Machine learning-guided repurposing of FDA-approved quinolones as dual cholinesterase inhibitors: A multi-level docking, molecular dynamics, DFT, and SHAP-based analysis


Türkeş C.

Journal of Molecular Graphics and Modelling, cilt.143, 2026 (SCI-Expanded, Scopus) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 143
  • Basım Tarihi: 2026
  • Doi Numarası: 10.1016/j.jmgm.2025.109259
  • Dergi Adı: Journal of Molecular Graphics and Modelling
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, BIOSIS, Chemical Abstracts Core, Chimica, Compendex, EMBASE, INSPEC, MEDLINE
  • Anahtar Kelimeler: Acetylcholinesterase, Butyrylcholinesterase, Drug repurposing, Molecular docking, Molecular dynamics, Neurodegenerative diseases, Quinolone antibiotics, SHAP analysis
  • Erzincan Binali Yıldırım Üniversitesi Adresli: Evet

Özet

Alzheimer's disease (AD) involves progressive cholinergic degeneration, with acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) playing key enzymatic roles in its pathology. In this study, we computationally repurposed four FDA-approved quinolone antibiotics, Norfloxacin, Sparfloxacin, Gatifloxacin, and Nalidixic acid, as potential dual-site cholinesterase (ChE) inhibitors using a hybrid in vitro/in silico workflow. Enzyme inhibition assays identified Norfloxacin as the most potent AChE inhibitor (KI = 1.08 μM), while all compounds displayed non-competitive inhibition toward BChE. Molecular docking and MM-GBSA binding free energy analyses revealed key interactions within the catalytic gorge of AChE, supported by hydrogen bonding with Phe295 and Arg296, as well as π–π contacts with Tyr124. Density functional theory computations highlighted the influence of frontier orbital distribution on binding affinity, particularly for Norfloxacin and Sparfloxacin. An explicit-solvent molecular dynamics simulation of the AChE–Norfloxacin complex further confirmed the stability of the docking-derived binding mode over 100 ns. In an exploratory fashion, SHAP-based machine learning models were applied to a descriptor set derived from QikProp, SwissADME, and Jaguar outputs, suggesting that BBB-related indices and HOMO energy contribute to AChE inhibition, whereas the energy gap is more relevant for BChE; these trends, however, are constrained by the small four-compound dataset and should be regarded as hypothesis-generating. In silico ADME/Tox profiling indicated favorable oral drug-like properties, low predicted CYP450 inhibition liabilities, and physicochemical profiles compatible with CNS-oriented optimization, although passive BBB permeability was not predicted to be high. Finally, systems-level enrichment (STRING, GeneCards) provided a qualitative network context linking ACHE and BCHE to neurodegeneration. Together, these data position Norfloxacin and Sparfloxacin as computationally prioritised candidates whose ChE-related repurposing potential warrants further validation in dedicated cellular and in vivo models.