Abstract
Fermentation determines the physicochemical properties and flavor characteristics of sour meat. However, research on yak sour meat fermentation remains limited, compared to pork's counterpart. This study combined physicochemical analysis, intelligent sensory techniques, gas chromatography - ion mobility spectrometry (GC-IMS), 16S rRNA sequencing and machine learning (ML) to explore bacterial dynamics and flavor development during a 45-day fermentation procedure. E-sensing distinguished stage-specific sensory profiles, with day-45 samples showing enhanced umami and sweetness. GC-IMS identified 42 volatiles, with 3-methylbutanal isomers, acetic acid and α-terpinolene increasing and 2-butoxyethanol, hexanal-D decreasing during fermentation. Bacterial diversity declined, with Staphylococcus and Lactobacillus dominating, and Lactobacillus was confirmed as the key genus for flavor formation. Fermentation caused a pH reduction, and a hardness elevation, accompanied by modifications to the color's profile. ML models demonstrated high predictive performance: Support Vector Machine (SVM) achieved 100% accuracy in classifying fermentation stages using E-nose/E-tongue data, while k-Nearest Neighbors (k-NN) optimally differentiated stages based on GC-IMS volatile profiles. This study elucidates the interplay between microbial succession and flavor evolution in yak sour meat, establishing a scientific foundation for optimizing fermentation protocols and enhancing product quality control.