Integrating AI/ML-Enabled Sonochemical Bioprocessing for Sustainable Valorization of Sweet Potato Leaves Waste into Functional Food Ingredients: Structural, Nutritional, and Metabolic Insights

Abstract Sweet potato leaves are agro-waste produced in substantial amounts during the post-harvest period. In this study, sweet potato leaves were converted into functional food ingredients to achieve zero waste, using AI/ML in the process system, and their metabolic potential was evaluated. The protein- and fiber-rich fractions were extracted using single-stage sonication, and the fiber-rich residue (OPSF) was subsequently treated by dual-stage citric acid-assisted sonication (TPSF). ANN, Bayesian, and NSGA-II models were independently coupled with the extraction process to identify optimal processing conditions for protein fractions, with training R² values of 0.980, 0.998, and 0.942 and cross-validation values of 0.854, 0.872, and 0.807, respectively. The protein extract exhibited a diverse HPLC amino acid profile. ATR-FTIR from 4000–400 cm⁻¹ was employed to evaluate the effects of sonication on the protein extract and the introduction of C = O, indicating enhanced hydrophilicity of the citric acid-treated fiber fraction (TPSF). TPSF showed a 19.4% increase in glucose adsorption capacity compared to OPSF. TPSF also increased cholesterol adsorption capacity by 10.3% at pH 2.0, and at pH 7.0 it increased from 25.03 ± 0.17 to 26.15 ± 0.41 mg/g compared to OPSF, reflecting its potential as a functional food ingredient for improving metabolic function.