Integration Boruta-SVM-RFE model with flavoromics: Unveiling the aroma enhancement mechanism of enzyme-microbe synergy in soymilk.

Abstract

Traditional methods for removing the beany flavor from soymilk often result in losing its original flavor, resulting in a bland flavor profile. So, this study proposes a synergistic aroma-enhancing strategy combining enzymolysis with fermentation. Flavoromics and machine learning techniques were utilized to elucidate the mechanisms by which enzymolysis co-fermentation impacts soymilk flavor. Results indicated that enzymolysis co-fermentation significantly enhances the functional and storage characteristics, with substantial increases in free fatty acids and free amino acids contents, providing abundant precursors for aromatic compounds formation. A 80 volatile and 76 non-volatile flavor compounds were detected by GC-IMS and LC-MS. To accurately identify key flavor compounds for the "soy aroma", a progressive machine learning selection strategy combining Boruta with SVM-RFE was employed. Key contributors identified include γ-linolenic acid, glutamic acid, 2,3-butanedione, and 3-hydroxy-2-butanone. Important pathways for aroma formation include Ehrlich pathway, Maillard reaction, lipid oxidation, and β-oxidation. This research provides theoretical insights and technical support for the development of high-quality soymilk.