PocketDock¶

An automated, web-based molecular docking pipeline.

PocketDock combines P2Rank for binding pocket prediction with AutoDock Vina for ligand docking, then renders the results in your browser with an interactive 3Dmol.js viewer. Upload a protein and a ligand — or a whole batch of ligands — get ranked docking poses with detailed binding-site interaction analysis a few minutes later.

PocketDock results page

What you can do with PocketDock¶

Predict binding pockets

P2Rank scans the protein surface and ranks druggable pockets by probability — no need to know the binding site in advance.
Dock single ligands or batches

Submit one ligand for a focused run, or up to 100 in a single batch — multi-molecule SDFs are split automatically and tracked on a batch dashboard.
Ensemble docking

Generate 2–10 receptor conformations with normal-mode analysis (NMA, ~30 s) or short OpenMM MD (5–15 min), then dock across all of them with consensus scoring.
ADMET property panel

Every job ships with RDKit-computed drug-likeness descriptors — MW, logP, TPSA, QED, plus Lipinski and Veber pass/fail flags.
Pose refinement & MM-GBSA rescoring

Optionally minimize each pose with OpenMM (AMBER14 + implicit solvent), and add an MM-GBSA-style per-pose ΔG column to the results table.
Analyze interactions interactively

Inspect H-bonds, hydrophobic contacts, salt bridges, π-stacking, π-cation, and halogen bonds in 3D — toggle each type, customize distance thresholds, export PNGs and CSVs.

Quick links¶

Getting Started — install with Docker and run your first job
Concepts — what pockets, poses, and binding affinities actually mean
The Results Page — tour of every panel and control
Batch Docking — screen libraries with one submission
Ensemble Docking — flexible-receptor docking via NMA or MD
ADMET Properties — read the drug-likeness panel
MM-GBSA Rescoring — opt-in per-pose ΔG
Interpreting Results — how to read the affinity, combined score, and Kd estimates
API Reference — script PocketDock from Python or curl

How the pipeline works¶

flowchart LR
    A[Upload protein + ligand] --> B0{Ensemble?}
    B0 -- yes --> B1[NMA / MD<br/>conformations]
    B0 -- no --> B
    B1 --> B
    B[P2Rank<br/>pocket detection] --> C[Meeko<br/>structure prep]
    C --> CA[RDKit<br/>ADMET]
    CA --> D[AutoDock Vina<br/>docking]
    D --> E[Interaction<br/>analysis]
    E --> F{Refine?}
    F -- yes --> F1[OpenMM<br/>minimization]
    F -- no --> G
    F1 --> G
    G{MM-GBSA?}
    G -- yes --> G1[Per-pose<br/>ΔG rescoring]
    G -- no --> H
    G1 --> H
    H[Interactive results]

Each stage runs as a Celery task; the web UI polls the job status every 5 seconds and redirects to the results page when docking completes.

License¶

PocketDock is released under the MIT License.