Integrated multi-omics and machine learning analysis of 3-phenyllactic acid variation across diverse kimchi types

3-Phenyllactic acid (PLA) is a bioactive metabolite in kimchi with antifungal activity. We investigated PLA formation across kimchi types using targeted metabolomics and bacterial community profiling, including lactic acid bacteria (LAB) composition. Baechu, chonggak, and radish kimchi showed high PLA levels, whereas mustard and godeulppaegi showed lower levels associated with precursor imbalance. Watery kimchi types, including dongchimi and nabak, showed low PLA levels, likely due to high water content. Multivariate analysis based on metabolite profiles and microbial composition classified samples into four groups reflecting PLA levels, precursor balance, and ingredient origin. Random Forest and XGB classified the four groups with out-of-fold accuracies of 0.994 and 0.929, respectively, while comparisons with simpler linear and kernel-based models supported the robustness of predictive patterns. Integrated analysis indicated that PLA accumulation was associated with precursor availability, ingredient matrix, and LAB composition, providing an integrated assessment of factors associated with PLA distribution among kimchi types.