configuration
├── requirements.txt           # Python dependencies
├── tower.yml                 # Seqera Platform configuration
├── LICENSE                   # MIT License
├── README.md                 # This file
├── results/                  # Output directory
│   ├── *.pdb                # Protein structure files
│   ├── analysis/            # Antibody analysis results
│   ├── sequences/           # Extracted sequences
│   ├── variants/            # Generated variants
│   └── *.html               # Summary reports
└── work/                    # Nextflow work directory

The workflow produces PDB structure files directly:

• Format: Standard PDB format
• Location: results/ directory
• Naming: Based on sequence ID from FASTA header

The workflow produces comprehensive analysis results:

• PDB files: Downloaded antibody structures
• Sequence files: Extracted chain sequences (JSON format)
• Variant files: Generated antibody variants (JSON format)
• Analysis files: CDR diversity analysis (JSON + CSV)
• HTML report: Interactive summary with statistics and visualizations

• PDB Files: Standard protein structure files ready for visualization
• Clean Output: No intermediate JSON files, just the structures you need
• Scalable: Can process single sequences or entire FASTA files

• Multi-Target Analysis: EGFR, PDL1, MBP, and IL-7RALPHA antibody targets
• CDR Diversity: Comprehensive analysis of antibody diversity
• Professional Reports: HTML summaries with statistics and visualizations
• Complete Pipeline: From PDB download to final analysis
• Real BioLM Integration: Uses actual AntiFold API for variant generation

• Visualize: Open PDB files in tools like PyMOL, Chimera, or online viewers
• Customize: Modify the workflow for your specific needs
• Scale Up: Process larger datasets with more compute resources

• Analyze Results: Review the HTML summary and diversity analysis
• Explore Variants: Examine generated antibody sequences and their properties
• Customize Targets: Modify the workflow to work with your own antibody targets
• Production Use: Scale up for large-scale antibody engineering projects

• API Token Issues: Ensure BIOLMAI_TOKEN is set correctly. The workflows now include graceful token validation and will provide clear error messages if the token is missing.
• Import Error: Run pip install biolmai
• Workflow errors: Check the .nextflow.log file for details
• API Rate Limits: BioLM has rate limits; wait between requests if needed

The workflow can be easily customized by:

1. Modifying sequences: Edit the hardcoded sequences in the workflow
2. Adding parameters: Extend the parameter list for additional options
3. Changing output format: Modify the PDB extraction logic

• Blog Post: Scaling BioLM Workflows with Nextflow: From Notebooks to Production Pipelines - Learn more about integrating BioLM with Nextflow workflows
• BioLM Documentation: https://biolm.ai/ - Official BioLM platform and API documentation
• Nextflow Documentation: https://www.nextflow.io/ - Nextflow workflow framework documentation
• Seqera Platform: https://cloud.seqera.io/ - Cloud-native platform for running Nextflow workflows