Integrated Computational Strategy for Predicting the Binding Site of PD-L1 IgV-like Domain by a Short Linear Peptide
발표자
손한빈 (KAIST)
연구책임자
남윤성 (KAIST)
공동저자
손한빈 (KAIST), 남윤성 (KAIST)
초록
내용
Computational approaches have gained increasing attention in the development of new therapeutic agents to minimize experimental trial-and-error. To overcome the limitations of static prediction for peptide-protein interactions, we propose a robust ‘filtering funnel’ initiating with docking simulations using CABS-dock (global search) and FlexPepDock (atomic refinement). Candidates were screened via PRODIGY and PLIP to optimize the pool for molecular dynamics. Validated using a known PD-L1 binder, trajectory analysis revealed that authentic interactions are characterized by a rigid central core (lowest root mean square fluctuation), high hydrogen bond occupancy and sustained buried surface area, contrasting with flexible termini. By distinguishing robust binders from unstable decoys, this study suggests a scalable framework applicable to newly discovered peptides, offering a rational strategy to prioritize high-confidence candidates prior to wet-lab synthesis. Funding: NRF grant funded by the Korea government (MSIT) (RS-2025-02217286).