Skip to content

Code implementation of "Diffusion probabilistic modeling of protein backbones in 3D for the motif-scaffolding problem" https://arxiv.org/abs/2206.04119

Notifications You must be signed in to change notification settings

blt2114/ProtDiff_SMCDiff

Repository files navigation

Diffusion Probabilistic Modeling of Protein Backbones in 3D for the Motif-Scaffolding Problem

Implementation for "Diffusion Probabilistic Modeling of Protein Backbones in 3D for the Motif-Scaffolding Problem" paper link. (Accepted at International Conference on Learning Representations 2023)

LICENSE: MIT

If you use our work then please cite

@article{trippe2022diffusion,
  title={Diffusion probabilistic modeling of protein backbones in 3D for the motif-scaffolding problem},
  author={Trippe, Brian L and Yim, Jason and Tischer, Doug and Broderick, Tamara and Baker, David and Barzilay, Regina and Jaakkola, Tommi},
  journal={arXiv preprint arXiv:2206.04119},
  year={2022}
}

Installation

To install, you may use miniconda (or anaconda). Run the following to install a conda environment with the necessary dependencies.

conda env create -f protein_diffusion.yml

Next, we recommend installing our code as a package. To do this, run the following.

pip install -e .

Running

You can run examples of unconditional generation and motif-scaffolding with SMC-Diff with the included ipython notebook.

inference_example.ipynb

Related works

This implementation has largely been superceded by FrameDiff and RFdiffusion.

Third party source code

Our model code adapts equivariant graph convolutionanal neural networks (code link). If you adapt this code further, please credit them as well.

About

Code implementation of "Diffusion probabilistic modeling of protein backbones in 3D for the motif-scaffolding problem" https://arxiv.org/abs/2206.04119

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published