Genome Sentinel is a self-contained, offline-capable molecular docking workspace. It combines an automated Python pipeline for protein and ligand preparation with a modern web dashboard featuring real-time 3D visualization via 3Dmol.js.
No cloud dependencies. No subscriptions. Runs entirely on your laptop.
- Curated Target Presets — Alzheimer's (BACE1 / 2B8L), Malaria (PfDHFR / 1LD3), Diabetes (DPP-4 / 2OQV), Breast Cancer (HER2 / 3PP0) with pre-configured binding site coordinates
- Built-in Ligand Library — 7 FDA-approved drugs (Donepezil, Chloroquine, Metformin, Lapatinib, Aspirin, Acetaminophen, Ibuprofen) with automatic PubChem 3D structure fetching
- Custom PDB & SMILES Support — Download any RCSB PDB target by ID; input any custom SMILES ligand
- AutoDock Vina Integration — Automated execution of rigid-receptor molecular docking with full log parsing and binding affinity extraction
- AI Molecule Generation — Generate novel drug-like molecules via NVIDIA GenMol NIM integration (requires free NVIDIA API key)
- Interactive 3D Viewer — 3Dmol.js-based protein cartoon + docked ligand sticks with style controls (cartoon, sphere, stick, line)
- Built-in Manuscript Generator — Auto-compiles results into a markdown academic paper draft
- No RDKit/Meeko Required — Uses PubChem API + direct PDBQT conversion as fallback (RDKit C++ DLLs commonly blocked on Windows)
 |
| Dashboard & Environment Status |
 |
| Docking Grid Configuration |
 |
| 3D Complex Visualization |
Genome Sentinel/
├── bin/ # AutoDock Vina executable
├── data/ # Working data
│ ├── proteins/ # Downloaded & prepared target structures (.pdb, .pdbqt)
│ ├── ligands/ # Prepared ligand libraries (.pdbqt)
│ └── results/ # Docking output poses + logs + summary.json
├── pipeline/ # Core Python pipeline
│ ├── setup_env.py # Download Vina + install Python packages
│ ├── prep_protein.py # RDKit-based protein preparation (original)
│ ├── prep_protein_no_rdkit.py # Protein preparation without RDKit (fallback)
│ ├── prep_ligands.py # RDKit/Meeko-based ligand preparation (original)
│ ├── prep_ligands_no_rdkit.py # Ligand preparation via PubChem API (fallback)
│ └── run_docking.py # AutoDock Vina execution + log parser
├── app/ # Web dashboard
│ ├── index.html # Single-page application
│ ├── style.css # Clean professional UI theme
│ ├── app.js # SPA logic, API client, 3Dmol.js integration
│ └── 3Dmol-min.js # Local copy of 3Dmol.js (no CDN needed)
├── server.py # Zero-dependency HTTP server + REST API
├── run.bat # Double-click launcher (Windows)
├── implementation_plan.md # Original design document
└── README.md # This file
- Windows 10/11 (the pipeline uses
vina.exefor Windows; Linux/macOS would need a different Vina binary) - Python 3.12+ with
pip
# 1. Clone or download this repository
cd Genome-Sentinel
# 2. Install Python dependencies
pip install requests
# 3. (Recommended) Install RDKit and Meeko for the primary pipeline
pip install rdkit meekoOption A — Double-click run.bat
Option B — From PowerShell / Terminal
python server.pyThen open http://localhost:8000 in your browser.
The server will automatically download AutoDock Vina v1.2.5 on first launch when you click Initialize Environment in the dashboard.
