Python-based Molecular Docking Platform for Drug Discovery, Bioinformatics, and Computational Chemistry.

PandaDock is a powerful, Python-based molecular docking toolkit combining:

• Traditional docking strategies
• Physics-based scoring functions (MM-GBSA-inspired)
• CHARMm-style dynamics and sampling
• CPU and GPU acceleration
• Automatic flexible residue detection
• Comprehensive reporting with interactive HTML visualization

It is designed for high-accuracy drug discovery, computational chemistry, and next-generation bioinformatics workflows.

• 🔬 Flexible Input Parsing: Supports PDB, MOL, SDF files
• 🎯 Binding Site Definition: Manual or automatic pocket detection
• 📊 Multiple Scoring Functions:
  • Basic (VDW + H-bond)
  • Enhanced (VDW + H-bond + electrostatics + desolvation + hydrophobic)
  • Physics-based (full MM-GBSA inspired energy decomposition)
• 🔥 Hardware Acceleration:
  • Native GPU (PyTorch/CuPy) and multi-core CPU parallelization
• 🧬 Advanced Search Algorithms:
  • Genetic Algorithm (Parallelized)
  • Monte Carlo Simulated Annealing
  • PANDADOCK (Simulated Annealing + Final Minimization)
  • Random Search, Gradient-based, Replica-Exchange
• 🧪 Flexible Residue Docking: Auto-flex and custom-flex options
• 📈 Batch Screening: High-throughput screening with scoring, filtering, and summaries
• 🧩 Comprehensive Reports:
  • Energy component breakdown
  • Interaction analysis
  • Pose clustering
  • RMSD validation
  • Full HTML visualization
  • Binding Affinity calculations
  • All score plots
• 🛠️ Extensible Python API for custom workflows and integrations

PandaDock requires Python 3.6+ and the following dependencies:

• NumPy, SciPy, Matplotlib
• scikit-learn
• RDKit (optional but recommended)
• PyTorch or CuPy (optional for GPU acceleration)

# Create virtual environment
python -m venv .venv
source .venv/bin/activate

# Install dependencies (recommended)
pip install rdkit
pip install torch  # For GPU acceleration

# Install PandaDock
pip install pandadock

# First install RDKit with conda
conda install -c conda-forge rdkit

# Then install PandaDock
pip install pandadock

Tip: Install torch with CUDA for GPU acceleration:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

pip install cupy-cuda11x

🚀 Pro Tip:
Always provide an active site center using -s X Y Z.
It significantly boosts docking speed, precision, and reliability in PandaDock. If GPU is available then it would run smoothly and efficiently or if you have multi-core CPU

# Simple run (default Genetic Algorithm)
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --grid-radius 10.0 --grid-spacing 0.375  -i 10 -a genetic --use-gpu

# Physics-Based docking 
pandadock -p protein.pdb -l ligand.sdf -a genetic --physics-based -s -15.7 -17.7 8.1 --grid-radius 10.0 --grid-spacing 0.375  --use-gpu

# Physics-Based docking (PANDADOCK + MMGBSA scoring)
pandadock -p protein.pdb -l ligand.sdf -a pandadock --physics-based
pandadock -p protein.pdb -l ligand.sdf -a pandadock --physics-based -s -15.7 -17.7 8.1 --grid-radius 10.0 --grid-spacing 0.375 

# Use GPU
pandadock -p protein.pdb -l ligand.sdf --use-gpu --physics-based -s -15.7 -17.7 8.1 --grid-radius 10.0 --grid-spacing 0.375 

# Automatic algorithm selection
pandadock -p protein.pdb -l ligand.sdf --auto-algorithm -s -15.7 -17.7 8.1 --grid-radius 10.0 --grid-spacing 0.375 

PandaDock offers flexible running modes to balance between speed and accuracy:

