Optimization of the microwave-assisted extraction of bioactive compounds from Satureja hortensis L.: an artificial neural network approach, chemical profile, extraction kinetics, and thermal properties

Introduction Summer savory is still not investigated thoroughly despite its potential and beneficial effects. Therefore, we aimed to investigate its chemical profile, thermal properties, extraction kinetics, and to optimize extraction process. Methods Herein, an artificial neural network (ANN) was used as a nonlinear regression-based optimization model to optimize the microwave-assisted extraction of summer savory leaves. To achieve the goal, 17 experiments were conducted, combining different extraction times (20–40 min), ethanol concentrations (68–80%), and irradiation powers (400–800 W). Investigated responses included total phenolic content (TPC), total flavonoid content (TFC), DPPH, and ABTS assays. Kinetics was investigated by using four models, while thermal behavior was studied using DSC and TGA. Results The highest outputs were: 256.36 mg GAE/g (GAE-gallic acid equivalents) (40 min, 40% ethanol, and 600 W) for TPC, 35.78 mg RU/g (rutin equivalents) (40 min, 60% ethanol, and 400 W) for TFC, 15.89 μg/mL (20 min, 60% ethanol, and 400 W) for DPPH, and 23.06 μg/mL (30 min, 60% ethanol, and 600 W) for ABTS. As a result of optimization, we obtained optimal extraction conditions (40 min, 52.8% ethanol, and 656.1 W) and predicted responses (246.50 mg GAE/g for TPC, 35.66 mg RU/g for TFC, and IC 50 and EC 50 values of 17.79 μg/mL and 25.79 μg mL for DPPH and ABTS assays, respectively). The experimentally obtained values for the investigated responses were 242.25 mg GAE/g, 36.30 mg RU/g, 17.10 μg/mL, and 24.48 μg/mL for TPC, TFC, IC 50 , and EC 50 , respectively. Total content of the quantified phenolic compounds was 91.47 μg/mL. The principal compound was rosmarinic acid (80.99 μg/mL), followed by chlorogenic acid (2.64 μg/mL), rutin (1.45 μg/mL), and apigenin (1.31 μg/mL). Discussion The validity of the developed ANN model has been confirmed experimentally after preparing the extract under optimal conditions. Kinetic modeling showed that Model II provided the best fit for TFC, while Model IV provided the best fit for TPC. Optimally prepared extract showed high antioxidant activity and thermal behavior and could be used in food and pharmaceutical industry as an additive.