Rational design for improved thermostability of methionine adenosyltransferase based on FoldX, Rosetta, and...

S-adenosyl-L-methionine (SAM) is a vital functional food ingredient and nutraceutical, but its industrial production is severely limited by the poor thermal stability of methionine adenosyltransferase (MAT). To overcome this bottleneck, this study presents a rational design strategy that integrates physical force fields with deep learning algorithms. Based on the crystal structure, molecular docking and B-factor analysis identified 11 flexible hot spots near the substrate-binding pocket. A multi-level virtual screening workflow was implemented: a high-throughput saturation mutagenesis primary screen was performed using FoldX, followed by a refined evaluation using Rosetta, resulting in 18 selected candidates from 209 mutants. To mitigate the bias of individual algorithms, a multidimensional orthogonal validation system was introduced incorporating deep learning-based tools. Finally, six identified core mutants were integrated into a computational framework involving all-atom molecular dynamics for 200 ns, which was complemented by subsequent experimental verification. The optimal mutant achieved synergistic improvements in both catalytic activity and thermostability. Compared with the wild-type, the optimal mutant exhibited a 4.73 °C increase in melting temperature, a 93.49% extension in half-life, and a 70.65% increase in specific activity. Mechanistically, the enhanced performance of this mutant stems from increased local structural compactness and reduced conformational fluctuation near the catalytic region. This work provides a thermostable MAT mutant for the efficient biosynthesis of SAM and a generalizable computational design framework for balancing stability, solubility, and activity in industrial food enzyme engineering.