- Open the dashboard at
http://localhost:8000 - Click Control Center → Check Status to verify Vina and Python are ready
- Go to Step 1 → Select BACE1 (2B8L) preset → Click Download
- Go to Step 2 → Click Prepare Preset Library (fetches 7 drugs from PubChem)
- Go to Step 3 → Grid coordinates auto-fill for the selected preset
- Go to Step 4 → Select protein + ligand → Run Screen
- Go to Step 5 → View binding scores → Click View 3D to visualize
| Method | Endpoint | Description |
|---|---|---|
GET |
/api/status |
Environment status (Vina, RDKit, Meeko, pipeline readiness, GenMol) |
POST |
/api/genmol/generate |
Generate molecules via NVIDIA GenMol NIM (requires api_key) |
POST |
/api/setup |
Download Vina binary + install pip packages |
POST |
/api/download_protein |
Download PDB by ID, clean, convert to PDBQT |
POST |
/api/prep_presets |
Prepare all 7 built-in ligand presets |
POST |
/api/prep_custom_ligand |
Prepare a custom ligand from name + SMILES |
POST |
/api/run_docking |
Execute AutoDock Vina with specified grid parameters |
GET |
/api/results |
Get all docking results summary |
GET |
/api/log?file=... |
Get raw log file content |
POST |
/api/delete_protein |
Delete prepared protein files |
POST |
/api/delete_ligand |
Delete prepared ligand files |
POST |
/api/clear_results |
Clear all docking results |
POST |
/api/download_manuscript |
Save manuscript markdown to data/ |
POST |
/api/genmol/generate |
Generate molecules via NVIDIA GenMol NIM |
GET |
/data/... |
Static file serving (proteins, ligands, results) |
| PDB ID | Protein | Disease | Grid Center (x, y, z) | Box Size |
|---|---|---|---|---|
| 2B8L | BACE1 | Alzheimer's | (16.0, 10.0, 15.0) | 22 Å |
| 1LD3 | PfDHFR | Malaria | (32.5, 14.8, -3.2) | 22 Å |
| 2OQV | DPP-4 | Diabetes | (40.1, 38.5, 50.3) | 22 Å |
| 3PP0 | HER2 | Breast Cancer | (14.8, 17.5, 94.6) | 22 Å |
| Ligand | SMILES | Target Context |
|---|---|---|
| Donepezil | COC1=...CC4=CC=CC=C4)OC |
Alzheimer's (BACE1 inhibitor) |
| Chloroquine | CCN(CC)CCCC(C)NC1=... |
Malaria (PfDHFR candidate) |
| Metformin | CN(C)C(=N)N=C(N)N |
Diabetes (DPP-4 pathway) |
| Lapatinib | CS(=O)(=O)CCNCC1=... |
Breast Cancer (HER2 inhibitor) |
| Aspirin | CC(=O)OC1=CC=CC=C1C(=O)O |
Control anti-inflammatory |
| Acetaminophen | CC(=O)NC1=CC=C(O)C=C1 |
Control analgesic |
| Ibuprofen | CC(C)CC1=CC=C(C=C1)C(C)C(=O)O |
Control NSAID |
- Protein Download — RCSB PDB fetched by ID; water molecules and crystallization agents removed; converted to PDBQT format
- Ligand Preparation — 3D coordinates fetched from PubChem (or generated via RDKit if available); converted to PDBQT with Vina-compatible atom types
- Grid Configuration — Search box centered on the binding pocket; size and exhaustiveness configurable
- Docking Execution — AutoDock Vina 1.2.5 performs rigid-receptor docking; outputs up to 9 binding modes with RMSD clustering
- Results — Binding affinities displayed in a sortable table; 3D complex viewable with multiple rendering styles
Some Windows configurations block the C++ extension DLLs shipped with RDKit/Meeko (Windows Defender Application Control / Smart App Control). Genome Sentinel includes fallback scripts that work without RDKit:
- Protein preparation: Strips water/heteroatoms from raw PDB → direct PDBQT conversion with element-based atom typing
- Ligand preparation: Fetches 3D SDF coordinates from PubChem's public API → writes Vina-compatible PDBQT
The pipeline_ready status field reports true when either the primary (RDKit+Meeko) or fallback pipeline is available.
Genome Sentinel integrates with NVIDIA GenMol NIM for AI-powered de novo molecule generation.
- Get a free API key from NVIDIA build.nvidia.com
- Set it as an environment variable:
$env:NGC_API_KEY = "nvapi-..."or enter it directly in the dashboard UI
- Go to Step 2: Ligand Library
- Find the Generate with AI (GenMol) card
- (Optional) Enter a scaffold SMILES for decoration, or leave blank for fully novel molecules
- Choose molecule count and scoring function (QED for drug-likeness, LogP for lipophilicity)
- Click Generate Molecules — GenMol produces novel drug-like candidates
- Click Prepare on any molecule to add it to your ligand library and proceed to docking
- GenMol uses SAFE notation internally; scaffold SMILES are automatically converted
- Molecules are scored by QED (0-1, higher = more drug-like) or LogP
- Generated molecules can be docked against any prepared protein target using AutoDock Vina
- Requires network access to
https://health.api.nvidia.com
- Backend: Python 3.12 standard library (
http.server— zero dependencies) - Docking Engine: AutoDock Vina 1.2.5 (auto-downloaded)
- Frontend: Vanilla JS SPA, clean professional CSS theme with onboarding tour, step progress bar, and elapsed-time progress indicators
- 3D Visualization: 3Dmol.js (local copy, no CDN)
- 3D Structure Source: PubChem PUG REST API (free, no API key)
MIt