Modeling and Optimization of α‑Amylase Immobilization on Chitosan-Based Supports: A Comparison of Response Surface...

α-Amylases are among the most important enzymes used in various industries; however, their large-scale application is restricted by limited stability, low reusability, and low activity loss during operation. In this study, immobilization of α-amylase from Priestia megaterium (PmAmy) onto glutaraldehyde-activated chitosan beads (PmAmy@GA-CS) was optimized using response surface methodology (RSM) and machine learning (ML) approaches, including support vector regression (SVR), Gaussian process regression (GPR), and random forest regression (RFR). The effects of the CS concentration, GA concentration, and enzyme binding time were investigated. RSM analysis identified optimal immobilization conditions at 2.6% CS, 0.5% GA, and 6 h of binding time, resulting in approximately 90% immobilization efficiency and high enzymatic activity. ML models captured similar immobilization trends, with GPR exhibiting the highest predictive accuracy (R ² ≈ 0.98) and superior generalization performance compared to SVR, RFR, and RSM. Immobilization produced nearly uniform porous beads and induced a shift in the pH optimum toward slightly alkaline conditions (from pH 7.0 to 8.0), while the temperature optimum remained unchanged at 40 °C. Furthermore, the immobilized enzyme exhibited significantly improved stability, with a 2.1-fold increase in thermal stability and enhanced storage stability. Moreover, PmAmy@GA-CS retained approximately 50% of its initial activity after ten consecutive reuses, demonstrating strong operational stability. Overall, this study highlights the complementary strengths of the RSM and ML in designing robust immobilized enzymes and provides a solid foundation for the development of reusable biocatalysts.