Plasma metabolomic signatures differentiate Duroc pigs with divergent feed intake: insights into the molecular architecture of a complex trait

Feed intake is a biologically complex and economically relevant trait in pig production, influencing growth rate, efficiency, sustainability, and overall production costs. Its variability results from the interaction between environmental factors and genetically regulated biological processes operating across multiple molecular and physiological levels. Dissecting the molecular basis of feed intake is therefore essential to better understand its biology and to identify biomarkers that may improve breeding, nutritional strategies and the sustainability of pig production systems. Metabolomics, defined as the comprehensive analysis of low molecular weight molecules in biological systems, enables the exploration of intermediate molecular phenotypes linking genotypes to complex traits, potentially providing simpler and biologically informative proxies. The objective of this study was to identify plasma metabolites discriminating pigs with extreme and divergent feed intake and to evaluate their biological relevance. A total of 290 growing male Duroc pigs were selected from a larger performance-tested population and classified into high (n=150) and low (n=140) feed intake groups. Untargeted plasma metabolomics quantified 756 metabolites. After quality control, 587 metabolites spanning major biological classes were retained. Missing values were handled, generating 25 minimally different datasets. A robust feature selection pipeline based on the Boruta algorithm (Random Forest classifier) was applied across the 25 datasets, combined with 10-fold cross-validation, resulting in 1,250 feature selection runs. Metabolites consistently selected across runs were considered discriminant. Their discriminatory power was further evaluated using the ROC-AUC, principal component analysis (PCA), and standard Random Forest. Pearson correlations with raw feed intake were calculated, and biological interpretation was supported by pathway enrichment and metabolite network analyses. Twenty-nine metabolites (~5% of the filtered dataset) consistently discriminated between the two groups. Most were amino acids (66%), followed by nucleotides (17%), lipids (10%), one carbohydrate, and one unnamed compound. Amino acids were significantly over-represented (PFDR<0.05), whereas lipids were under-represented relative to the whole metabolome. Correlations with feed intake ranged from -0.40 to +0.34. AUC values ranged from 0.61 to 0.75, indicating moderate but consistent discriminatory capacity. PCA based on the 29 selected metabolites improved group separation compared with the whole-metabolome PCA. Pathway analyses highlighted enrichment of glutamate, histidine, branched-chain amino acids, and lysine metabolisms. These results support the hypothesis that specific plasma metabolites reflect biological mechanisms underlying feed intake variability. The predominance of amino acid signatures suggests an active involvement of protein and energy metabolism in differentiating intake extremes. In conclusion, the pig plasma metabolome contains reproducible molecular phenotypes associated with feed intake. These findings contribute to a better understanding of the biological architecture of feed efficiency and may support the development of metabolomics-informed breeding strategies. This study received funding from the European Union’s Horizon Europe research and innovation program under grant agreement No. 101059609 (Re-Livestock project).