🧪 Automated Virtual Screening Pipeline

A modular and reproducible virtual screening system for drug discovery.
This pipeline integrates ADMET filtering, 3D structure generation, ligand preparation, and docking with AutoDock Vina, with robust provenance tracking and clean restarts.

---

📂 Pipeline Overview

The system is divided into four modules, each of which can be run independently or sequentially:

1. Module 1 – ADMET Screening

   • Input: input/input.csv (id, SMILES, notes, params_json)
   • Computes descriptors via RDKit (MW, logP, TPSA, HBD, HBA, RotB, InChIKey).
   • Applies drug-likeness rules (Lipinski, Veber, optional Egan/Ghose).
   • Outputs:
     • output/admet.csv
     • state/admet_pass.list, state/admet_fail.list
     • Updates state/manifest.csv

2. Module 2 – SMILES → 3D (SDF)

   • Uses RDKit ETKDGv3 + UFF/MMFF minimization.
   • Converts SMILES into clean 3D .sdf files.
   • Outputs:
     • 3D_Structures/<id>.sdf, <id>.smi
     • Logs (<id>_rdkit.log)
     • Updates manifest with sdf_* status

3. Module 3 – SDF → PDBQT (Ligand Prep)

   • Uses Meeko to prepare ligands for docking.
   • Outputs:
     • prepared_ligands/<id>.pdbqt
     • Logs (<id>_meeko.log)
     • Updates manifest with pdbqt_* status

4. Module 4 – Docking with AutoDock Vina

   • Runs AutoDock Vina using VinaConfig.txt placed next to the binary.
   • Performs docking against target receptor.
   • Outputs:
     • results/<id>_out.pdbqt (poses)
     • results/<id>_vina.log
     • results/summary.csv (scores + metadata)
     • results/leaderboard.csv (ranked ligands)
     • Updates manifest with vina_* status

---

🧹 Utility Script

WARNING_PURGE_PIPELINE.py

• Cleans pipeline outputs for a fresh run.
• Preserves folder structure + CSV headers.
• Deletes generated ligands/receptors/results.
• Keeps VinaConfig.txt intact.

---

📊 Manifest Tracking

A central state/manifest.csv records per-ligand progress through the pipeline:

• ADMET: status/reason, descriptors, InChIKey
• SDF: file path, validation reason
• PDBQT: ligand prep status
• Vina: docking score, pose path, receptor SHA1
• Provenance: config hashes, RDKit/Meeko/Vina versions, timestamps

This makes the pipeline idempotent (safe to re-run without duplication).

---

⚙️ Installation

Requirements

• Python 3.10+
• RDKit
• Meeko
• AutoDock Vina (binary placed in project root)
• PyYAML (optional, for configs)

Clone and Setup

git clone https://github.com/yourusername/virtual-screening-pipeline.git
cd virtual-screening-pipeline
pip install -r requirements.txt

(Create a requirements.txt like below)

rdkit-pypi
meeko
pyyaml

---

🚀 Usage

1. Prepare Input

Create input/input.csv with columns:

id,smiles,notes,params_json

2. Run Modules in Order

# 1. ADMET filtering
python Module\ 1.py

# 2. SMILES → 3D SDF
python Module\ 2.py

# 3. Ligand preparation
python Module\ 3.py

# 4. Docking
python Module\ 4.py

3. Reset Pipeline

python WARNING_PURGE_PIPELINE.py

---

📂 Project Structure

├── input/                  # input.csv (SMILES)
├── output/                 # admet.csv
├── 3D_Structures/          # generated SDFs
├── prepared_ligands/       # ligand PDBQTs
├── results/                # docking poses, summary, leaderboard
├── state/                  # manifest.csv, pass/fail lists
├── config/                 # optional run.yml, machine.yml
├── logs/                   # logs from modules
├── Module 1.py             # ADMET filtering
├── Module 2.py             # SMILES → SDF
├── Module 3.py             # SDF → PDBQT
├── Module 4.py             # Docking
└── WARNING_PURGE_PIPELINE.py  # reset script

---

🔑 Key Features

• ✅ Modular: each stage can be run independently
• ✅ Validated: descriptors, SDF, PDBQT, docking pose checks
• ✅ Idempotent: safe re-runs without duplication
• ✅ Atomic writes: avoids corrupted/empty files
• ✅ Provenance: config hash + tool versions logged
• ✅ Resettable: one-command clean restart

---

📜 License

MIT License – feel free to use and adapt.

LigandForge