Ultrasound‐assisted protein extraction from mosambi peel support vector regression and genetic algorithm‐based modeling and optimization

Abstract

This experiment aimed to optimize the protein extraction from mosambi peel powder using Support Vector Regression and Genetic Algorithm. The protein was extracted using the pH shift method with ultrasound treatment. A model based on Support Vector Regression (SVR) was created to predict protein yield extraction from mosambi peel powder based on extraction parameters such as particle size (150, 375, and 600 m), ultrasonic (US) time (5, 8, and 11 min), and amplitude (50%, 70%, and 90%). The model accurately predicted both seen and unseen data (R2 = 0.9934 and 0.989, respectively). The trained SVR model was optimized using the Genetic Algorithm to maximize the extraction yield. The optimized model predicted the yield of 55.24% at particle size 187.496 μm, time 10.79 min, and 53.15% amplitude. The model prediction was validated experimentally, and 54.24% yield with 90.2% crude protein was obtained at the optimal settings. The recovered mosambi peel protein isolate shows 26.3% emulsion stability, 1.34 ml/g water‐holding capacity, 3.1 ml/g oil‐holding capacity, and 30% foaming capacity. The XRD of the protein isolate suggested the dominating amorphous nature with a crystallinity index of 27.35. In conclusion, the mosambi peel protein isolate has good functional properties and can be used as a functional ingredient in the food industry. Practical Implications The fruit industry by‐products have been used for the extraction of phytochemicals and dietary fibers. However, there are a few investigations on protein extraction from these by‐products. Protein derived from these sources can be employed as a functional component in the development of new food products or to boost the protein content of existing ones. This will also aid farmers and producers by reducing waste and increasing the value of juice by‐products.