• Fast Mode: --fast-mode - Quick docking with minimal enhancements
• Default Mode: Basic docking with standard scoring
• Enhanced Mode: Use any combination of enhancement flags:
  • --enhanced-scoring - More accurate scoring with electrostatics
  • --local-opt - Fine-tune top poses for better results
  • --exhaustiveness 5 - Run multiple independent searches
  • --prepare-molecules - Optimize molecule geometry before docking
  • --flex-residues - Use flexible residues
• Physics Mode: More advanced algorithm
• Pandadock: Pandadock algorithm with conformer generation

After running PandaDock, a comprehensive report is automatically generated:

├── binding_affinity_report.csv
├── binding_affinity_report.txt
├── complex_pose_1_score:-14.25.pdb
├── complex_pose_2_score:-14.07.pdb
├── complex_pose_3_score:-13.94.pdb
├── complex_pose_4_score:-12.35.pdb
├── detailed_docking_report.txt
├── docking_report.html
├── docking_report.json
├── docking_results.csv
├── docking_scores.txt
├── energy_breakdown.csv
├── parameters.txt
├── plots/
│   ├── combined_metrics_vs_rank.png
│   ├── energy_breakdown.png
│   ├── energy_stacked.png
│   ├── ic50_vs_deltag.png
│   ├── kd_vs_deltag.png
│   ├── score_analysis.png
│   ├── score_distribution.png
│   ├── score_improvement.png
│   ├── score_plot.png
│   ├── score_rank.png
├── pose_1_score:-14.2.pdb
├── pose_2_score:-14.1.pdb
├── pose_3_score:-13.9.pdb
├── pose_4_score:-12.3.pdb
├── sphere.pdb
└── status.json

Even generates a grid sphere where you define your grid center Example:

• 📄 Reports in CSV, TXT, JSON, and interactive HTML formats
• 📈 Analysis plots (binding energy breakdowns, score distributions, ranking curves)
• 🧬 Pose files (.pdb) for each predicted ligand conformation
• 🛠️ Parameters and status tracking for reproducibility

One example output in detailed_docking_report.txt is appened below:

ALGORITHM DETAILS
-----------------
Algorithm: Genetic
Population Size: 100
Scoring Function: Advanced
Hardware Acceleration: GPU (ID: 0)
GPU Precision: float32
Exhaustiveness: 1
RESULTS SUMMARY
--------------
Total Poses Generated: 4
Best Score: -14.2475
Worst Score: -12.3464
Average Score: -13.6515

Score Standard Deviation: 0.7612

TOP 10 POSES
--------------
Rank	Score	File
ENERGY COMPONENT BREAKDOWN (TOP 10 POSES)
----------------------------------------
Pose       Clash    Desolvat    Electros     Entropy      H-Bond    Hydropho    Van der 
----------------------------------------------------------------------------------------
   1       -0.46       -0.00       -0.09       -0.12       -0.09       -0.09       -0.14
   2       -0.93       -0.00       -0.19       -0.23       -0.19       -0.19       -0.28
   3       -1.39       -0.01       -0.28       -0.35       -0.28       -0.28       -0.42
   4       -1.86       -0.01       -0.37       -0.46       -0.37       -0.37       -0.56

BINDING AFFINITY ESTIMATION
---------------------------
Pose 1: Score = -14.25 kcal/mol
  Estimated ΔG: -14.25 kcal/mol
  Estimated Kd: 3.60e-11 M
  Estimated IC50: 7.20e-11 M

Pose 2: Score = -14.07 kcal/mol
  Estimated ΔG: -14.07 kcal/mol
  Estimated Kd: 4.87e-11 M
  Estimated IC50: 9.75e-11 M

Pose 3: Score = -13.94 kcal/mol
  Estimated ΔG: -13.94 kcal/mol
  Estimated Kd: 6.02e-11 M
  Estimated IC50: 1.20e-10 M

