Integrating lipidomics and machine learning to characterize lipid profile differences among goose breeds

This study combined lipidomics technology with machine learning models to identify differences in intermuscular fat (IMF) between goose breeds and to screen for key lipid molecules. Here, we compared the meat quality characteristics and IMF lipid profiles of two goose breeds. The results showed that the meat quality traits of Yangzhou goose (YG) and Zhedong white goose (ZG) exhibited significant breed-specific differences, particularly in muscle color and physical properties. Using UPLC-ESI-MS/MS, we established a comprehensive lipid profile of YG and ZG breast muscles, identifying 35 lipid subclasses encompassing 564 lipid molecules. Based on the lipid dataset, we performed OPLS-DA analysis and identified 328 differential lipids. Further bioinformatics analysis showed that lipid deposition in YG and ZG was primarily influenced by glycerophospholipid and glyceride metabolic pathways, respectively. Subsequently, we used supervised PCA models and unsupervised machine learning models to screen characteristic lipids, and took the intersection of multiple models. Finally, we identified 8 key characteristic lipids to evaluate the differences in IMF between the two goose meats. These results show that the combination of lipidomics and machine learning not only reveals the dynamic changes of lipid molecules in complex biological samples, but also offers valuable insights for goose meat quality evaluation and biomarker screening.