ROCKET User Guide

Documentation, installation instructions, and tutorials for ROCKET

ROCKET refines protein structure predictions against X-ray and cryo-EM/ET data.

It combines OpenFold prediction, Phenix preprocessing, and gradient-based maximum-likelihood refinement.

New to ROCKET? Install first. Then follow the setup tutorial for your data type.

Read our paper in Nature Methods.

GitHub - rs-station/ROCKET: AI-powered macromolecular refinementGitHub

Start here

Keeping it simple: install once, preprocess datasets, then iterate in rk.refine.

Title

Install ROCKET

Target

Installing ROCKET

Title

Set up with X-ray data

Target

Launch with Your Own X-ray Data

Title

Set up with cryo-EM data

Target

Launch with Your Own Cryo-EM Data

Title

CLI reference

Target

API

Example workflows

Title

MSA subsampling with data-based scoring

Target

Latent State of a Human Serpin with MSA Subsampling

Title

Low-resolution cryo-EM refinement

Target

Low Resolution GroEL Sub-Tomogram Average

Citing

@article{fadini2026alphafold,
  title={AlphaFold as a prior: experimental structure determination conditioned on a pretrained neural network},
  author={Fadini, Alisia and Li, Minhuan and McCoy, Airlie J. and Banjara, Suresh and Okumura, Hiroki and Napier, Eve and Fontana, Pietro and Khan, Amir R. and Jovine, Luca and Terwilliger, Thomas C. and Read, Randy J. and Hekstra, Doeke R. and AlQuraishi, Mohammed},
  journal={Nature Methods},
  volume={23},
  number={4},
  pages={785--795},
  year={2026},
  doi={10.1038/s41592-026-03047-4}
}