Identification of Novel Extracellular-Signal-Regulated Kinase 2 Inhibitors Through Machine Learning-Driven De Novo...

Background: The extracellular-signal-regulated kinase (ERK) cascade regulates cell proliferation, differentiation, and survival, and ERK2 mediates substrate phosphorylation, influencing gene expression and cellular functions. Methods: In the current study, a pool of new molecules was generated using the DeLA-Drug, a machine learning (ML)-assisted de novo design tool. The chemical space was reduced through a similarity search against active ERK2 inhibitors and molecular docking with AutoDock vina, followed by pharmacokinetic assessment in DeepPK. Poses of the final selected molecules were refined in DiffDock, and dynamicity was assessed through molecular dynamics (MD) simulation. Finally, the free-energy perturbation (FEP)-based binding affinity was explored in Gromacs2023.4. Results: From the above approaches, four molecules (Ek1, Ek2, Ek3, and Ek4) were identified as promising candidates with favorable binding interactions. Molecular docking revealed that the selected molecules exhibited higher binding affinity for ERK2, ranging from -9.50 to -10.50 kcal/mol. The dynamics assessment via MD simulation clearly revealed their strong association with ERK2, corroborated by the lower deviation of the ERK2 backbone in dynamic states. All four screened molecules have satisfactory pharmacokinetic properties, medicinal chemistry properties, and good synthetic accessibility scores, indicating their potential as drug-like compounds under Lipinski's rule of five to inhibit or modulate ERK2 activity. The FEP energy of Ek1 was found to be -26.85 kJ/mol, which is higher than the standard molecule (-22.77 kJ/mol) and indicates its strong affinity toward ERK2. Conclusions: These results suggest that all proposed ERK2 modulators are potential avenues for future drug discovery targeting ERK2, subject to experimental validation.