Impact of Aging Microbiome on Metabolic Profile of Natural Aging Huangjiu through Machine Learning

Abstract

Aging is a time-consuming step in the manufacturing of fermented alcoholic beverages. Natural-aging huangjiu sealed in pottery jars was taken as an example to investigate the changes of physiochemical indexes during aging and to quantify intercorrelations between aging-related factors and metabolites through machine learning methods. Machine learning models provided significant predictions for 86% of metabolites. Physiochemical indexes well reflected the metabolic profile, and total acid was the most important index that needed to be controlled. For aging-related factors, several aging biomarkers of huangjiu were also well predicted. Feature attribution analysis showed aging year was the most powerful predictive factor, and several microbial species were significantly associated with aging biomarkers. Some of the correlations, mostly connected to environmental microorganisms, were newly found, showing considerable microbial influence on aging. Overall, our results reveal the potential determinants that affect the metabolic profile of aged huangjiu, paving the way for a systematical understanding of changes in metabolites of fermented alcoholic beverages.

Keywords

• aging microbiome
• metabolic profiling
• multi-omics
• machine learning
• aging regulation
• fermented alcoholic beverages