Molecular Docking & MD Pipeline

A high-performance, automated pipeline for large-scale virtual screening and molecular dynamics simulations. Built with Nextflow, it leverages GPU-accelerated docking engines and provides integrated post-docking analysis and MD workflows.

Quick Start

The fastest way to get started is using the one-shot installer.

# Clone the repository
git clone https://github.com/jsphu/docking_pipeline.git
cd docking_pipeline

# Run the installer (Docker mode is recommended)
sudo ./install.sh --mode docker

# Run a sample docking session
nextflow run main.nf --smiles_file data/sample_ligands.smi --receptor data/receptor.pdbqt

Key Features

• GPU Acceleration: Support for AutoDock-GPU, Vina-GPU, and QuickVina-GPU for ultra-fast screening.
• Flexible Inputs: Support for 2D (SMILES) and 3D (PDBQT) ligand inputs.
• Automated Downloader: Directly fetch ligands from ZINC or other URI links.
• Containerized: Full support for Docker and Singularity, ensuring reproducibility and easy deployment.
• Post-Docking Suite: Integrated scripts for filtering (Lipinski, PAINS), scoring, and lead collection.
• Automated MD: One-command Molecular Dynamics workflow using GROMACS (via md_workflow.py).

Installation & Setup

Prerequisites

• Linux (Ubuntu/Debian recommended) or WSL2.
• NVIDIA GPU with latest drivers (required for GPU acceleration).
• Docker or Singularity/Apptainer (optional but highly recommended).

Using the Installer

The install.sh script automates the installation of Java, Nextflow, Docker, and the NVIDIA Container Toolkit.

Mode	Description
docker	(Recommended) Installs Docker, NVIDIA toolkit, and pulls images from GHCR.
wsl	For WSL2 users with Docker Desktop on Windows.
native	Builds everything on the host (Nextflow, Java, QuickVina-GPU binaries).

# Example: Install for Linux with Docker
sudo ./install.sh --mode docker

Usage Guide

1. Basic Docking (CPU/Vina)

Run docking on a local SMILES file using standard AutoDock Vina.

nextflow run main.nf \
  --smiles_file data/ligands.smi \
  --receptor data/receptor.pdbqt \
  --outdir results

2. GPU Accelerated Docking (QuickVina-GPU)

Enable GPU acceleration and provide a specialized configuration.

nextflow run main.nf \
  --use_gpu \
  --smiles_file data/ligands.smi \
  --receptor data/receptor.pdbqt \
  -c config/5TBM-GPU-ACCELERATED.config

3. Downloading from ZINC

Specify a file containing download links (URIs) to fetch and dock ligands directly.

nextflow run main.nf --links_file data/ZINC-links.uri --use3d_downloader

⚙️ Main Parameters

Parameter	Default	Description
--receptor	data/...	Path to the receptor PDBQT file.
--smiles_file	''	Input SMILES file for local screening.
--pdbqt_file	''	Input PDBQT file for local screening.
--use_gpu	false	Enable GPU-accelerated docking.
--center_x/y/z	-	Coordinates of the docking box center.
--size_x/y/z	20.0	Dimensions of the docking box (Angstroms).
--exhaustiveness	8	Search exhaustiveness for Vina.

Post-Docking Analysis

After docking, use the provided Python suite to filter and organize your results.

1. Filter Ligands: Applies Lipinski rules (scripts/rules.txt) and PAINS filters to remove false positives.

   python3 scripts/filter_ligands.py
   python3 scripts/pains_filter.py

2. Prepare SwissADME: Standardize SMILES for online medicinal chemistry tools.

   python3 scripts/prepare_swissadme.py

3. Collect Leads: Organizes top-performing PDBQT poses into a single directory for visualization.

   python3 scripts/collect_leads.py

Molecular Dynamics (MD)

Perform automated GROMACS simulations for your top lead compounds:

cd scripts/md_workflow
python3 md_workflow.py \
  --protein protein.pdb \
  --ligand ligand.pdbqt \
  --config config.json \
  --docker

For more details, see the MD Workflow documentation.

License

This project is licensed under the LICENSE - see the file for details.