This record provides bibliographic details and links to the original research. An editorial study breakdown is not available.
What the researchers found
AlphaFold2 was adapted to predict and design cyclic peptide structures. Over 10,000 designs were generated, with 8 tested sequences matching predictions closely and some binding targets with nanomolar affinity.
Why it matters
Designing cyclic peptides has been limited by lack of deep learning tools. This method enables rapid computational design of therapeutic peptide candidates.
The numbers in context
Over 10,000 designs. 8 tested matched X-ray structures (RMSD under 1 angstrom). IC50 under 50 nM against MDM2 and Keap1.
How the study worked
Deep learning approach adapting AlphaFold2 for cyclic peptides. Structure prediction, sequence redesign, and de novo hallucination. X-ray crystallography and binding assay validation.
Who was studied
Computationally designed cyclic peptides
What this study cannot tell us
Only 8 of 10,000+ designs experimentally tested. Two targets only. In vitro binding does not guarantee in vivo efficacy.
Read the original research
Cyclic peptide structure prediction and design using AlphaFold2.
Nature communications, 16(1), 4730
Citation
Rettie, Stephen A; Campbell, Katelyn V; Bera, Asim K; Kang, Alex; Kozlov, Simon; Bueso, Yensi Flores; De La Cruz, Joshmyn; Ahlrichs, Maggie; Cheng, Suna; Gerben, Stacey R; Lamb, Mila; Murray, Analisa; Adebomi, Victor; Zhou, Guangfeng; DiMaio, Frank; Ovchinnikov, Sergey; Bhardwaj, Gaurav. (2025). Cyclic peptide structure prediction and design using AlphaFold2.. Nature communications, 16(1), 4730. https://doi.org/10.1038/s41467-025-59940-7