What question did this study set out to answer?

The aim is to optimize the extraction of bioactive compounds from Ballota limbata using sustainable methods.

April 4, 2026Open Access

Artificial neural networks (ANN) and response surface methodology (RSM) for optimizing antioxidant and α-amylase inhibitory compound extraction from Ballota limbata

Key Points

The aim is to optimize the extraction of bioactive compounds from Ballota limbata using sustainable methods.
Compared microwave-assisted extraction (MA) and conventional heat-assisted extraction (HA).
Utilized 70% hydro-ethanol solvent and optimized processes with response surface methodology (RSM).
Applied a Box-Behnken design within an artificial neural network (ANN) framework to analyze extraction efficiency.
Microwave-assisted extraction yielded lower total phenolic content (3.27 mg GAE/g) than heat-assisted extraction (3.85 mg GAE/g).
Heat-assisted extraction (86.22%) had higher alpha-amylase inhibitory activity compared to microwave-assisted extraction (60.99%).
ANN demonstrated superior predictive accuracy with higher R2 and lower error metrics compared to RSM.

Abstract

Ballota limbata (syn. Otostegia limbata), a medicinal plant with various ethno-therapeutic uses, was investigated for the recovery of its bioactive compounds using sustainable extraction strategies. This study compared microwave-assisted extraction (MA) and conventional heat-assisted extraction (HA), employing 70% hydro-ethanol as the solvent and optimizing both techniques via response surface methodology (RSM) based on a 3-factor-3-level Box-Behnken design and Artificial Neural Network (ANN), comprising three input variables and one hidden layer. The response factors included total phenolic content (TPC), total flavonoid content (TFC), DPPH radical scavenging activity (RSA), and α-amylase inhibitory activity (AAI). MA under optimized conditions (power 220 W, time 20 s, solvent-to-solid ratio (SSR) 40 mL/g) yielded TPC, TFC, RSA, and AAI as 3.27 mg gallic acid equivalents/g (GAE/g) dry weight (DW), 2.43 mg rutin equivalents/g (RE/g) DW, 2.16 mg ascorbic acid equivalents/g (AAE/g) DW, and 60.99%, respectively; while HA under optimized conditions (temperature 57 °C, time 150 min, SSR 40 mL/g) produced 3.85 mg GAE/g DW TPC, 2.40 mg RE/g DW TFC, 2.50 mg AAE/g DW RSA, and 86.22% AAI. Statistical modeling showed that MA and HA both were reliable, with low p-values, strong fit statistics, and small differences between predicted and observed values. The yield of bioactives obtained from HA was higher than MAE, but MAE was more efficient in terms of time, which aligns directly with green chemistry principles. Both ANN and RSM performed well; however, ANN had superior predictive accuracy with higher R2, lower root mean square error (MSE), and mean absolute error (MAE). The independent t-test confirmed that there is a significant difference (p < 0.001) between HA and MA when compared for AAI. Thus, the study provided valuable insights into sustainable phytochemical recovery from B. limbata and supported the industrial application of MA for rapid, eco-friendly extraction of medicinal plant constituents.

Bookmark

View Full Paper

Cite This Study

Namra et al. (Thu,) studied this question.

synapsesocial.com/papers/69d0ae94659487ece0fa491a https://doi.org/https://doi.org/10.1038/s41598-026-37738-x

Bookmark

View Full Paper