Anti-Inflammatory Potential of Bioactive Metabolites of Lactobacillus -Fermented Milk and Their Analysis through Machine Learning

This is the authors' abstract. We don't add key points for this paper.

This study aimed to evaluate the fermentation capacity of twoLactiplantibacillus plantarum and two Limosilactobacillus fermentum strains to assess their in vitro anti-inflammatory potential and to develop an artificial intelligence (AI)-based predictive model. Milk was fermented with Lactiplantibacillus plantarum J25 (FMJ25), Lactobacillus pentosus J27 (FMJ27), Limosilactobacillus fermentum J23 (FMJ23), or L. fermentum J24 (FMJ24). Fermentation was characterized by pH, titratable acidity, viable cell counts, proteolysis, and metabolites, including exopolysaccharides (EPS) and short-chain fatty acids (SCFA). The inhibition of lipoxygenase (LOX), cyclooxygenase-2 (COX-2), and the reduction of nitric oxide (NO) were evaluated, and a random forest regression model was used to identify the main predictors of activity. Overall, the metabolites present in fermented milks exhibited anti-inflammatory activity; however, FMJ23 and FMJ27 showed the highest LOX inhibition (14.4% and 15.6%, respectively) and NO production (9.68 and 10.60 μM) (p < 0.05). These FM also inhibited COX-2 activity by 60% and 51% (p < 0.05). FMJ23 exhibited the strongest anti-inflammatory effect (0.76), primarily attributable to its peptide fractions (<3 and 3–10 kDa) and EPS. Machine learning identified peptide fractions and butyric acid as the main predictors of LOX and COX-2 inhibition and NO reduction. The findings support FMJ23 as a promising functional starter culture for producing fermented dairy foods with anti-inflammatory potential.