Accelerated Enzyme Design: A Novel CSFp Pipeline for Enhanced Activity and Enantioselectivity

The development of stable and functionally diverse enzymes has gained significant traction with the advent of structure prediction algorithms such as trRosetta. Building upon this foundation, we demonstrate that AlphaFold2, a state-of-the-art structure prediction tool, can serve as a reliable starting point for enzyme design. Here, we introduce a novel approach termed CSFp, a smart combination of SMD, FuncLib, and pSUFER, for rapid enhancement of enzyme activity. This strategy enables computationally focused, ultra-low throughput screening for enzyme design without relying on experimental structural data. The CSFp pipeline utilizes AlphaFold2 for structure prediction, employing Pointsite for pocket prediction and molecular docking. Furthermore, SMD is employed to identify key residues associated with (un)binding tunnels, while FuncLib predicts small but strategically designed libraries of mutants. The results of these steps are then filtered based on predicted relative reactivity using MD simulations. Suboptimal residues in protein folding are tagged by pSUFER, and the decomposition and recombination of combinatorial mutations are implemented to prevent superior mutants from becoming trapped in local optima. The efficacy of this strategy was tested through the challenging design of D-lactonohydrolase (D-Lac), targeting high catalytic activity and enantioselectivity. This research exemplifies the potential of CSFp to accelerate the design of novel enzymes with desirable properties.