AlphaFold2 and Deep Learning for Elucidating Enzyme Conformational Flexibility and Its Application for Design

Abstract

The recent success of AlphaFold2 (AF2) and other deep learning (DL) tools in accurately predicting the folded three-dimensional (3D) structure of proteins and enzymes has revolutionized the structural biology and protein design fields. The 3D structure indeed reveals key information on the arrangement of the catalytic machinery of enzymes and which structural elements gate the active site pocket. However, comprehending enzymatic activity requires a detailed knowledge of the chemical steps involved along the catalytic cycle and the exploration of the multiple thermally accessible conformations that enzymes adopt when in solution. In this Perspective, some of the recent studies showing the potential of AF2 in elucidating the conformational landscape of enzymes are provided. Selected examples of the key developments of AF2-based and DL methods for protein design are discussed, as well as a few enzyme design cases. These studies show the potential of AF2 and DL for allowing the routine computational design of efficient enzymes.

Scaffold-Lab: Critical Evaluation and Ranking of Protein Backbone Generation Methods in A Unified Framework

“…Figure 2E shows the differences between the distribution of pdb-TM and novelty score, where we conducted one-sided Mann-Whitney U test 72 across different groups distinguished by with or without the penalty term. The results show that some of the groups have significant differences with p-values < 0.001, including FrameDiff , Chroma in medium-protein group (6.11 × 10 −8 and 5.19 × 10 −4 , respectively) and RFdiffusion, Chroma in long-protein group ( 2.46 × 10 −1 5 and 3.33 × 10 −1 6 , respectively) . We also noticed that RFdiffusion and FrameFlow in the short protein group did not show significant differences in results with or without penalty, indicating that most proteins within those groups pass the designability threshold, which aligns with our previously proposed hypothesis.…”

Perspectives on Computational Enzyme Modeling: From Mechanisms to Design and Drug Development

“…Nevertheless, challenges remain, including the need to incorporate aspects of protein dynamics, , allostery and entropy-enthalpy compensation into enzyme design principles, , as well as the design of enzymes capable of catalyzing multistep reactions. , From the perspective of computational enzymology, the physics-based computational methods discussed above have proven instrumental in exploring these complexities of enzymes and understanding their mechanisms . Consequently, these methods need to be fully integrated with machine learning techniques into the enzyme design framework to streamline the effective refinement of the catalytic properties of the designed enzymes. ,,,,,,, On the activity prediction side, regression models, including linear regression and neural networks, have long been used to decipher the sequence–activity relationship of enzymes. Such relationships have then been used to guide (a) the optimization of enzyme variants , and (b) the directed evolution of enzyme activity ,, and product enantioselectivity. , …”

“…On the structure side, achievements such as AlphaFold2 and RoseTTAFold2 have significantly improved the accuracy of protein structure prediction. These structure prediction techniques have the potential to greatly improve our ability to design robust enzymes with high success rates . Nevertheless, challenges remain, including the need to incorporate aspects of protein dynamics, , allostery and entropy-enthalpy compensation into enzyme design principles, , as well as the design of enzymes capable of catalyzing multistep reactions. , From the perspective of computational enzymology, the physics-based computational methods discussed above have proven instrumental in exploring these complexities of enzymes and understanding their mechanisms .…”

Fishing for Catalysis: Experimental Approaches to Narrowing Search Space in Directed Evolution of Enzymes

“…Future work will also shed light onto how close the transition state analogs need to be to the transition state and how sensitive the approach is to it. Further experimental validation of this approach can result in rapid advance of ML approaches, as at least dynamic information can be obtained using already established tools . The principal limitation of the approach is data availability, and while NMR works well for smaller, soluble proteins, its applications for larger enzymes is limited.…”