Fermentation modeling and machine learning for flavor prediction in low-sodium radish paocai with potassium chloride substitution.

Abstract

To address the challenge of high sodium in paocai, this study evaluated the partial substitution of NaCl with KCl during radish paocai fermentation, focusing on microbial kinetics and flavor. The methodology integrated microbial growth modeling with comprehensive flavor analysis (HS-SPME-GC-MS, HS-GC-IMS, E-tongue) and Random Forest (RF) machine learning. Substituting 30% NaCl with KCl (K30) significantly increased mannitol and glutamic acid, enhancing desirable fresh, sweet, and umami tastes. RF modeling identified erucin, 1-hexanol, 3-methylbutan-1-ol, and 2-methoxy-4-vinylphenol as potential key aroma compounds. The K30 treatment also improved the aroma profile by increasing volatile compounds associated with cabbage, fruit, and sweet notes. Conclusively, sensory analysis confirmed that K30 paocai exhibited superior sourness, sweetness, umami, and overall acceptability. These findings support that 30% NaCl substitution with KCl is a valuable strategy for producing low-sodium radish paocai with an enhanced flavor profile, offering a practical framework for sodium reduction across traditionally fermented foods.