Protein Design at Library Scale

Abstract

Recent advances in de novo protein design have made it increasingly feasibleto create proteins with novel functions, driven by rapid progress in both com- putational modeling and high-throughput experimentation. Modern tools can explore vast sequence-structure spaces and evaluate biomolecular interactions, while experimental assays can now screen billions of variants in parallel. Yet, a key limitation remains: our current predictive models still struggle to capture the complex physical and dynamical factors that underlie enzyme function. My the- sis addresses this gap by developing an integrated experimental–computational framework for enzyme design that couples large-scale protein library construc- tion with data-driven model development. I design and test extensive libraries of enzyme variants to both optimize catalytic activity and generate training data for next-generation predictors of protein function. Ultimately, this approach ad- vances our ability to connect sequence, structure, and function.