StrataDock was built around a practical frustration: molecular docking workflows are often scattered across scripts, binary installs, manual file preparation, and one-off result folders. The goal was to turn that into a Linux-first docking workstation that a user can run repeatedly.
The app supports the messy real case. A user might have a receptor PDB, a few ligands, no reference ligand, and no polished docking box. StrataDock gives them a structured path from upload to ranked results.
What the workflow handles
The workflow covers receptor cleanup, ligand preparation, binding-site definition, docking, scoring, visualization, and export. It supports AutoDock Vina by default and GNINA as an optional CPU/CUDA path.
Binding-site setup is deliberately flexible:
- reference ligand mode for redocking and known binding sites
- fpocket mode when only a receptor structure is available
- manual box mode when coordinates are already known
That design makes the app useful beyond a demo dataset, because real docking users often do not start with a perfect co-crystal setup.
Validation and results
The repo includes five co-crystal redocking validation systems:
- HIV protease, 1HSG
- EGFR kinase, 1M17
- ABL kinase, 1IEP
- trypsin, 3PTB
- HSV thymidine kinase, 1KIM
Each validation case includes receptor files, native ligand files, ligand SDF inputs, docking box coordinates, and acceptance checks such as pose RMSD thresholds, score presence, and pose-file existence.
That validation set is the most important result of the engineering work. It turns StrataDock from a UI wrapper into a workflow with testable behavior.
Engineering choices
The repo is split into core modules for docking, reports, pockets, receptors, ligands, sessions, and visualization. The Streamlit app is large, but the important logic lives in reusable modules under src/stratadock/core.
The test suite covers ADMET-style profiles, batch runs, binding boxes, diagnostics, GNINA handling, pocket prediction, receptor and ligand preparation, reports, sessions, UI tables, and validation data.
Why this is portfolio-worthy
This project shows a different kind of bioinformatics skill: not just running an analysis, but building an actual tool. It combines scientific domain logic, workflow ergonomics, UI design, validation data, and reproducible exports.
For a future version, I would keep simplifying the Streamlit app into smaller panels and add more visual regression checks around the UI. The core direction is solid: make docking practical, repeatable, and inspectable.