Volatile markers as indicators of salmon freshness states identified via volatolomics and machine learning

Freshness is a key determinant of salmon quality and consumer acceptance. However, volatile markers and metabolic pathways driving salmon spoilage, particularly during the critical transition to the sub-fresh state, remain poorly characterized. This study systematically identified volatile markers of salmon freshness during refrigerated storage using headspace solid-phase microextraction coupled with gas chromatography-mass spectrometry (HS-SPME-GC-MS) and a random forest machine learning approach. Salmon samples were categorized into three distinct freshness stages (fresh: 0-6 days, sub-fresh: 7-8 days, and spoiled: ≥ 9 days) to investigate candidate markers under 4 °C storage. Initially, univariate ANOVA and orthogonal partial least squares discriminant analysis (OPLS-DA) identified 189 significantly different volatile compounds. Subsequent random forest modeling narrowed these to 40 potential markers with high predictive accuracy (AUC = 0.988). Further integrative analysis using O2PLS and multivariate linear regression pinpointed methyl glyoxal, benzyl alcohol, and acetic acid as robust candidate markers indicative of progressive freshness degradation. These three volatiles were validated through targeted analysis. Metabolic pathway analysis highlighted styrene degradation, phenylalanine metabolism, sulfur metabolism, glycolysis/ gluconeogenesis, pyruvate metabolism, and lysine degradation as critical spoilage pathways. Overall, this volatolomics-based strategy provides valuable insight into volatile spoilage indicators across freshness transitions and offers an effective alternative for objective salmon quality evaluation.