RFdiffusion Background
RFdiffusion
Background
We will start by using the RFdiffusion binder design protocol, as described in the following papers:
Watson, J.L., Juergens, D., Bennett, N.R. et al. “De novo design of protein structure and function with RFdiffusion.”, Nature, 620, 1089–1100 (2023). https://doi.org/10.1038/s41586-023-06415-8 - Github: https://github.com/RosettaCommons/RFdiffusion
Bennett, N.R., Coventry, B., Goreshnik, I. et al. Improving de novo protein binder design with deep learning. Nat Commun, 14, 2625 (2023). https://doi.org/10.1038/s41467-023-38328-5 - Github: https://github.com/nrbennet/dl_binder_design
Dauparas, J. et al. Robust deep learning–based protein sequence design using ProteinMPNN. Science, 378,49-56(2022). https://doi.org/10.1126/science.add2187 - Github: https://github.com/dauparas/ProteinMPNN
This method combines:
- RFdiffusion for generating the backbone of a designed binder for a target structure
ProteinMPNN-FastRelax for ‘inverse folding’ to generate a sequence for the binder backbone
Alphafold2 ‘initial guess’ structure prediction to quickly score binders in silico
- To make predictions faster, Alphafold2 ‘initial guess’ doesn’t use the multiple sequence alignment (MSA) input and provides the initial target+binder complex coordinates as a starting point in the first recycle.
- ‘Full’ Alphafold2 on best binders to check conformational stability of binder monomers, verify complexes
Note: RFdiffusion (and BindCraft) depend on PyRosetta/Rosetta, which is free for non-commercial use. Commercial use requires a paid license agreement with University of Washington - see https://github.com/RosettaCommons/rosetta/blob/main/LICENSE.md and https://rosettacommons.org/software/licensing-faq/