Ant colony algorithm-enabled back propagation neural network and response surface methodology based ultrasonic optimization of safflower seed alkaloid extraction and antioxidant

Safflower seed extract has antioxidant, anti-inflammatory and other effects. The difference in antioxidant activity of safflower seed extracts obtained by different prediction models remains to be studied. In the present study, single factor screening of safflower seed was used to determine the level of optimization, and a four-factor and three-level optimization system was developed. The optimal extraction conditions were obtained by response surface methodology (RSM) and Ant colony algorithm-enabled back propagation neural network (ACO-BPNN) optimization, respectively. The two optimal conditions were verified and in vitro antioxidant studies were performed. Based on the single factor results, the optimization range was determined to be the ultrasonic frequency of 160–320 W, the ethanol concentration of 70–90 %, the liquid-to-material ratio of 15–25 mL/g, and the extraction temperature of 60–80 °C. Validation of both models and in vitro antioxidant results showed that ACO-BPNN results were superior to RSM. The prediction (actual) conditions of ACO-BONN were ultrasonic frequency of 303.4576 (320) W, ethanol concentration of 79.6549 (80)%, liquid-to-material ratio of 15.0027 (15) mL/g and extraction temperature of 82.2419 (80) °C. The predicted value of ACO-BPNN was 0.2042, the actual value was 0.1871 ± 0.0110, and the results of DPPH and FRAP were 2.28 ± 0.42 and 1.64 ± 0.12 mmol Trolox/g. In vivo antioxidant results showed that safflower seed extract could significantly increase the level of SOD in blood (P < 0.01), reduce the level of MDA (P < 0.05). Simultaneously, it significantly increased the expression of HO-1 (P < 0.01), and inhibited the expression of NOX2 (P < 0.05) in brain tissue. It provids experimental basis for clinical development of safflower seed. It provides reference for the large-scale extraction of safflower seed.