Computational identification of novel inhibitors for norovirus 3CL protease using DeepPurpose, Lipinski rule-based screening, and molecular dynamics simulations

The rapid transmission of norovirus among individuals of all ages makes it a major issue in public health. Targeting the 3CL protease, an essential enzyme in viral protein processing, is one potential strategy for the development of antiviral drugs. Using computational methodologies, this study investigates inhibitors that could disrupt the norovirus 3CL protease's activity. The Lipinski rule was used to screen a total of 9699 compounds, resulting in 7171 compounds. Machine learning-based screening using the DeepPurpose model further refined the dataset to 408 compounds, among which 162 demonstrated enhanced docking scores ranging from −6.29 to −7.8 kcal/mol compared with the control (−6.28 kcal/mol). Clustering analysis identified 77 compounds exhibiting structural similarity to the control ligand. Subsequent PSICHIC-based binding affinity prediction identified two top compounds, 124122368 and 73,224,028, with predicted affinities of 7.67 and 7.22, respectively. Molecular dynamics simulations conducted for 300 ns revealed that compound 124,122,368 exhibited lower and more stable fluctuations (0.1–0.13 nm) than the control (0.13–0.15 nm), indicating superior binding stability. Free energy landscape (FEL) analysis demonstrated compact conformational clustering and well-defined minima, suggesting energetically favorable binding. MM/GBSA calculations yielded a binding free energy of −36.01 kcal/mol for 124,122,368, comparable to the control (−43.23 kcal/mol), confirming its stable interaction within the active site. These findings indicate that compound 124,122,368 exhibits strong and stable binding to the norovirus 3CL protease, establishing it as a promising lead molecule for further in vitro and in vivo validation toward antiviral drug development.