RareFoldGPCR

Agonist Design Beyond Natural Amino Acids

By transfer learning from RareFold on high-quality structurtes from GPCRdb, we can learn how to incorporate noncanonical amino acids (NCAAs) seen in RareFold to the structure prediction of GPCRs without ever having seen NCAA-based GPCR modulators. Read more here

RareFoldGPCR (RFG) supports 49 different amino acid types.
The 20 regular ones, and 29 RARE ones: MSE, TPO, MLY, CME, PTR, SEP,SAH, CSO, PCA, KCX, CAS, CSD, MLZ, OCS, ALY, CSS, CSX, HIC, HYP, YCM, YOF, M3L, PFF, CGU,FTR, LLP, CAF, CMH, MHO

Click the image for a video explanation!

If you like RareFoldGPCR - please STAR the repo!

Table of Contents

1. Agonist Design Beyond Natural Amino Acids
2. LICENSE
3. Installation
4. Design using RareFoldGPCR
5. Citation
6. Data
7. The EvoBind ecosystem

LICENSE

RareFoldGPCR is available under the Apache License, Version 2.0.
The RareFoldGPCR parameters for prediction and design are made available under the terms of the CC BY-NC 4.0 license. The design protocol is made available under the terms of the CC BY-NC 4.0 license.

You may not use these files except in compliance with the licenses.

Installation

The entire installation takes <1 hour on a standard computer.
We assume you have CUDA12. For CUDA11, you will have to change the installation of some packages.
The runtime will depend on the GPU you have available and the size of the protein you are predicting.
On an NVIDIA A100 GPU, the prediction time is a few minutes on average.

First install micromamba (you can also build from conda or mamba if you already have these installed) https://mamba.readthedocs.io/en/latest/installation/micromamba-installation.html

bash install_dependencies.sh

Run the script to:

1. Install the RareFoldGPCR environment
2. Get the RareFoldGPCR parameters for binder design
3. Get Uniclust for MSA search
4. Install HHblits for MSA search

Design using RareFoldGPCR

Run the test case (a few minutes)

micromamba activate rarefold
bash design.sh

If you want to use your target, simply replace the structure file for processing in "design.sh" with a path for a new one. This will take care of all feature generation. Note that RareFoldGPCR is purely sequence based, but can use structural input for e.g. scaffolding or for biasing the design towards certain states.
We have included the structure as a starting point here to make it easy for you to start designing.

Citation

If you use RareFoldGPCR in your research, please cite:

RareFoldGPCR: Agonist Design Beyond Natural Amino Acids Qiuzhen Li, Thomas Helleday, Patrick Bryant bioRxiv 2025.10.01.679733; doi:10.1101/2025.10.01.679733 link to paper

and

Li Q, Daumiller D, Zuo F, Marcotte H, Pan-Hammarstrom Q and Bryant P. RareFold: Structure prediction and design of proteins with noncanonical amino acids. bioRxiv. 2025. p. 2025.05.19.654846. doi:10.1101/2025.05.19.654846 link to paper

Data

https://zenodo.org/uploads/15180406

The EvoBind ecosystem

EvoBind - designs novel [cyclic] peptide binders based only on a protein target sequence.
RareFold - prediction & design with noncanonical amino acids
RareFoldGPCR - GPCR agonist design with noncanonical amino acids