Dual Modeling for Optimization of Corosolic Acid Extraction Conditions From Lagerstroemia speciosa Leaves Using RSM and ANN–GA

Abstract

Lagerstroemia speciosa leaf (banaba; LSL) extract has been used as a functional ingredient for blood glucose control, and corosolic acid (CA) is commonly used as a marker compound for standardization of LSL-based products. This study screened ethanol concentration (0%–100%) and extraction time (1–2 h) at 80°C and optimized CA content using central composite design (CCD) based response surface methodology (RSM) and an artificial neural network coupled with a genetic algorithm (ANN–GA). CA was the highest at 75% and 100% ethanol, with no significant difference among them. In CCD modeling, ethanol concentration was significant, whereas extraction time was not significant within the experimental range. The RSM optimum (86.4% ethanol, 109.4 min) yielded 1.028 ± 0.09 μg/mg (3.9% relative error), while the ANN–GA optimum (98.3% ethanol, 106.5 min) yielded 0.974 ± 0.03 μg/mg (2.3% relative error). Extracts prepared under the validated RSM and ANN–GA conditions also showed strong in vitro α-glucosidase inhibitory activity, with IC₅₀ values of 15.55 ± 1.30 and 6.89 ± 0.24 μg/mL, respectively. These results provide a methodological basis and supporting data for efficient standardization and systematic optimization of LSL extract production as a functional food ingredient.