Protein–Ligand Dynamics · Helmholtz Munich

MISATO

Molecular dynamics and quantum-mechanical properties for thousands of protein–ligand complexes.

Use it with the research agent Project page Paper

What MISATO is

MISATO takes experimental protein–ligand structures and adds the two things a static structure cannot give a machine-learning model: dynamics and electronic structure. Its molecular-dynamics half holds 16,972 simulated complexes at 10 ns each; its quantum-mechanics half holds 19,443 curated and refined ligands with calculated electronic properties. The authors publish PyTorch dataloaders and baseline models alongside the data, and the repository ships a small sample so the format can be inspected before the full archive is fetched. The full MD archive is 133 GiB, which is the practical constraint on using it: the QM half is a 0.3 GiB download and is where most work starts.

Get the data

Its authors publish MISATO outside the Hugging Face Hub, so the dataset explorer has no table to read and there are no rows on this page. The files are linked below, and the research agent can fetch them onto its machine and work through them there.

Try asking

“Download the MISATO QM archive onto the machine, read its structure, and summarise which electronic properties are available per ligand.”

Open in the research agent

Licence

Commercial use permitted · CC BY 4.0 · code LGPL-2.1

Read 2026-10-24: the Zenodo record for the data declares CC BY 4.0 (attribution, commercial use permitted), and the repository holding the dataloaders and baseline models declares LGPL-2.1. The two are separate — using the data carries the attribution requirement, while linking the loader code into your own package carries the LGPL's obligations as well.

Questions

Can I browse MISATO in the browser?

No. Its authors publish it outside the Hugging Face Hub, so the dataset explorer has no table to read and there are no rows to show here (format: HDF5 (MD and QM), with PyTorch dataloaders). This page describes what it holds and links to the files; the research agent can fetch and read them on its machine.

What licence is MISATO under?

CC BY 4.0, with the accompanying code under LGPL-2.1. Read 2026-10-24: the Zenodo record for the data declares CC BY 4.0 (attribution, commercial use permitted), and the repository holding the dataloaders and baseline models declares LGPL-2.1. The two are separate — using the data carries the attribution requirement, while linking the loader code into your own package carries the LGPL's obligations as well.

How to cite it

Siebenmorgen, T. et al. MISATO: machine learning dataset of protein–ligand complexes for structure-based drug discovery. Nature Computational Science 4, 367–378 (2024).