AI-assisted in silico design and multi-target docking analysis of a novel gentamicin-ferrostatin hybrid compound for reduced nephrotoxicity and ototoxicity.

Gentamicin, a widely used aminoglycoside antibiotic, causes severe nephrotoxicity and ototoxicity through ferroptosis-mediated mechanisms involving glutathione peroxidase 4 (GPX4), the cystine/glutamate antiporter SLC7A11, NADPH oxidase 4 (NOX4), and Prestin (SLC26A5). Ferrostatin-1, a selective ferroptosis inhibitor, demonstrates preclinical cytoprotective efficacy but lacks antimicrobial activity and carries active cardiotoxicity. We designed a novel hybrid compound by covalently conjugating a garosamine-derived aminoglycoside fragment of gentamicin C1 with ferrostatin-1 via an amide bond (-CO-NH-), yielding a pharmacophore-based fragment conjugate - not the complete gentamicin C1 trisaccharide - with molecular formula C19H31N5O4 (MW 393.48 g/mol); no antibacterial activity is claimed for this fragment. Multi-target blind molecular docking was performed against five therapeutically relevant proteins - GPX4, SLC7A11, NOX4, Caspase-3, and Prestin - using CB-Dock2. The hybrid showed higher predicted binding scores than gentamicin at four of five targets and than ferrostatin-1 at all five, though binding energy differences (0.4-1.3 kcal/mol) are semi-quantitative and should be interpreted cautiously given the differing molecular weights compared. Strongest binding was observed at Prestin (-9.6 kcal/mol) with seven hydrogen bond interactions at residues TYR16, ASP518, ASP645, THR647, SER705, ILE706, and HIS707. In silico ADMET profiling using SwissADME and ProTox-3.0 revealed Toxicity Class VI for the hybrid (LD50 = 12,800 mg/kg) versus Class V for gentamicin C1 (LD50 = 5000 mg/kg), with inactive predictions for cardiotoxicity and immunotoxicity in the hybrid, compared to active predictions for gentamicin C1. Bioavailability score improved from 0.17 to 0.55. These findings provide preliminary computational evidence supporting this hybrid scaffold as a hypothesis-generating candidate warranting experimental validation. This study is entirely computational; no synthesis or biological testing has been performed.