All-Atom Structure Dataset · Koes Lab, University of Pittsburgh
SidechainNet
An all-atom extension of the standard structure-prediction splits, with side-chain angles and coordinates.
What SidechainNet is
The established structure-prediction datasets describe the backbone and stop there, which quietly excludes every question about side chains — packing, rotamer preference, the atoms that actually form a binding site. SidechainNet keeps the same protein entries and splits and adds the measurements that complete them: torsion angles and heavy-atom coordinates for the whole residue, hydrogens excluded. It ships as a Python package with loaders and batching for PyTorch, structure-writing helpers that emit PDB and glTF for visualisation, and an interface to OpenMM for computing energies and forces on a predicted structure. Users can extend it with new PDB entries as they are released rather than waiting for a new version of the dataset.
Get the data
Its authors publish SidechainNet 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
“Install the SidechainNet package on the machine, load one protein from a CASP12 release and report its side-chain torsion angles.”
Licence
Commercial use permitted · BSD-3-Clause
Read 2026-10-24: the repository and the published Python package both declare BSD-3-Clause. The dataset files themselves are served from the authors' university host and derive from the Protein Data Bank, which places no restriction on reuse.
Questions
Can I browse SidechainNet 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: Pickled Python dictionaries, one file per release and sequence-identity thinning). 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 SidechainNet under?
BSD-3-Clause. Read 2026-10-24: the repository and the published Python package both declare BSD-3-Clause. The dataset files themselves are served from the authors' university host and derive from the Protein Data Bank, which places no restriction on reuse.
How to cite it
King, J. E. & Koes, D. R. SidechainNet: An all-atom protein structure dataset for machine learning. Proteins 89, 1489–1496 (2021).