Open Access

David Patsch

• In their idealized forms, enzymes can facilitate complex reactions with extreme specificity and selectivity. Additionally, in this imaginative form, they only require mild reaction conditions, resulting in low energy consumption, and they are biodegradable, efficient, reusable, and sustainable. Unfortunately, this idealized form often deviates significantly from reality, where enzymes are more likely to be associated with marginal stability and low reaction rates, leaving them less than desirable for many industrial applications. As such, if we could master the process of engineering the configuration of a protein towards a given task, the implications could be staggering. This thesis aims to contribute to the process of protein engineering, mainly how computational tools can be used to make the protein engineering process more efficient and accessible. Article I explores the current state of the art in machine learning-guided directed evolution and serves as a foundation for Article II, which is a concrete application of these techniques to an engineering campaign. Despite successfully improving overall activity and selectivity, we also observe limitations and constraints within the methodology. Article III then delves into these drawbacks and attempts to lay the foundation for a more generalizable and, more importantly, efficient engineering workflow, balancing the strengths and weaknesses of computational techniques with advances in gene synthesis. We then validated this novel pipeline in Article IV, where we show the potential of this methodology. Article V describes a more standard protein engineering campaign on squalene-hopene cyclases for potentially interesting products in the flavor and fragrance industry. Lastly, Article VI outlines a PyMol plugin for molecular docking.