Pose 4: Score = -12.35 kcal/mol
  Estimated ΔG: -12.35 kcal/mol
  Estimated Kd: 8.91e-10 M
  Estimated IC50: 1.78e-09 M
════════════════════════════════════════════════════════════════════════════════
    Binding Affinity Report
================================================================================

Pose         ΔG (kcal/mol)              Kd (M)            IC50 (M)
------------------------------------------------------------------
1                   -14.25            3.60e-11            7.20e-11
2                   -14.07            4.87e-11            9.75e-11
3                   -13.94            6.02e-11            1.20e-10
4                   -12.35            8.91e-10            1.78e-09

# Random Search Docking
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a random -i 100

# Genetic Algorithm Docking (Default)
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a genetic -i 100

# Auto Algorithm Selection
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --auto-algorithm

# CPU Physics-Based Docking
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a genetic --physics-based

# GPU-Accelerated Physics-Based Docking
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a genetic --physics-based --use-gpu

# Monte Carlo Sampling with Genetic Optimization
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a genetic --monte-carlo

# Monte Carlo + Physics-Based Scoring
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a genetic --monte-carlo --physics-based

# PANDADOCK Docking (Default Parameters)
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a pandadock

# PANDADOCK + Physics-Based Scoring
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a pandadock --physics-based

# PANDADOCK Using Grid-Based Energies
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a pandadock --use-grid

# Fine-Tune PANDADOCK MD Parameters
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a pandadock --high-temp 500 --target-temp 300 --cooling-factor 0.95

# GPU-Only Docking
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a genetic --use-gpu

# CPU/GPU Hybrid Workload Balancing
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 -a genetic --use-gpu --workload-balance 0.8

# GPU Float64 Precision Docking
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --use-gpu --gpu-precision float64

# Gradient-Based Search
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --advanced-search gradient

# Replica Exchange Monte Carlo
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --advanced-search replica-exchange

# ML-Guided Docking (Random Forest Surrogate)
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --advanced-search ml-guided --surrogate-model rf

# Fragment-Based Docking
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --advanced-search fragment-based

# Hybrid Genetic Algorithm + L-BFGS Docking
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --advanced-search hybrid

# Generate Detailed HTML Analysis Report
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --generate-analysis-report --analysis-report-format html

# Save Results as Text, CSV, and JSON
pandadock -p protein.pdb -l ligand.sdf -s -15.7 -17.7 8.1 --report-format all

✅ Always add --prepare-molecules to automatically protonate and minimize molecules.
✅ Use --enhanced-scoring or --physics-based for better electrostatics (even in non-physics mode).
✅ Use --local-opt to refine poses with local energy minimization.
✅ Add --generate-analysis-report for detailed visual inspection after docking.

PANDADOCK is a CHARMm-inspired docking algorithm that combines conformational sampling with molecular dynamics simulation. The protocol includes:

1. Conformer Generation: Multiple ligand conformers are generated by systematic rotation of rotatable bonds
2. Random Orientation: Each conformer is positioned in the binding site with multiple initial orientations
3. Simulated Annealing: Poses are refined using simulated annealing molecular dynamics
4. Final Minimization: A final energy minimization optimizes the poses
5. Scoring and Ranking: All poses are scored and ranked using the physics-based scoring function

(Over 70+ options for full control!)

▶️ See full list below

PandaDock includes physics-based molecular modeling capabilities that significantly enhance the accuracy of docking results:

• MMFF94 Force Field Minimization: Full molecular mechanics energy minimization
• Enhanced Electrostatics: Poisson-Boltzmann inspired model with distance-dependent dielectric
• Implicit Solvation (GB/SA): Generalized Born model with surface area-based nonpolar term
• Monte Carlo Sampling: Enhanced conformational sampling with Metropolis criterion
• Physics-Based Scoring: Complete MM-GBSA inspired scoring function

• Physics-Based Energy Decomposition:
  • VDW, Electrostatics, H-Bond, Desolvation, Hydrophobic, Clash, Entropy
• Replica-Exchange Docking (Multiple temperatures)
• Gradient Descent (L-BFGS-B) Local refinement
• ML-Guided Docking (Experimental)
• Rotamer sampling for flexible residues
• Tethered docking (if a reference ligand is provided)

Yes! You can absolutely showcase multiple PandaDock Python APIs in your README or docs for different use cases.

