Welcome to PandaDock Documentation

PandaDock Logo

PandaDock is a next-generation molecular docking platform featuring an SE(3)-equivariant Graph Neural Network scoring function that achieves R > 0.8 correlation with experimental binding affinities.

Note

PandaDock v4.0 features the PandaDock-GNN scoring function with state-of-the-art performance: Pearson R = 0.88 on PDBbind, R = 0.82 on ULVSH test set, and R = 0.68 on novel GABA receptor dataset. Significantly outperforms all baseline methods including Vina, Gnina, and MM-GBSA.

Quick Start

Install PandaDock using pip:

git clone https://github.com/pritampanda15/PandaDock.git
cd PandaDock
pip install -e ".[gnn]"

Basic usage:

# Traditional docking with Vina-style scoring
pandadock dock -r protein.pdb -l ligand.sdf \
               --center 10 20 30 --box 20 20 20

# Hybrid docking with GNN rescoring (RECOMMENDED)
pandadock hybrid -r protein.pdb -l ligand.sdf \
                 --center 10 20 30 --box 20 20 20 \
                 --model models/best_model.pt

Key Features

🧠 SE(3)-Equivariant GNN Scoring (NEW)

  • Rotation and translation invariant predictions

  • Heterogeneous graph representation (protein + ligand)

  • Multi-task learning: pEC50 regression + activity classification

  • Pearson R > 0.8 on held-out test sets

  • Outperforms all traditional scoring methods

🔬 Advanced Docking Algorithm

  • Hierarchical Search: Multi-resolution coarse-to-fine sampling

  • Vina-style empirical scoring function

  • Physics-based scoring with Lennard-Jones + electrostatics

⚡ Hybrid Workflow (Recommended)

  • Traditional pose generation + GNN rescoring

  • Combines speed of physics-based docking with GNN accuracy

  • Best of both worlds approach

🎯 Specialized Docking Modes

  • Flexible Docking (pandadock-flex): Induced-fit with receptor flexibility

  • Metal Docking (pandadock-metal): Specialized for metalloproteins (Zn, Fe, Mg, Ca, etc.)

  • Tethered Docking (pandadock-tethered): Constrained near reference positions

  • Analysis & Reporting (pandadock-report): Publication-ready figures

📊 Benchmark Performance

  • PandaDock-GNN outperforms 8+ baseline methods on ULVSH dataset

  • Better than VM2, MMGBSA, Gnina, Hyde, and other scoring functions

Performance Benchmarks

PDBbind Benchmark (5,316 complexes):

Method

Type

Pearson R

PandaDock-GNN

SE(3)-EGNN

0.88 ⭐

OnionNet-2

3D-CNN

0.86

RF-Score v3

Random Forest

0.80

AutoDock Vina

Empirical

0.60

ULVSH Benchmark (942 compounds, 10 targets):

Method

Type

Pearson R

PandaDock-GNN (test set)

ML Scoring

0.82 ⭐

VM2

Free energy

0.15

PM6

Semi-empirical

0.08

Gnina, MMPBSA, MMGBSA, Vina

Baselines

< 0.02

BindingDB Benchmark (8,891 complexes):

Training Configuration

Type

Pearson R

BindingDB Only

ML Scoring

0.81 ⭐

BindingDB + ULVSH

ML Scoring

0.79

Documentation Contents

Getting Started

PandaDock-GNN

Algorithms

Command Line Interface

About

Indices and tables