Computational dual-loop frameworks bridging single-enzyme design and cascade tunnel network engineering for next-generation biosynthetic systems.

Advances in computational biology are reshaping enzyme engineering from empirical practice into a predictive, model-driven discipline. Traditional strategies such as directed evolution and rational design have improved catalytic properties but often fail to resolve system-level bottlenecks involving stability, cofactor incompatibility, and intermediate leakage. This review summarizes recent advances in computational enzyme engineering, with a focus on structure prediction, mutation-effect modeling, multiscale molecular simulations, and multi-enzyme network engineering. We frame the reviewed studies within a dual-loop Design-Build-Test-Learn (DBTL) perspective to connect molecular-level enzyme design with pathway-level integration. The first loop summarizes computational strategies for single-enzyme optimization, including deep learning-based structure and fitness prediction, molecular dynamics (MD) simulations, and quantum mechanics/molecular mechanics (QM/MM) analyses that elucidate structure-function relationships and guide iterative refinement. The second loop extends these approaches to multi-enzyme systems, reviewing recent studies on substrate tunnel engineering, electrostatic coupling, and temporal coordination that facilitate efficient intermediate transfer and balanced flux in enzyme cascades. In contrast to existing reviews that primarily focus on individual computational methods or single-enzyme optimization, this review emphasizes the integration of mechanistic enzyme modeling with multi-enzyme network design within a unified DBTL framework. By integrating advances across these scales, we provide a structured overview of current capabilities, limitations, and emerging directions in computationally guided enzyme engineering.