A hybrid statistical-machine learning framework for optimization, prediction, and model-fitting of xylose and total sugar production from sugarcane bagasse using recombinant Escherichia coli-mediated hydrolysis.

Efficient conversion of lignocellulosic biomass utilizing microbes to release fermentable sugars remains a major challenge. This study reports the integrated process involving Trichoderma reesei for pretreatment and recombinant whole-cell Escherichia coli expressing endo-β-1,4-xylanase for hydrolysis. Maximum cellulase activity 178.61 ± 1.98 IU/g dry substratewas observed on the 10th day at pH 5.5 during pretreatment. A Box-Behnken Design was applied to optimize the reducing sugar (xylose equivalent) and total sugar production during hydrolysis. The maximum reducing sugar (xylose equivalent) was 589.8 µg/mL, and the total sugar was 7785.6 µg/mL, achieved after 6 h and 24 h of induction, respectively. ANOVA demonstrated significant models with strong predictive performance, R² = 0.9866 for reducing sugar (xylose equivalent) and 0.9815 for total sugar. Machine learning predictions using Support Vector Machine (SVM) and Artificial Neural Network (ANN) identified SVM as the superior predictive model. Structural changes in the process strategies were confirmed using compositional analysis, XRD, FE-SEM, and FTIR. Overall, this strategy provides an efficient insight into the biological degradation of SCB for sustainable sugar production.