In silico directed evolution of peptide binders

EvoBind (v2) designs novel peptide binders based only on a protein target sequence. It is not necessary to specify any target residues within the protein sequence or the length of the binder (although this is possible). Cyclic binder design is also possible.

Read more here

EvoBind2 accounts for adaptation of the receptor interface structure to the peptide being designed during optimisation: sequence and structure is generated simultaneously. This consideration of flexibility is crucial for binding. EvoBind is the first protocol that only relies on a protein sequence to design a binder with experimentally verified cyclic design capacity.

Receptor in green and peptide in blue.

1. EvoBind
2. LICENSE
3. Colab
4. Computational requirements
5. Setup
6. Design binders
   • Cyclic design
   • Adversarial evaluation with AlphaFold-multimer
7. Citation
8. Examples of studies with EvoBind
9. The EvoBind ecosystem

EvoBind2 is based on AlphaFold2, which is available under the Apache License, Version 2.0.
The AlphaFold2 parameters are made available under the terms of the CC BY 4.0 license and have not been modified.
The design protocol EvoBind2 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.

It is possible to run EvoBind2 online in the Google colab here

Before beginning the process of setting up this pipeline on your local system, make sure you have adequate computational resources. Make sure you have an available GPU as this will speed up the prediction process substantially compared to using a CPU. EvoBind2 assumes you have NVIDIA GPUs on your system, readily available. A Linux-based system is assumed.

To setup this pipeline, clone this github repository:

git clone https://github.com/patrickbryant1/EvoBind.git

Then do

bash setup.sh

This script fetches the AlphaFold2 parameters, installs a conda env and downloads uniclust30_2018_08 which is used to generate the receptor MSA.

To design binders the following needs to be specified:
Receptor fasta sequence
Optional arguments:
Peptide length - default=10
Target residues within the raceptor sequence - default=all

If you want to design a cyclic peptide, add the flag --cyclic_offset=1 in the design script when calling mc_design.py. Based on cyclic offset.

A test case is provided in design_local.sh.
This script can be run by simply doing:

bash design_local.sh

See src/AFM_eval for instructions

If you use EvoBind in your research, please cite

Li, Q., Vlachos, E.N. & Bryant, P. Design of linear and cyclic peptide binders from protein sequence information. Commun Chem 8, 211 (2025). https://doi.org/10.1038/s42004-025-01601-3

1. Daumiller D*, Giammarino F*, Li Q, Sonnerborg A, Cena-Diez R, Bryant P. Single-Shot Design of a Cyclic Peptide Inhibitor of HIV-1 Membrane Fusion with EvoBind. bioRxiv. 2025. p. 2025.04.30.651413. doi:10.1101/2025.04.30.651413
2. Li Q, Wiita E, Helleday T, Bryant P. Blind De Novo Design of Dual Cyclic Peptide Agonists Targeting GCGR and GLP1R. bioRxiv. 2025. p. 2025.06.06.658268. doi: https://doi.org/10.1101/2025.06.06.658268

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