Analysis of lipid–protein interactive oxidation and underlying pathways in aquatic products via multi-omics and machine learning: a case study of air-dried hairtail

The interactive oxidation mechanisms between lipids and proteins during aquatic product processing remain poorly understood, limiting quality control strategies. This investigation employed a comprehensive approach by integrating multi-omics data with seven machine learning models to elucidate the mechanisms of lipid and protein interactive oxidation in air-drying hairtail. The optimal XGBoost model achieved 0.98 AUC with 0.90 accuracy on test set. Machine learning combined with multi-omics has identified six oxidative biomarkers of lipid-protein interactions: receptor expression-enhancing protein 6-like, glutathione-S-transferase 1-like, peroxisome biogenesis factor 19, protein-glutamine gamma-glutamyltransferase 2-like isoform X2, dMePE (18:4/16:0), and PE (20:1/18:1). The analysis of the underlying pathways associated with these oxidative biomarkers revealed their involvement primarily in ferroptosis and the protein kinase A pathway. The lipid-protein interaction oxidation process in air-dried hairtail is facilitated by a TG2-crosslinked gel network. This study offers novel molecular targets and theoretical insights for developing targeted antioxidant interventions during processing.