GAPO-Ab: Genetic Algorithm for Protein Optimization - Antibody Version

GAPO-Ab is a specialized version of GAPO specifically designed for antibody optimization. This tool implements a genetic algorithm approach to optimize antibody sequences for improved binding affinity and stability.

Features

• Optimization of specific antibody regions:
  • CDRs (Complementarity-Determining Regions)
  • Framework regions
  • VH/VL interface
• Multiple mutation strategies using state-of-the-art language models:
  • ESM2 (Evolutionary Scale Modeling)
  • AbLang (antibody-specific language model)
  • Sapiens (human antibody repertoire-based model)
  • IgBert (paired antibody sequences model)
  • Random mutations
• Various objective functions for ΔGbind estimation:
  • Rosetta REF15 scoring function
  • Protein Binding Energy Estimator (PBEE)
  • Interface descriptors with Shape Complementarity

Installation

# Clone the repository
git clone https://github.com/jsartori12/GAPO_Ab
cd GAPO_Ab

# Install dependencies (requirements.txt should be included in your repository)
pip install -r requirements.txt

Dependencies

• PyRosetta
• Python 3.7
• Additional requirements will be listed in requirements.txt

Algorithm Overview

1. Initial Population Generation

   • Random generation: Introduces random mutations at allowed positions
   • Optimized generation: Uses ESM2 model for naturally plausible sequences

2. Structure Modeling

   • Uses PyRosetta's Repack function for mutation modeling
   • Optimizes side chains within 8Å radius
   • Performs energy minimization using FastRelax

3. Region Selection

   • Implements ANARCI algorithm for CDR and framework identification
   • Interface detection between VH/VL domains (8Å distance criterion)

4. Optimization Process

   • Tournament selection
   • Crossover using binary mask
   • Multiple mutation strategies
   • Objective function evaluation

Usage

[Add specific usage instructions and code examples here]

Objective Functions

1. Rosetta ΔGbind

ΔGbind = ΔGcomplex - (ΔGAntibody + ΔGAntigen)

2. Protein Binder Energy Estimator (PBEE)

Uses machine learning model PBEE to predict binding free energy based on Rosetta structural descriptors.

3. Interface composed score

ΔGInterface = Σ(InterfaceGiComplex) - (Σ(InterfaceGiAntibody) + Σ(InterfaceGiAntigen))
Final Score = ΔGInterface × Shape_Complementarity

Citation

If you use GAPO-Ab in your research, please cite: [Add citation information when available]

License

[Add your chosen license]

Contributors

[Add contributor information]

Related Projects

• GAPO (Original version): https://github.com/izzetbiophysicist/GAPO
• ESM2_NanoGEN: https://github.com/jsartori12/ESM2_NanoGEN

Contact

[Add contact information]