Design of target specific peptide inhibitors using generative deep learning and molecular dynamics simulations
Problem
Designing peptide inhibitors that disrupt specific protein-protein interactions, here β-catenin with TCF/LEF transcription factors, for therapeutic discovery.
Why it matters
Targeted PPI inhibitors could enable new antibiotics and therapeutics that are hard to find by conventional methods.
What the AI did
A GRU-based variational autoencoder generated candidate peptide sequences, scored by Rosetta FlexPepDock and refined with molecular dynamics; eight inhibitors were designed, four synthesized and tested.
Results
Eight peptides designed; four synthesized and experimentally tested; two showed enhanced binding affinity for β-catenin versus the best-known peptide
Impact
Evidence supports AI-designed peptides with experimentally demonstrated enhanced binding affinity in vitro.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Sijie Chen, Tong Lin, Ruchira Basu, Jeremy L. Ritchey, Shen Wang, Yichuan Luo, Xingcan Li, Dehua Pei et al. |
| Citations | 319 (OpenAlex, at import time) |
| References | 80 |
| Model or method | GRU-based variational autoencoder (VAE) with Rosetta FlexPepDock |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-024-45766-2 |
| Open access | gold |
| Keywords | generative deep learning, peptide inhibitors, protein-protein interactions, VAE, molecular dynamics |
Importance rating
real but narrower advance
Targeted PPI peptide inhibitors could enable new antibiotics/therapeutics; only in vitro binding shown for one target, no clinical or scale evidence.