Related Experiment Video
Updated: Jun 13, 2025

06:50
Author Spotlight: A Computational Approach to Decipher Amino Acid Preferences in Multispecific Protein-Protein Interactions
Published on: January 26, 2024
1.8K
Target-Specific De Novo Peptide Binder Design with DiffPepBuilder
Fanhao Wang1, Yuzhe Wang1, Laiyi Feng2
1Center for Quantitative Biology, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing 100871, China.
Journal of Chemical Information and Modeling
|September 12, 2024
Summary
DiffPepBuilder designs novel peptide binders using an SE(3)-equivariant diffusion model. This method effectively generates stable peptide sequences and structures, outperforming existing tools in binding affinity and diversity.
Area of Science:
- Computational biology
- Protein engineering
- Drug discovery
Background:
- Peptide binder design is challenging due to structural flexibility and limited complex data.
- Existing *de novo* protein binder design methods struggle with peptide specificity.
Purpose of the Study:
- To develop a *de novo* method for generating target-specific peptide binders.
- To improve the stability and binding affinity of designed peptides.
Main Methods:
- Curated a large synthetic dataset (PepPC-F) from protein-protein interface data.
- Developed DiffPepBuilder, an SE(3)-equivariant diffusion model for peptide sequence and structure co-design.
- Incorporated disulfide bonds for enhanced peptide structural stability.
Main Results:
- DiffPepBuilder successfully recalled native peptide structures and sequences.
- Generated novel peptide binders with improved binding free energy.
- Outperformed AfDesign and RFdiffusion in sequence/structure recall, interface quality, and diversity.
- Disulfide bonds enhanced structural rigidity and binding performance.
Conclusions:
- DiffPepBuilder is an effective *de novo* peptide binder design tool.
- The method shows promise for designing peptide binders for specific protein targets.
- DiffPepBuilder advances the field of peptide-based therapeutics design.

