High-throughput sequencing reveals microbial transitions in refrigerated sturgeon meat: Implications for quality...

This study investigated the microbially driven spoilage mechanism of sturgeon (Acipenser baerii) under refrigerated (4 °C) aerobic storage. High-throughput sequencing analysis revealed that Pseudomonas and Shewanella dominated late-stage spoilage, which was strongly positively correlated with volatile base nitrogen (VBN) accumulation and microbial metabolic shifts toward amino acid degradation pathways. By applying machine learning (random forest coupled with SHAP analysis, AUROC = 0.96) and graph neural networks (GAT, recall = 89.7 %), we pinpointed key spoilage-associated taxa and their interaction dynamics. Furthermore, numerous chemical descriptors have revealed that spoilage-associated enzymes present elevated molecular electrostatic potential (MEPs >25 kcal/mol), which facilitates the nucleophilic attack of amino acids and accelerates spoilage reactions. Time series forecasting (multivariate Prophet model) accurately predicted critical spoilage thresholds (96.5 ± 4.2 h postprocessing) with high accuracy (MAPE = 12.3 %). Additionally, metabolic modeling has demonstrated microbial cold-adapted energy strategies, including a significant increase in succinate fermentation flux (3.8 ± 0.5 mmol·g^-1 DW·h^-1) and the suppression of TCA cycle activity. This study establishes a multiscale framework linking microbial ecology, enzymatic quantum mechanics, and metabolic dynamics, offering mechanistic insight into spoilage and providing a foundation for precise sturgeon preservation strategies in cold chain logistics.