A Multistage Virtual Screening Strategy Integrating Molecular Similarity, Deep Learning Scoring, and Molecular Docking toward the Discovery of Novel LRRK2 Inhibitors.
AI interpretation is pending for this paper.
What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
Leucine-rich repeat kinase 2 (LRRK2) has emerged as an attractive molecular target for Parkinson's disease therapeutics. To discover novel and potent LRRK2 inhibitors, we employed an integrated virtual screening workflow comprising a similarity search, deep learning-guided compound filtering, molecular docking, and experimental validation. Starting from three previously identified hit compounds, a structurally diverse compound library was generated and subsequently prioritized using the Ouroboros molecular representation model, which incorporates conformational and pharmacophore features. Following docking-based virtual screening, 15 candidate compounds were selected for enzymatic evaluation, leading to the identification of four novel LRRK2 inhibitors. Among them, compound C-298 exhibited the highest potency with IC50 values of 315.7 nM against wild-type LRRK2 and 255.4 nM against the G2019S mutant. The cell viability assay indicated comparable cytotoxicity of compound C-298 to the positive control MLi-2. Furthermore, compound C-298 effectively reduced reactive oxygen species levels and inhibited the phosphorylation of LRRK2 (Ser935) and Rab10 (Thr73) in a concentration-dependent manner. Molecular dynamics simulations elucidated that hydrogen bond interactions with residues Glu1948 and Ala1950, as well as molecular rigidity, play critical roles in inhibitory activity. Our study demonstrates the utility of AI-assisted virtual screening in accelerating LRRK2 inhibitor discovery and identifies compound C-298 as a promising inhibitor, providing valuable insights for further rational design of LRRK2 inhibitors.