We use cookies to ensure that we give you the best experience on our website. By continuing to browse this repository, you give consent for essential cookies to be used. You can read more about our Privacy and Cookie Policy.

Durham e-Theses
You are in:

On deep generative modelling
methods for protein-protein

LEACH, ADAM (2023) On deep generative modelling
methods for protein-protein
Doctoral thesis, Durham University.

PDF - Accepted Version
Available under License Creative Commons Attribution Non-commercial Share Alike 3.0 (CC BY-NC-SA).



Proteins form the basis for almost all biological processes, identifying the interactions that proteins have with themselves, the environment, and each other are critical to understanding their biological function in an organism, and thus the impact of drugs designed to affect them. Consequently a significant body of research and development focuses on methods to analyse and predict protein structure and interactions. Due to the breadth of possible interactions and the complexity of structures, \textit{in sillico} methods are used to propose models of both interaction and structure that can then be verified experimentally. However the computational complexity of protein interaction means that full physical simulation of these processes requires exceptional computational resources and is often infeasible. Recent advances in deep generative modelling have shown promise in correctly capturing complex conditional distributions. These models derive their basic principles from statistical mechanics and thermodynamic modelling. While the learned functions of these methods are not guaranteed to be physically accurate, they result in a similar sampling process to that suggested by the thermodynamic principles of protein folding and interaction. However, limited research has been applied to extending these models to work over the space of 3D rotation, limiting their applicability to protein models. In this thesis we develop an accelerated sampling strategy for faster sampling of potential docking locations, we then address the rotational diffusion limitation by extending diffusion models to the space of $SO(3)$ and finally present a framework for the use of this rotational diffusion model to rigid docking of proteins.

Item Type:Thesis (Doctoral)
Award:Doctor of Philosophy
Keywords:Protein Docking, Diffusion Models, Machine Learning, Generative Modelling,
Faculty and Department:Faculty of Science > Computer Science, Department of
Thesis Date:2023
Copyright:Copyright of this thesis is held by the author
Deposited On:11 Dec 2023 09:52

Social bookmarking: del.icio.usConnoteaBibSonomyCiteULikeFacebookTwitter