Search

Streptomyces mobaraensis transglutaminase (smTG) has been widely used in the food processing industry for protein crosslinking. However, its poor thermostability becomes a major obstacle to further applications. It is significant to develop a feasible strategy to improve the thermostability of smTG. Here, we developed a computational model based on the Siamese graph neural network framework to identify residues critical to the thermostability of smTG by predicting changes in folding free energy (ΔΔG) between the wild type and mutants. Four candidate residues were selected for mutation experiments, and the single mutant H44C exhibited 2.7 U mg−1 residual enzyme activity after 10 min incubation at 60 °C, twice as much as the wild type. Mutants Q74F and E87W also exhibited 3- and 2.3-times greater activity after 20 min. Molecular dynamics (MD) simulations revealed that smTG mutants (H44C, Q74F, and E87W) improved thermostability by enhancing hydrogen bond interactions, increasing additional residue interactions, and reducing loop flexibility. The MM-GBSA calculation demonstrated that mutants H44C and Q74F enhanced binding affinity with the substrate, and six residues crucial for substrate binding were identified. This study combines computational analysis with mutation assays for the rational design of smTG and offers a facile and efficient strategy to understand and improve the thermostability of proteins for industrial applications. [ABSTRACT FROM AUTHOR]