Accurate de novo design of high-affinity protein binding macrocycles using deep learning
Problem
No robust method existed for de novo design of macrocyclic peptide binders against therapeutic protein targets.
Why it matters
Large-scale screening for macrocyclic binders is resource-intensive and gives little control over binding mode.
What the AI did
RFpeptides, a denoising diffusion-based pipeline, designed macrocyclic binders that were experimentally tested against four diverse protein targets.
Results
Medium-to-high-affinity binders against all four targets; KD 1-10 μM for MCL1/MDM2; anti-GABARAP KD 6 nM with sub-nanomolar IC50; RbtA binder KD < 10 nM.
Impact
Experimentally validated binders against four targets suggest a working de novo design pipeline.
Record
| Venue | bioRxiv (Cold Spring Harbor Laboratory) |
|---|---|
| Published | 2024 |
| Work type | preprint |
| Authors | Stephen Rettie, David Juergens, Victor Adebomi, Yensi Flores Bueso, Qinqin Zhao, Alexandria N. Leveille, Andi Liu, Asim K. Bera et al. |
| Citations | 21 (OpenAlex, at import time) |
| References | 35 |
| Model or method | RFpeptides |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1101/2024.11.18.622547 |
| Open access | green |
| Keywords | macrocycles, diffusion model, RFpeptides, binding affinity, de novo design |
Importance rating
significant — thousands to millions affected
First robust de novo design of protein-binding macrocycles; validated binders to four diverse therapeutic targets could ease resource-intensive drug-discovery screening.