Model-driven high-level expression of β-mannanase in Pichia pastoris for degrading mannan in palm kernel expeller

• Deep learning model (PMEPE) boost CsMan5A-Q7S expression to 16.38 g/L. • Deep learning model (PMEPE) enhanced CsMan5A-Q7S activity to 147,597 U/mL. • Cell wall engineering increased CsMan5A-Q7S expression to 18.70 g/L. • CsMan5A-Q7S hydrolysis increased protein and oil content, and reduced fiber content of PKE. Palm kernel expeller (PKE), a copious byproduct, is limited in its application to non-ruminant diets due to the presence of antinutritive factors, predominantly galactomannan, necessitating effective enzymatic hydrolysis by β-mannanase to enhance its nutritional value. Here, a cross-strategy combining the p redicting m utations with e nhanced p rotein e xpression (PMEPE) model with B-factor analysis was employed to rationally modify the gene encoding β-mannanase CsMan5A in Pichia pastoris . The expression level of mutant CsMan5A-Q7S improved 17% compared to the wild type, reaching 16.38 g/L, which resulted in an increased enzyme activity titer of 147,597 U/mL and specific enzyme activity of 9010.81 U/mg, respectively increasing by 32.87% and 13.81% compared to the wild-type. Then, the expression of CsMan5A-Q7S was further elevated to 18.70 g/L through cell wall engineering. Computational simulations explained the mechanism of the superior thermostability and activity observed in the CsMan5A-Q7S mutant, which presented a more stable overall structure than the wild type. The residue left after hydrolysis by CsMan5A-Q7S had lower fiber content, as well as higher protein content, thereby effectively enhancing its feed quality. This study achieved exceptional phenotypic enhancement of β-mannanase production in Pichia pastoris by an effective strategy through computational modeling, protein engineering, and cell wall modification strategies.