Here’s a well-structured section with three API blocks you can include:

from pandadock.protein import Protein
from pandadock.ligand import Ligand
from pandadock.unified_scoring import EnhancedScoringFunction
from pandadock.search import GeneticAlgorithm
from pandadock.utils import save_docking_results

protein = Protein("protein.pdb")
ligand = Ligand("ligand.sdf")
protein.define_active_site([10.0, 20.0, 30.0], 12.0)

scoring = EnhancedScoringFunction()
search = GeneticAlgorithm(scoring, max_iterations=1000, population_size=100)

results = search.search(protein, ligand)
save_docking_results(results, "docking_results")

from pandadock import Protein, Ligand
from pandadock.unified_scoring import PhysicsScoringFunction
from pandadock.pandadock import PandaDockSearch

protein = Protein("protein.pdb")
ligand = Ligand("ligand.sdf")
protein.auto_detect_pocket()  # Auto-detect binding pocket

scoring = PhysicsScoringFunction()
search = PandaDockSearch(scoring, md_steps=100, minimize_steps=50)

results = search.search(protein, ligand)

from pandadock import Protein, Ligand
from pandadock.unified_scoring import EnhancedScoringFunction
from pandadock.search import GeneticAlgorithm

protein = Protein("protein.pdb")
ligand = Ligand("ligand.sdf")
protein.define_active_site([35.0, 22.0, 18.0], 10.0)
protein.auto_define_flexible_residues(max_residues=5)

scoring = EnhancedScoringFunction()
search = GeneticAlgorithm(scoring, max_iterations=500)

results = search.search(protein, ligand)

from pandadock.batch_screening import run
from pandadock.utils import setup_logging

# Define batch screening configuration
config = {
    "protein": "protein.pdb",                     # Path to protein file
    "ligand_library": "ligands/",                  # Folder containing many ligands (MOL/SDF)
    "output_dir": "batch_screening_results",       # Output folder
    "screening_params": {
        "hardware": {
            "use_gpu": True,                       # Use GPU if available
            "gpu_id": 0,
            "gpu_precision": "float32"
        },
        "scoring": "enhanced",                      # Use 'basic', 'enhanced', or 'physics-based'
        "algorithm": "genetic",                     # Docking algorithm: 'genetic', 'random', 'pandadock', etc.
        "iterations": 500,
        "population_size": 50,
        "detect_pockets": True,                     # Auto detect binding site
        "prepare_molecules": True                   # Auto prepare ligands
    }
}

# Setup logger (optional)
setup_logging("batch_screening_results")

# Run batch screening
results = run(config)

# Access results
for ligand_name, info in results.items():
    print(f"Ligand: {ligand_name}, Best Score: {info['score']:.2f}, Runtime: {info['runtime']:.2f} sec")

PandaDock incorporates concepts from several established molecular docking and computational chemistry packages:

• CDOCKER (Wu et al. 2003)
• AutoDock Vina (Trott & Olson, 2010)
• DOCK (Allen et al., 2015)
• RDKit (http://www.rdkit.org)

MIT License — free for academic and commercial use.

If you use PandaDock in your research:

Pritam Kumar Panda (2025). PandaDock: Python-Based Molecular Docking. GitHub.
https://github.com/pritampanda15/PandaDock

• GitHub: @pritampanda15
• Email: [email protected]
• Issue Tracker: Open an Issue

PandaDock is intended for research purposes.
Always verify docking predictions through experimental validation.

Dock Smarter. Discover Faster