Antioxidant Potential in Kombucha. Application of Penalized Models

This project evaluated the antioxidant potential of kombucha using a three-factor design: extracts from Guayusa (Ilex guayusa) and Moringa (Moringa oleifera) leaves; at variable concentrations (1% and 2%); and with or without pH adjustment to 3.5 using 0.1 N citric acid; across four fermentation times over 14 days. The experiment was conducted at ambient temperature between 19 and 21 °C with 6% honey as a carbohydrate source due to its hypoglycemic potential. Antioxidant activity was quantified in Trolox equivalents (mg/100 mL) measured by DPPH assay. Additionally, physicochemical properties were evaluated: °Brix, acetic acid percentage, and pH, using refractometry (AOAC 934.14), titration (INEN 1091:1984), and potentiometry, respectively. For statistical analysis, penalized regression models were employed with performance metrics RMSE and R², respectively, Lasso (0.0917, 0.9867), Ridge (0.2884, 0.8932), and Elastic Net (0.0925, 0.9866), showing better goodness-of-fit indicators than classical regression models. R Studio was used with the dplyr, MASS, and glmnet packages, evaluating an 80/20 train–test data split. The conclusions highlight the robustness of Elastic Net in predicting antioxidant capacity, offering a reliable statistical tool for future fermentation systems, where the considered factors enable estimation of antioxidant potential.