Ultidock v1.1.2
Open-source molecular docking

Molecular docking.Automated
at scale.

From receptor preparation to ranked poses.
Ultidock connects automatic binding-site discovery, multi-GPU docking, and structured results in one reproducible workflow.

Get started Explore the source
Linux · Python 3.10+ · MIT licensed
CaV-EMPS / SITE DISCOVERY[ RECEPTOR → BOX ]
Conceptual binding-site discoveryAn abstract contour surface surrounds a highlighted pocket and ligand-like shape, with a wireframe docking box. This is a schematic, not a calculated protein structure. POCKET SURFACE zxy
PHYSICS + GEOMETRYConceptual illustration
A connected research workflowmolguardCaV-EMPSAutoDock-GPUAutoDock VinaSQLite
01 / Architecture

The whole screening workflow.
Connected, from the start.

Prepare structures, propose search boxes, schedule docking, and collect results on a Linux workstation or HPC compute node.

Reliable inputs.

molguard

Deterministic receptor handling, fixed-width coordinate formatting, and input validation help make preparation traceable and repeatable.

Explore preparation

Discover the site.

CaV-EMPS

Hybrid physics–geometry pocket proposals combine geometric cavity information with AutoGrid4 interaction maps. Optional fpocket and P2Rank methods are also available.

Explore site discovery

Scale the screen.

AutoDock-GPU / Vina

Bounded workers and per-device concurrency controls coordinate docking. CUDA mode distributes work across detected NVIDIA GPUs; CPU Vina is also supported.

Explore batch screening

Keep results connected.

SQLite + reports

Save parsed docking records as the run progresses. Keep scores, binding sites, and pose paths together for inspection and downstream analysis.

Explore results
From structure
to scored poses
  1. Receptors + ligandsStage prepared inputs
  2. Validate & preparemolguard
  3. Propose sitesCaV-EMPS / adapters
  4. Run dockingGPU or CPU engines
  5. Inspect resultsSQLite / pose files
02 / In the terminal

A clear path to
your first run.

Start with the guided example, then move to your own receptors and ligand library. The same CLI supports interactive learning and unattended screening.

  • Guided examples with real docking inputs
  • Receptor-only binding-site proposals
  • Incremental results during each run
Follow the quickstart
ultidock / bash
# After installing Ultidock and native tools
$ ultidock doctor

# Read the guided lesson without docking
$ ultidock example run quickstart --dry-run

# Open the interactive lesson
$ ultidock example run quickstart
Documented CLI examples · Commands are shown, not executed

New installation? Follow the installation guide for Python, native tools, and GPU requirements.

03 / Evaluation & reproducibility

Results with context.

Explore benchmark methods
50.73jobs/s
Multi-GPU throughput
SERT / 5I6X
5 × NVIDIA RTX A40
65.13%
Top-(n+2) site recall
COACH420
DCA ≤ 4.0 Å
140,056
Docked complexes
DUD-E evaluation
Automated box placement
420/ 420
Targets processed
COACH420 evaluation
Zero target-processing failures

Reported evaluation results. Throughput depends on hardware, settings, and workload; target-processing coverage is separate from site-recovery accuracy. For protocol details, metric definitions, and per-target results, consult the benchmark materials in the data archive.

Open science

Inspect the work behind the numbers.

Explore benchmark tables and supporting materials for the site-discovery and docking evaluations.

Open data archive
Read / learn / run

Documentation

Installation, first-run guides, scientific background, and command references for v1.1.2.

Read the documentation
Inspect / adapt / contribute

Built in the open

Browse the implementation, report an issue, or contribute improvements through GitHub.

Explore the repository
Evaluate / compare / reproduce

Benchmark tooling

Explore site-recovery protocols, COACH420 and HOLO4K workflows, and active/decoy evaluation tools.

Inspect benchmark code
04 / Cite the software

Make your workflow
traceable.

If Ultidock supports your research, cite the software version you used. The repository maintains the project’s citation metadata.

View CITATION.cff
BibTeX / software reference
@misc{turgut2026ultidock,
  author       = {Turgut, Taha},
  title        = {Ultidock},
  year         = {2026},
  howpublished = {GitHub},
  note         = {Software, version 1.1.2},
  url          = {https://github.com/taka78/ultidock}
}
Use the version-specific Zenodo DOI when available.
Also cite the docking engines and site-prediction methods used in your study.