Remarkably efficient metabolic enzymes play a pivotal role in converting readily available, simple starting materials into valuable products within microbial cells. Furthermore, these enzymes facilitate a diverse array of chemical reactions, typically operating under mild reaction conditions and exhibiting high selectivity (Hollmann and Fernandez-Lafuente, 2021). These advantages have led to the incorporation of enzymes in various industries, including food (Flynn et al., 2021; Lin et al., 2022; Punia, 2020; Rentschler et al., 2015; van Donkelaar et al., 2016; Xu et al., 2020), agricultural (Costa-Silva et al., 2021; Sijinamanoj et al., 2021; Tingley et al., 2021), cosmetics (Almeida et al., 2021; Fournière et al., 2021) and pharmaceuticals (Meghwanshi et al., 2020; Park et al., 2017; Rosenthal and Lütz, 2018). Regrettably, enzymes come with certain drawbacks such as low turnover rates, limited stability and a narrow substrate scope. Consequently, there is a necessity for enzyme engineering to tailor them for diverse industrial applications (Li et al., 2018b).

Utilizing both directed evolution and semi-rational approaches, enzyme engineering involves iterative steps and mutation libraries, with the process repeated until the desired variant is obtained (Dinmukhamed et al., 2021; Gargiulo and Soumillion, 2021). However, the unknown structure-function relationship introduces uncertainty in discovering desired protein variants through a few rounds of iterative mutagenesis. It's more likely that the researchers may never come across it even with a comprehensive library, leading them to rely on insights from rational design and evolutionary analysis (Blazeck et al., 2022). In contrast, rational enzyme design stands out for its ability to provide higher predictive accuracy and streamline the screening library (Jumper et al., 2021; Kuhlman and Bradley, 2019). Proposed mutations are introduced after evaluation and design, increasing the likelihood of beneficial mutations while saving time and labor. When high-throughput screening is not feasible, this approach stands out as particularly useful (Cui et al., 2022; Steiner and Schwab, 2012).

Based on the design principles, rational computational enzyme design is divided into three classes, structure-based computational design, sequence-based computational design, and data-driven machine learning computational design. Here, we assess the recent advancements in rational computational enzyme design, emphasizing the effectiveness and versatility of different approaches in enhancing catalytic activity, stability and substrate selectivity. Successful cases are cited to illustrate these improvements. Lastly, the article provides a thorough analysis of these approaches, highlights existing challenges and potential solutions, and offers insights into future development directions.