Microbiomics and machine learning-assisted approaches reveal amino acid patterns in high-temperature Daqu

Abstract

Amino acids are crucial nitrogen sources in high-temperature Daqu (HTD), and they can significantly influence the quality of HTD. This study investigated amino acid patterns by analyzing fermentation parameters and microbial communities. Correlation analysis and machine learning methods were utilized to identify 6 key amplicon sequence variants (ASVs) from Saccharopolyspora, Bacillus, Lactobacillus, and Virgibacillus. Functional predictions revealed that these ASVs exhibited high enzymatic activity in amino acid metabolic pathways during the first and second flipping stages, consistent with observed metabolic phenotypes. The ensemble machine learning models successfully predicted the concentrations of most amino acids in HTD, with coefficients of determination (R²) ranging from 0.70 to 0.95, and the robustness of the models was validated in an independent HTD dataset. This study provides a strategy for predicting and regulating metabolite profiles in traditional fermented foods.