Computer-directed evolution of enzyme selectivity for bioproduct synthesis

Abstract

Global environmental issues and sustainable development call for new technologies for fine chemical synthesis and waste valorization. Biocatalysis has attracted great attention as the alternative to the traditional organic synthesis. To navigate the vast sequence space to identify those proteins with admirable biocatalytic functions is challenging. The expansion of the 3D structures aided by the recent development of deep-learning based structure prediction methods such as AlphaFold2 enabled structure-based design which are further reinforced by different computational simulations or multiscale calculations combining molecular mechanics and quantum mechanics. However, the structure-based approaches are not suitable for large-scale screening of the potential biocatalysts from large sequence space. Machine learning techniques can be applied to comprehensive datasets (e.g., protein sequences, 3D structural, functional annotations, and enzymatic activities, etc.) to enable more efficient and accurate predictive modeling. Chapter 1 covers the theoretical background and computational methods, including traditional structure-based rational design approaches and the fundamental concepts of the state-of-the-art machine learning techniques. Chapter 2 presents a case study demonstrating rational design strategies where key catalytic motifs identification and co-evolution analysis were employed to guide the selective synthesis of bicyclogermacrene enantiomers. Chapter 3 presents a case study on the selective synthesis of farnesene diastereomers guided by computational simulations, combining structural modelling, molecular docking and molecular dynamics simulations. Chapter 4 presents theoretical study on the origin of enantioselectivity in alcohol dehydrogenase using QM/MM methods. Chapter 5 presents how machine learning algorithms were utilized in overcoming the limitations of rational enzyme design in sequence landscape exploration, we were able to obtain enantioselective products by large-scale predictions and evaluation of mutations. Chapter 6 concludes this thesis and summarizes the strategy of integrating computational simulations and machine learning techniques in enzyme design.

Thesis embargoed until 31st July 2030

Date of Award	Jul 2025
Original language	English
Awarding Institution	Queen's University Belfast
Sponsors	QUB-China Scholarship Council
Supervisor	Meilan Huang (Supervisor) & Peijun Hu (Supervisor)