Exploring the potential of Pediococcus pentosaceus PPF28-8 in a rapid fermentation model for soy sauce flavor enhancement using machine learning

Abstract

This study developed a rapid fermentation model for soy sauce using Daqu enzymatic hydrolysates and investigated the role of Pediococcus pentosaceus in modulating physicochemical properties and volatile compounds. Daqu hydrolysates from different enzymatic hydrolysis times were used as substrates. The results showed that P. pentosaceus PPF28-8 contributed significantly to acidification, increasing total acidity and amino acid nitrogen levels. Glucose supplementation promoted microbial growth, lactic acid production, and amino acid metabolism. Volatile compound analysis revealed a notable increase in aroma compounds, including alcohols, aldehydes, and ketones, enhancing the flavor profile. Machine learning methods, including partial least squares discriminant analysis (PLS-DA) and principal component analysis (PCA), were applied to compare the rapid fermentation model with the validation experiment. Consistent patterns emerged, confirming the model's predictive ability for P. pentosaceus PPF28-8's aroma production. These findings underscore the model's potential for precise control over fermentation dynamics and flavor enhancement, offering an efficient approach for optimizing soy sauce production while preserving flavor complexity.