Integrated machine learning and structure-based discovery of noncovalent natural product inhibitors targeting SARS-CoV-2 papain-like protease

Papain-like protease is an essential severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) enzyme involved in viral polyprotein processing and modulation of host innate immune responses, supporting its consideration as a viable target in antiviral research. Here, we present an integrated computational and experimental strategy to identify noncovalent inhibitors of SARS-CoV-2 papain-like protease from a natural product chemical space. A curated dataset of 893 reported papain-like protease inhibitors was used to train machine learning classification models, among which a light gradient boosting model achieved a testing area under the receiver operating characteristic curve of 0.904. This model was applied to screen a natural product library of 4,945 compounds, yielding 355 candidates with predicted inhibitory probabilities greater than 0.995. Structure-based docking against the GRL0617-bound protease structure prioritized five compounds with docking scores ranging from -8.86 to -8.14 kcal/mol, compared with -7.39 kcal/mol for the reference inhibitor GRL0617. Molecular dynamics simulations of 500 ns revealed stable protein–ligand complexes, with root-mean-square deviation values stabilizing between 1.6-3.5 Å and preservation of global compactness across all systems. Collective motion analysis showed first principal component contributions ranging from 27.43% to 50.28%, indicating ligand-dependent modulation of dominant conformational dynamics. Molecular mechanics–generalized born surface area (MM-GBSA) calculations yielded total binding free energies ranging from -28.71 ± 0.27 to -40.20 ± 0.47 kcal/mol. Biochemical evaluation using a fluorometric assay demonstrated that the five selected compounds inhibit papain-like protease with IC ₅₀ values between 2.62 and 3.89 μM, comparable to the IC ₅₀ of 2.83 μM for GRL0617 measured under identical conditions. Together, these results highlight prioritizing noncovalent papain-like protease inhibitors and support further development of natural products targeting this enzyme.