Mapping Microbial Interaction Networks of Alcoholic Fermentation

Stacking fermentation in sauce‐flavor liquor production involves complex microbial interaction networks that drive variation in chemical compounds of the liquor. However, how to infer such meaningful interaction networks from high‐dimensional common microbial abundance data is computationally challenging. We propose a generalized statistical mechanics model to coalesce all microbes into multilayer and multiplex networks (idopNetworks) that capture a full set of interaction properties. By analyzing the abundance data of microbes derived from 16S rRNA gene and ITS amplicon sequencing across four subsequent batches of liquor brewing (denoted as XD, ZD, YD, and ED), we find that fungal species more aggressively regulate microbial communities than bacterial species. Lactic acid bacterial species that metabolize flavor‐related chemicals in the liquor are inhibited mostly by fungal species. Although it is a hub in the second stage ZD, Lactobacillus is inhibited by a non‐brewing yeast Issatchenkia in XD. A newly annotated lactic acid bacterium Acetilactobacillus jinshanensis increases its abundance with subsequent batches, but it is inhibited by fungal species. We further reconstruct causal idopNetworks to explain how microbial communities and chemical components of the liquor are reciprocally influenced. idopNetworks could serve as a powerful tool for detecting the optimal composition of microbial communities for improving the production and quality of sauce‐flavor liquor and beyond.