DiffDock-PP

Protein-protein docking using diffusion models for predicting how proteins interact and form complexes. DiffDock-PP combines advanced AI diffusion models with ESM-2 protein language models to accurately predict protein-protein interactions, generating multiple binding orientations with confidence scoring. This tool is particularly powerful for antibody-antigen complex prediction, enzyme-substrate interactions, and understanding protein assembly mechanisms, providing detailed interface analysis and residue-level interaction mapping crucial for therapeutic development.

Key Features

  • Protein Complex Prediction: Models how two proteins bind together
  • Multiple Poses: Generates diverse binding orientations with confidence scoring
  • Interface Analysis: Identifies key residues at binding interfaces
  • ESM-2 Encoding: Uses protein language models for evolutionary information

Input Requirements

Protein Structures

  • Format: PDB files for both proteins (receptor and ligand)
  • Preparation: Clean structures with proper formatting
  • Constraints: Works with rigid protein structures

Key Parameters

  • num_samples: 40 poses generated (typical setting)
  • run_name: Identifier for the docking job
  • Output: Ranked poses with confidence scores

Antibody-Antigen Example

Based on the 4G6K complex tutorial:

  • Input: Heavy/light chain sequences from Boltz-2 prediction
  • Docking: 40 poses generated with confidence scoring
  • Analysis: CAPRI metrics and interface characterization
  • Visualization: Top poses exported for Mol* analysis

Confidence Scoring

Confidence Thresholds:

  • High Confidence: score > 0
  • Moderate Confidence: -1.5 < score ≤ 0
  • Low Confidence: score ≤ -1.5

Integration Workflows

  • Boltz-2 → DiffDock-PP → Protein complex prediction
  • Structure preparation → DiffDock-PP → Docking Report analysis
  • Complex prediction → GROMACS → MD validation
  • Docking Report - Protein-protein analysis with CAPRI metrics
  • GROMACS - MD validation of predicted complexes
  • Mol* - Interactive complex visualization

Tutorials