SeonghwanSeo/kfold

K-Fold

K-Fold is currently in the preview stage. Preprint will be released soon.

K-Fold predicts biomolecular complex structures and binding-induced conformational changes. It supports proteins, DNA, RNA, small molecules, and chemical modifications without multiple sequence alignments (MSAs).

Protein component structures are generated with AtlasFold by default or supplied from other predictors or experiments.

This repository provides K-Fold model weights. Source code, examples, and documentation are available on GitHub.

Installation and inference

K-Fold requires Python 3.11 or later.

pip install 'kfold[cuequiv]'
kfold predict --input query.yaml --out-dir predictions/ --seed 42

The cuequiv extra installs cuEquivariance kernels for faster inference on NVIDIA GPUs. Model weights and the chemical component dictionary (CCD) are downloaded automatically on first use.

See the inference guide for input formats and options, or the Python API guide for use in Python.

Preview release

Stabilization, integration of optimized kernels, and inference pipeline optimization are ongoing. Prediction results may change between versions before the stable release. The stable 1.0.0 release is planned to coincide with the preprint release.

Planned

• Stability improvements
• Faster inference
• Inference pipeline optimization

License

Copyright © 2026 Korea Advanced Institute of Science and Technology (KAIST).

K-Fold is licensed under the Apache License 2.0.