SeonghwanSeo/kfold

K-Fold

K-Fold v1.0.0b1 is a beta release under active development.

K-Fold predicts biomolecular complex structures and binding-induced conformational changes. Through an apo-to-holo diffusion bridge, K-Fold aims to capture conformational changes in systems such as G protein-coupled receptors (GPCRs).

K-Fold supports proteins, DNA, RNA, small molecules, and chemical modifications without multiple sequence alignments (MSAs).

By default, K-Fold prepares protein component structures with AtlasFold. You can also provide apo structures from experiments or other prediction tools.

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
kfold --input query.yaml --out-dir predictions/ --seeds 42

Use --kernel {triton,cuequiv,torch} to select the backend for K-Fold complex prediction. Automatic selection prefers Triton, then cuEquivariance, then PyTorch, depending on availability. Install the cuequiv extra (pip install 'kfold[cuequiv]') to enable cuEquivariance kernels.

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.

Beta release

K-Fold v1.0.0b1 is a beta release. Interfaces and prediction results may change before the stable release.

License

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

K-Fold is licensed under the Apache License 2.0.