Presentation Information
[2Biocat-07]Developing a Structure-Based AI Platform to Engineer Diverse Enzymatic Reactions
○Christopher J. Vavricka1 (1. Tokyo University of Agriculture and Technology (Japan))
Keywords:
Enzyme engineering,Enzymatic reaction prediction,AI model development
[Purpose]
Development of structure-based AI models can enable the engineering of new enzymatic reactions. Accordingly, our lab has developed AI-guided workflows that expand the accessible enzymatic reaction space for the production of diverse non-natural products.
[Method]
Structure- and machine learning-based approaches were applied to select and engineer enzymes from multiple classes, including peroxidases (EC 1.11.1), monooxygenases (EC 1.14), FAD-dependent halogenases (EC 1.14.19), esterases (EC 3.1.1) and PLP-dependent decarboxylases/oxidases (EC 4.1.1). In parallel, we developed the AI framework GEnESIS (Graph-based Enzyme Evolution with Structure-Informed Scoring) for advanced structure-based enzyme design and reaction prediction.
[Results]
Our platform enabled the engineering of enzymes for diverse reaction types, including decarboxylation, oxidation, halogenation, hydrolysis, and hydroxylation. Notably, our computational workflows successfully guided the engineering of regioselective hydroxylation reactions, as well as peroxidase-mediated substrate activation for theranostic applications. These approaches are now being extended toward more complex coupling reactions.
[Consideration]
While several projects remain at an early stage, the ability to systematically engineer multiple enzyme classes across different EC number groups demonstrates the robustness and versatility of a structure-based AI approach.
[Conclusion]
This presentation highlights the potential of AI-guided enzyme engineering as a means to expand the accessible reaction space and enable the development of novel enzymatic reactions.
Development of structure-based AI models can enable the engineering of new enzymatic reactions. Accordingly, our lab has developed AI-guided workflows that expand the accessible enzymatic reaction space for the production of diverse non-natural products.
[Method]
Structure- and machine learning-based approaches were applied to select and engineer enzymes from multiple classes, including peroxidases (EC 1.11.1), monooxygenases (EC 1.14), FAD-dependent halogenases (EC 1.14.19), esterases (EC 3.1.1) and PLP-dependent decarboxylases/oxidases (EC 4.1.1). In parallel, we developed the AI framework GEnESIS (Graph-based Enzyme Evolution with Structure-Informed Scoring) for advanced structure-based enzyme design and reaction prediction.
[Results]
Our platform enabled the engineering of enzymes for diverse reaction types, including decarboxylation, oxidation, halogenation, hydrolysis, and hydroxylation. Notably, our computational workflows successfully guided the engineering of regioselective hydroxylation reactions, as well as peroxidase-mediated substrate activation for theranostic applications. These approaches are now being extended toward more complex coupling reactions.
[Consideration]
While several projects remain at an early stage, the ability to systematically engineer multiple enzyme classes across different EC number groups demonstrates the robustness and versatility of a structure-based AI approach.
[Conclusion]
This presentation highlights the potential of AI-guided enzyme engineering as a means to expand the accessible reaction space and enable the development of novel enzymatic reactions.
Comment
To browse or post comments, you must log in.Log in