Repurposing DrugBank compounds as NAD-dependent deacetylase sirtuin 2 inhibitors via QSAR modelling with gradient boosting algorithms and all-atom molecular simulations

Sirtuin 2 (SIRT2), a NAD+-dependent histone deacetylase implicated in α-synuclein aggregation, is an emerging target for disease-modifying therapies in Parkinson’s disease (PD). Here, we employed an integrated computational drug-repurposing strategy to identify potent SIRT2 inhibitors from the DrugBank database. A curated set of 949 inhibitors was used to construct quantitative structure–activity relationship (QSAR) models with four gradient-boosting algorithms, yielding CatBoost as the optimal predictor (= 0.74, = 0.72). The model screened 4947 drug-like compounds, from which 97 candidates with predicted pIC50 ≥ 6 were prioritized. Molecular docking against the SIRT2 crystal structure (PDB: 4RMG) revealed high-affinity binding modes for multiple hits, notably DB14822, DB03571, and DB06506, engaging conserved residues (Phe119, Tyr139, Phe190, Ile232) through hydrophobic and π-stacking interactions. ADMET profiling indicated favorable drug-likeness and acceptable pharmacokinetic/toxicity properties for most candidates. All-atom molecular dynamics simulations (250 ns) demonstrated that top ligands maintained compact, stable complexes with low RMSD, restricted radius of gyration, and minimal solvent exposure. Principal component and free energy landscape analyses confirmed constrained global motions, while MM/GBSA calculations yielded favorable binding free energies (− 32.6 to − 35.7 kcal/mol) for lead compounds. Given SIRT2’s established role in α-synuclein aggregation and neurodegeneration, these compounds represent potential therapeutic starting points for Parkinson’s disease and merit experimental validation.

Journal Article