Integrated metabolomics and machine learning reveal non-volatile metabolite changes in goose eggs during refrigeration

Abstract

Goose eggs are an important avian food source with higher lipid and protein content than chicken or duck eggs, but their relatively high unsaturated fat level makes them more prone to oxidation and quality deterioration during storage. Although metabolomics has been widely applied in meat science, information on changes in non-volatile metabolites during the refrigerated storage of goose eggs remains limited. Using a non-targeted metabolomics approach, this study systematically characterized the dynamic metabolic profiles of goose eggs during refrigeration. Machine learning algorithms were further integrated to screen potential marker metabolites and to identify key metabolic pathways associated with goose egg spoilage. Temporal trend analysis was performed using Mfuzz clustering to characterize metabolite dynamics across different storage stages. After variance analysis, a total of 52 consistently upregulated and 47 consistently downregulated differential metabolites were identified. Pathway enrichment analysis indicated that purine metabolism, pyrimidine metabolism, and tyrosine metabolism may play central regulatory roles during the spoilage process of goose eggs. Furthermore, four potential spoilage biomarkers—3-hydroxybutanoic acid, 6-aminonicotinamide, Gly-His, and maleic acid—were identified through random forest and LASSO regression analysis. These findings elucidate the key metabolic pathways associated with protein/lipid oxidation during goose egg refrigeration and establish a metabolite signature-based prediction framework for goose egg freshness to achieve accurate identification of early-stage spoilage.