Jingxian Zheng, Yijia Tang, Shu Xing, Weiwei Han
Antioxidant peptides from marine proteins are promising bioactive compounds, yet efficient discovery and multilevel validation remain challenging. Here, we established a proteome-scale workflow integrating virtual enzymatic hydrolysis, dual-model antioxidant prediction, consensus ranking, toxicity screening, sequence-diversity-informed selection, experimental validation, and molecular simulation to identify antioxidant short peptides from Apostichopus japonicus. Five newly identified antioxidant short peptides (DHE, HDHE, DDHN, EHED, and QHDE) were synthesized for experimental validation. All showed DPPH and ABTS radical-scavenging activity, good cytocompatibility, and protection against D-galactose-induced loss of HepG2 cell viability. HDHE showed the highest DPPH activity and QHDE showed the highest ABTS activity at higher concentrations, whereas EHED produced the greatest recovery of cell viability. Molecular docking and 100 ns molecular dynamics simulations indicated sequence-dependent Keap1 interaction patterns and preserved global receptor stability, but these computational features did not parallel cytoprotection. Western blotting further showed differential regulation of Keap1, Nrf2, and HO-1, with DDHN, EHED, and QHDE significantly increasing HO-1 expression. Across the workflow, predicted activity, chemical radical scavenging, Keap1 interaction characteristics, and cellular protection did not converge on the same peptide. These findings establish a proteome-scale discovery strategy and show that multilevel validation is essential because computational, chemical, molecular-interaction, and cellular readouts represent complementary rather than interchangeable dimensions of antioxidant peptide performance.