Transition State-Based Computational Enzyme Design

Our approach to computational protein design is physics-based. We develop a software called Proteus, allowing both physics-based energy evaluation and sequence-conformation exploration. Unlike knowledge-based models, a physics-based energy function facilitates the consideration of unusual chemical entities, such as novel ligands or transition states. Additionally, the adaptive landscape flattening method allows direct sampling on free energy difference between two states. We show here how these ingredients combined can benefit enzyme design. As an illustration, we revisit the stereospecificity inversion of tyrosyl-tRNA synthetase. Following a tutorial presentation, we explore various design criteria that can be related to enzyme experimental parameters. Our model is able to retrieve the native sequence when targeting L-tyrosine. Mutations predicted to favor D-tyrosine are proposed.