Structural and dynamic characterization of papain-like protease ligand recognition using integrated computational and biochemical approaches

The papain-like protease (PLpro) is a structurally conserved cysteine protease whose catalytic activity and regulatory functions are governed by a well-defined macromolecular architecture and a dynamically adaptable ligand-binding site. Understanding the structural determinants that govern ligand recognition and inhibition of PLpro remains essential for elucidating its structure-function relationships. Here, we describe an integrated computational and experimental framework to characterize small-molecule engagement of PLpro, combining supervised machine learning-guided ligand prioritization with structure-based modeling, molecular dynamics simulations, and biochemical functional evaluation. A LightGBM classification model trained on curated PLpro activity data exhibited stable predictive performance on an independent test set (AUC 0.904, average precision 0.892) and was applied to a library of 4779 natural products, yielding 243 high-confidence candidates for structure-guided screening. Molecular docking identified five compounds (HY-101399, HY-101406, HY-101410, HY-N7141, and HY-100196) with binding scores ranging from -7.6 to -7.9 kcal/mol, which were comparable to the reference ligand GRL0617 (-7.4 kcal/mol) and reproduced key interactions with Asp164, Arg166, Gln269, and adjacent residues within the PLpro binding pocket. MM-GBSA analysis indicated favorable binding energetics, including -35.04 ± 0.21 kcal/mol for HY-101399 and -37.47 ± 0.25 kcal/mol for HY-N7141. Biochemical assays confirmed that all five compounds inhibited PLpro protease activity with IC₅₀ values of 2.24-3.04 μM and Deubiquitinase activity with IC₅₀ values of 2.60-3.40 μM, while exhibiting low cytotoxicity in BJ fibroblasts. Collectively, these results define structurally diverse natural products as effective ligands for probing PLpro structure-function relationships and demonstrate how integrated computational strategies can support macromolecule-focused inhibitor characterization.