Related Experiment Video
Updated: Apr 12, 2026
06:50
Author Spotlight: A Computational Approach to Decipher Amino Acid Preferences in Multispecific Protein-Protein Interactions
Published on: January 26, 2024
2.7K
Accurate 3D Structure Prediction of Small Cyclic Peptides Containing Non-Canonical Amino Acid Residues Using an
1School of Pharmaceutical Sciences, Shanghai Jiao Tong University, Shanghai 200240, China.
Journal of Chemical Information and Modeling
|April 10, 2026
Summary
Cyclic peptides show drug potential but face prediction challenges. A diffusion model, AGDIFF, was adapted for accurate 3D structure generation of these peptides, including those with noncanonical amino acids.
Area of Science:
- Computational chemistry
- Drug discovery
- Structural biology
Background:
- Cyclic peptides offer drug advantages like stability and oral bioavailability.
- Predicting structures of cyclic peptides with noncanonical amino acids (NCAAs) and varied cyclization is difficult.
Purpose of the Study:
- To adapt and evaluate the AGDIFF diffusion model for high-precision 3D structure prediction of small cyclic peptides, including those with NCAAs.
- To assess AGDIFF's performance on diverse cyclic peptide structures and cyclization chemistries.
Main Methods:
- Retrained AGDIFF, an all-atom diffusion generative model, on the CREMP dataset (36,198 macrocyclic peptides).
- Utilized a 2D molecular graph representation to inherently support NCAAs and complex linkages.
- Implemented a stereochemical correction step for accurate enantiomer prediction.
Main Results:
- AGDIFF achieved high accuracy on benchmark cyclic peptides (average RMSD 0.79 Å, ring torsion deviation 6.55°).
- The model successfully resolved stereochemical insensitivity and predicted correct antipodes for enantiomeric residues.
- Ramachandran analyses validated the conformational plausibility of generated peptide ensembles.
Conclusions:
- Diffusion-based deep learning, exemplified by AGDIFF, is effective for cyclic peptide modeling.
- This approach holds significant potential for the rational design of novel NCAA-containing macrocyclic peptides in drug discovery.
