Pure NumPy · No GPU required · Training-free ΔΔG prediction

ΔΔG Scan &
Stability Prediction

Predict thermodynamic stability changes upon mutation — full saturation mutagenesis scan, thermal denaturation curves, and hotspot residue identification. All in pure Python.

20×
faster than Rosetta
19
mutation features
4
demo proteins
0
GPU required
01 · Capabilities

What can it do

From sequence to stability landscape — ΔΔG prediction, contact map estimation, and thermal denaturation all from first principles.

ΔΔG Prediction

Estimate binding free energy changes for single and multiple mutations using a 19-feature model combining MJ potential, hydrophobicity, and secondary structure context.

🧬

Saturation Mutagenesis

Scan all 19 possible amino acid substitutions at every position. Rank by ΔΔG to identify stabilizing and destabilizing mutations across the entire sequence.

🌡️

Thermal Denaturation

Estimate Tₘ from sequence composition and predict denaturation curves using a two-state model with Van't Hoff analysis.

🔥

Hotspot Identification

Combine conservation scores with ΔΔG magnitude to pinpoint mutation hotspots — critical residues whose perturbation has outsized stability effects.

🕸️

Contact Map Estimation

Derive coevolutionary-style contact maps from sequence alone using knowledge-based potentials — no MSA or homologs needed.

📊

ECharts Visualizations

Heatmaps, bar charts, scatter plots, and denaturation curves — publication-ready figures with T4 Lysozyme, Barnase, Ubiquitin, and GFP demos.

02 · Demo Results

T4 Lysozyme · ΔΔG Scan

Full saturation mutagenesis on T4 phage lysozyme (65 residues). G77A (ΔΔG = −1.4 kcal/mol) and C54T (ΔΔG = −0.5 kcal/mol) are known stabilizers — both ranked among the top predictions.

Top 20 Stabilizing Mutations
Ranked by ΔΔG (kcal/mol) · lower = more stabilizing
ΔΔG Distribution by Position
Mean ΔΔG across 19 substitutions per residue
Thermal Denaturation Curve
Fraction folded vs temperature · two-state model
Contact Map · T4 Lysozyme
Knowledge-based potential · sequence-derived
03 · Pipeline

How it works

From raw sequence to stability landscape in four stages.

1Sequence Input
2Contact Map
3Feature Extraction
4ΔΔG Prediction
5Visualization
# Full pipeline — 4 lines of Python from protein_stability import DDGPredictor, scan_all_mutations predictor = DDGPredictor() results = scan_all_mutations(sequence, predictor) # results['top_stabilizing'] · results['top_destabilizing'] # results['heatmap_data'] · results['denaturation_curve']
04 · Install

Get started

Requires Python 3.9+. No GPU, no external APIs, no training needed.

# Clone the repository git clone https://github.com/junior1p/ProteinStability.git cd ProteinStability # Run the T4 Lysozyme demo python main.py --demo T4_Lysozyme --run-full-scan --n-top 20 # Or use as a library from src.protein_stability import DDGPredictor, scan_all_mutations
Python 3.9+ NumPy only No GPU required Training-free