MATERIALS AND METHODS: Validated molecular docking protocols were employed to evaluate 52 flavonoids from the Indonesian Natural Products Database against the p53-MDM2 complex, yielding a validation RMSD value of 0.4935 Å. AI-assisted molecular design was subsequently applied to the top-performing compounds to generate optimized derivatives with improved binding affinity and pharmacokinetic properties. Binding interactions were analyzed using conventional molecular visualization approaches. Drug-likeness, ADME, and toxicity profiles were predicted using SwissADME, ProTox-II, and ADMETlab 2.0. Density Functional Theory (DFT) calculations at the B3LYP/6-31G(d,p) level were performed to evaluate electronic properties and molecular stability. Furthermore, Molecular Dynamics (MD) simulations and MM-GBSA calculations were conducted to investigate the dynamic stability and binding free energy of the selected complexes.

RESULTS: Among the AI-optimized derivatives, Amentoflavone-1 demonstrated the strongest binding affinity (-5.60 kcal/mol), outperforming Resveratrol (-4.24 kcal/mol). Interaction profiling revealed that hydrophobic interactions and π-π stacking with key residues within the p53-MDM2 binding pocket were the primary stabilizing forces. ADME predictions indicated improved drug-likeness and absorption properties, while toxicity analyses suggested reduced mutagenicity and mitochondrial toxicity, though a moderate hepatotoxicity risk remained. DFT calculations demonstrated favourable HOMO- LUMO energy gaps and frontier molecular orbital characteristics associated with molecular stability and chemical reactivity. Molecular dynamics simulations over 100 ns showed that the p53-MDM2- Amentoflavone-1 complex maintained a stable conformation comparable to the native ligand, as evidenced by RMSD, RMSF, SASA, and radius-of-gyration analyses. In addition, MM-GBSA calculations revealed a binding free energy of -18.37 kcal/mol for Amentoflavone-1, confirming spontaneous and thermodynamically favourable binding to the p53-MDM2 complex.

DISCUSSION: The integration of AI-driven de novo design with conventional computational modelling successfully enhanced and clarified the pharmacological potential of flavonoid scaffolds. Amentoflavone- 1 exhibited favourable binding interactions consistent with known p53-MDM2 inhibition mechanisms, supported by stable molecular dynamics behaviour and thermodynamically favourable binding profiles. Improvements in ADME and toxicity parameters further suggest that AI-guided structural modification can effectively optimize flavonoid-based inhibitor candidates.