Response Surface Methodology (RSM) and Artificial Neural Network (ANN) were employed to optimize saponin extraction from North Atlantic Sea cucumber (Cucumaria frondosa). Ultrasonication-assisted ethanol-based extractions were used in a second-order polynomial and 3n full factorial RSM interconnected with neural design ANN model. A 3-10-2 neural network architecture was constructed to predict the relationship between the independent variables and bioactive compounds adequately. The extracts with the highest frondoside A yield were characterized for different triterpene glycosides (saponins) by high-resolution mass spectrometry (HRMS). A total of ten saponins were detected and tentatively identified including fallaxoside, frondoside, cucumarioside, cercodemasoide, colochiroside, and lefevreioside, with two unknown saponins. Six of the saponins were detected in C. frondosa extracts for the first time. The extract of body walls have a higher concentration of frondoside A (0.73 mg/g DW) than internal organs and tentacles (flowers or aquapharyngeal bulb). The optimized extracts exhibit a significantly higher concentration of polyphenols and saponins when compared with extracts prepared from conventional methods. The ANN model demonstrated a low p and high f values to indicate a perfect good fit for RSM model. The advanced knowledge of saponins of C. frondosa can contribute to the development of novel functional foods and ingredients from C. frondosa and their processing byproducts.
Read full abstract