ppi-interface-skill

PPI Interface Analysis & Hotspot Prediction — BSA, alanine scanning, ColabFold, FreeBindCraft

Links

README

Protein-Protein Interface Analysis & Hotspot Prediction

A complete pipeline for analyzing protein-protein interfaces, identifying binding hotspots, predicting complex structures from sequences, and designing de novo binders.

Features

Installation

Feature	Description
BSA Analysis	Buried Surface Area per residue via ShrakeRupley SASA
Alanine Scanning	SASA-proxy hotspot identification (BSA ≥ 25 Å² = hotspot)
H-bonds & Salt Bridges	Distance-based detection at 3.5Å / 4.5Å cutoffs
Shape Complementarity	Sc proxy (Lawrence & Colman 1993)
ColabFold Mode B	AlphaFold2-Multimer v3 prediction from sequences
FreeBindCraft Mode C	De novo binder hallucination (GPU required)


pip install biopython numpy pandas matplotlib seaborn scipy requests
pip install colabfold[alphafold]  # for Mode B

Quick Start


Mode A: Analyze PD-1/PD-L1 complex (PDB 4ZQK)
python ppi_pipeline.py

Mode B: Predict complex from sequences
Edit run_mode_b.py and uncomment demo_mode_b()

Demo Results (PD-1/PD-L1, PDB 4ZQK)

**Total BSA**: 1823.7 Å² (typical Ab-Ag: 1200–2000 Å² ✓)
**Shape Complementarity**: 0.666 (>0.65 = good ✓)
**H-bonds**: 6 | **Salt bridges**: 2
**Hotspots Chain A (PD-L1)**: 12 residues → Top: A56TYR, A125ARG, A113ARG
**Hotspots Chain B (PD-1)**: 10 residues → Top: B134ILE, B128LEU, B78LYS

Output Files


results_pdl1/
├── interface_residues.csv   # All interface residues + scores
├── hotspots.json            # Hotspot list for FreeBindCraft
├── interface_bsa.png        # BSA bar chart
├── contact_map.png          # Inter-chain contact heatmap
├── composition_radar.png    # Polar/apolar composition radar
└── interface_report.txt     # Full text report

References

Mirdita et al. (2022). ColabFold. *Nature Methods*
Bogan & Thorn (1998). Anatomy of hot spots. *JMB*
Pacesa et al. (2024). BindCraft. *Nature*
Lawrence & Colman (1993). Shape complementarity. *JMB*