Coupling ensemble learning with multi-omics: a novel data-driven strategy for Daqu quality assessment and validation.

Area of Science:

• Food Science and Technology
• Microbiology
• Biotechnology

Background:

• Traditional fermented foods like Daqu rely on complex microbial communities.
• Current quality control methods for Daqu are subjective and lack repeatability.
• Industrial demands necessitate a shift from experience-driven to science-driven quality assessment.

Purpose of the Study:

• To establish a robust quality evaluation model for medium-high temperature Daqu used in Baijiu production.
• To integrate multi-omics data and machine learning for comprehensive Daqu analysis.
• To identify key functional microorganisms and elucidate their roles in Daqu fermentation.

Main Methods:

• Multi-omics analysis (genomics, transcriptomics, etc.) to profile microbial communities and gene expression.
• Development of an ensemble machine learning model combining Gradient Boosting, XGBoost, and Linear Regression.
• Targeted inoculation with identified core functional fungi (Rhizomucor, Saccharomycopsis) for validation.

Main Results:

• Eukaryotic microorganisms with 'low abundance and high expression' were identified as primary drivers of Daqu fermentation.
• Rhizomucor and Saccharomycopsis were confirmed as key contributors to hydrolase activity.
• The ensemble learning model enabled rapid quality grading of Daqu, and targeted inoculation improved its quality.

Conclusions:

• A systematic paradigm of 'multi-omics analysis → machine learning modeling → microbial targeted validation' was established.
• This approach transforms subjective 'ecological art' into quantifiable 'synthetic ecological engineering' for traditional fermented foods.
• The methodology offers a generalizable solution for quality evaluation challenges in the traditional fermented food industry.