Predicting the multispecies solid-state vinegar fermentation process using single-cell Raman spectroscopy combined with machine learning

Microbial community is a key contributing factor for flavor formation in natural food fermentation. However, it is a challenge to maintain batch-to-batch uniformity during the fermentation process due to the diversity and variability of microbial community. A rapid detection of the structure and function of the microbial community in the whole fermentation process is of great importance for quality control of the final fermentation products. Firstly, we employed amplicon sequencing to target the dominant operational taxonomic units in the microbial community of Zhenjiang aromatic vinegar, a traditional cereal vinegar. Secondly, we isolated and created a Raman database for 13 dominant bacterial species from vinegar culture, enabling us to establish a logistic regression model with 96.4% accuracy in species classification. Finally, a Raman-fermentation phase regression model was established, achieving an R2 of 0.952, accurately determining the actual fermentation phase of vinegar. This study offers a new method for dynamics monitoring of microbial community, prediction of fermentation state, and decision of subsequent production operations in multi-species fermentation processes.