이전 사이트 이동

This project aims to develop a diffusion-based cyclic peptide design model conditioned on the contact information of a target protein binding interface. Using protein–peptide complex structural data, the model learns residue-level contact patterns between target proteins and peptides to generate cyclic peptide structures with high binding potential. Geometric constraints such as head-to-tail cyclization are incorporated to ensure that generated peptides satisfy realistic cyclic peptide structural properties. The generated peptide candidates are evaluated through docking simulations and binding affinity estimation against the target protein. Through this approach, the study seeks to establish an AI-driven framework for cyclic peptide binder design and contribute to next-generation drug discovery research.