Artificial Neural Network-Guided Discovery of Antioxidant Peptides from Peony (Paeonia ostii) Seed Meal: Peptidomics, Molecular Mechanism, and Cellular Validation

Peony seed meal (PSM), a protein-rich by-product of oil extraction from Paeonia ostii , represents an underutilized resource with significant potential for functional ingredient development. In this study, an integrated strategy combining artificial neural network (ANN)-guided hydrolysis, peptidomics, molecular simulation, and cellular validation was employed to identify antioxidant peptides from PSM. Neutrase was selected as the optimal protease, and hydrolysis conditions were optimized using a backpropagation ANN model (R = 0.9935), yielding a hydrolysate with strong radical-scavenging activity (DPPH • IC 50 = 0.30 mg/mL; ABTS •+ IC 50 = 0.07 mg/mL). LC–MS/MS identified 364 peptides, predominantly low-molecular-weight sequences. In silico screening highlighted four candidates (FRF, WQFR, FEFR, and RWL) with favorable binding toward ABTS •+ , DPPH • , and Keap1. Molecular docking and 100 ns molecular dynamics simulations confirmed stable peptide–Keap1 interactions, particularly for FRF. Cellular assays demonstrated that FRF and RWL significantly protected HepG2 cells against H 2 O 2 -induced oxidative damage by restoring antioxidant enzyme activities (SOD, CAT, and GSH-Px). Collectively, this study establishes a systematic workflow for discovering plant-derived antioxidant peptides and supports the sustainable valorization of PSM as a functional food ingredient.