FoldKit#

FoldKit is a Python toolkit for working with AlphaFold3 results — loading predictions, computing per-chain and per-pair confidence metrics, and compressing large result directories down to a fraction of their size.

Installation

Get FoldKit installed from PyPI in one command.

Installation
Tutorial

Walk through loading single structures, ensembles, and the CLI.

Tutorial
API Reference

Full documentation for AF3Result, AF3Ensemble, and the CLI.

API Reference
GitHub

Source code, issues, and releases.

https://github.com/jonlevi/foldkit

Why FoldKit?#

AlphaFold3 output directories are large and awkward to work with at scale. FoldKit gives you a clean Python interface to the confidence metrics you actually care about — pLDDT, pAE/ipAE, pTM/ipTM, contact probabilities, and ipSAE — for both single predictions and multi-seed, multi-sample ensembles.

It also ships a compression format that shrinks result directories dramatically without losing any of the underlying data:

Scale

Raw AF3 output

FoldKit exported

One 4-chain complex

~7.8 MB

~1.9 MB

1,000 complexes × 4 seeds × 5 samples

~157 GB

~38 GB

Quick example#

import foldkit

result = foldkit.AF3Result.load_af3_result("tutorial_example/single_result")
result.get_ptm()                          # mean global PTM
result.get_ipae(chain1="B", chain2="P")    # interface PAE between two chains

See the Tutorial for the full walkthrough, including ensembles and the command-line interface.