Optimization of sequential solvent extraction of collagen and ommochrome from Indian squid, Uroteuthis duvauceli

This study explores and compares the predictive efficiency of one-factor-at-a-time (OFAT), response surface methodology (RSM), and various machine learning (ML) techniques in optimizing ultrasound and enzyme-assisted sequential extraction of collagen and ommochrome from Indian squid skin. A 3-level, 2-factor experimental design was implemented, focusing on two key variables: ultrasound exposure time and buffer temperature. Ultraviolet-visible absorbance readings at 230 and 280 nm served as the primary response indicators. Initial extraction conditions were identified using the OFAT approach. RSM modeling effectively captured the interactive effects of the two independent variables, confirming the suitability of a quadratic model. To further evaluate predictive accuracy, ML models including artificial neural networks (ANN), k-nearest neighbors, and Random Forest were trained and compared against the RSM model. Among these, the ANN model demonstrated superior performance, as evidenced by the highest coefficient of determination (R2), indicating its enhanced capability for accurate prediction of extraction outcomes.