Microbial Succession Patterns and Differences During Storage of Baobaoqu of Different Qualities

The quality of Baobaoqu directly determines the flavor quality and yield of baijiu. In this work, metagenomic approaches combined with machine learning, including a multilayer perceptron (MLP) neural network model and the SHapley Additive exPlanations (SHAP) method, were employed to investigate the succession patterns and differential characteristics of microbial communities and potential enzymatic systems between two grades of Baobaoqu during storage. The results showed that Premium Baobaoqu possessed a more diverse microbial community than Normal Baobaoqu: a total of 1764 genera and 6901 species were annotated in Premium Baobaoqu, versus 1656 genera and 6440 species in Normal Baobaoqu. The top five dominant microbial genera, namely Weissella, Staphylococcus, Thermoactinomyces, Limosilactobacillus, and Pediococcus, collectively occupied 46.3–87.1% of the total microbial abundance. Ten microbial taxa were identified as key differential candidate microbial markers—including Weissella confusa, Kluyveromyces marxianus, and Staphylococcus lloydii—eight of which exhibited significantly higher abundance in Premium Baobaoqu. Glycoside hydrolases (GHs) constituted the largest enzyme class, accounting for 40.99% of the total. The top ten potential enzyme families significantly differentiated between them were GH43, GH31, GT9, GH2, CBM50, GH5, GH4, GT4, and GT26, with higher abundance of GH43, GH31, GH5, and CBM50 in Premium Baobaoqu. This study systematically uncovered the distinct microbial and enzymatic profiles of Baobaoqu of different qualities throughout storage. The findings provide theoretical guidance for the optimization of industrial production workflows and storage strategies for Baobaoqu manufacturing.