Author ORCID Identifier
Year of Publication
Doctor of Philosophy (PhD)
Dr. Chang-Guo Zhan
Computational modeling is an invaluable tool in the drug discovery process either for small ligand or protein therapeutics. The widespread availability of protein X-Ray Crystal and Cryo-Electron Microscopy (Cryo-EM) structures has allowed for more accurate molecular dynamics (MD) simulations that are not reliant on methods such as homology modeling, which may produce structures that require significant computational time to demonstrate their stability. In this thesis we describe several novel methodologies for the computationally efficient modeling of protein/ligand and protein/protein complexes that may be employed within both large-scale virtual screenings and lead compound optimization. These methodologies may also be utilized in tandem with linear regression and neural network-based models to predict the binding affinity changes of both ligands with their respective proteins as well as with protein/protein interactions such as the SARS-CoV-2 Spike protein binding with human angiotensin type 2 receptor (ACE2) or antibody therapeutics. These methodologies are both computationally efficient and generalizable for a large selection of ligand and protein targets. This thesis consists of four chapters, Chapter One serving as an introduction to the current state of the art concerning the use of computational methodologies within the drug discovery space. Chapter Two will focus on two main projects focusing on structure-based drug design. The first subchapter of chapter 2 will focus on efforts concerning the discovery of new butyrylcholinesterase inhibitors, as well as the use of pharmacophore-based filtering to pare down large chemical libraries such as the NCI developmental therapeutics program (DTP) library. The second subchapter will focus on the methodologies employed to elucidate the binding mode of several long chain endocannabinoids and the creation of a linear regression model to predict the binding affinity and selectivity of these compounds towards Cannabinoid Receptors 1 and 2. Chapter Three will focus on four projects undertaken to study the interactions between the SARS-CoV-2 Spike protein and several protein targets. The first and second subchapters will focus on the methodologies employed to create a linear regression model to predict the binding affinity of the SARS-CoV-2 spike protein with human ACE2. The third subchapter will focus on the development of a multi-layer perceptron model artificial intelligence (AI) to predict how mutations of the SARS-CoV-2 virus would affect the efficacy of both human immune response antibodies as well as antibodies being developed as therapeutics. The fourth subchapter will focus on applying our methodology established within subchapters 1 and 2 to take a long-term approach to identifying new SARS-CoV-2 variants that would display increased affinity towards the ACE2 protein. Chapter 4 will summarize our findings and potential future experiments.
Digital Object Identifier (DOI)
These studies were funded by the following sources:
University of Kentucky College of Pharmacy: Pharmaceutical Sciences Excellence in Graduate Achievement Fellowship Award (2021-2022)
National Institute of Drug Abuse: T32 DA016176 (2018-2021)
National Institute of Health: UH2/UH3 DA041115 (2017-2022)
National Institute of Health: U18 DA052319 (2017-2022)
National Institute of Health: P20 GM130456 (2017-2022)
Williams, Alexander Hamilton, "DEVELOPMENT OF ACCURATE AND EFFICIENT COMPUTATIONAL METHODOLOGIES FOR PREDICTING PROTEIN-LIGAND AND PROTEIN-PROTEIN BINDING FREE ENERGIES" (2022). Theses and Dissertations--Pharmacy. 141.
Available for download on Tuesday, September 24, 2024