On The Physiological And Immunological Effects Of Protein Glycation And Its Prediction Using A Novel 3D Convolutional Neural Network
As a non-enzymatic reaction, protein glycation has a myriad of factors that govern the ability for a specific amino-group to react and if so, the rate of reaction. The first chapter of this dissertation is an overview of the reaction and these factors. Additionally, while in the modern, industrialized world glycation is associated with pathophysiology, one possible theoretical explanation for the physiological role in evolutionary history is put forth.One consequence of glycation being a non-enzymatic reaction is that no specific protein is required to catalyze the reaction and it therefore occurs ubiquitously within the body. This is reflected in the diverse pathophysiological effects elicited by elevations in protein glycation. Chapter 2 explores what this looks like in regards to the role protein glycation may play in COVID-19 complications experienced by diabetic patients. This chapter also examines inconsistent predictions made between existing protein glycation prediction models.In Chapter 3, a novel protein glycation model, SweetSMILE, is proposed. SweetSMILE is a 3D Convolutional Neural Network classifier that predicts the likelihood of lysine residues within a protein undergoing glycation utilizing simplified molecular-input line-entry (SMILE) representation of amino acid sequences converted into graphical representations of their chemical structures as input. The model demonstrates higher predictive quality than existing models, particularly for smaller datasets, and the novel input architecture permits it to be expanded to predicting glycation of amino acids besides lysine, as well as other types of biomolecules, when sufficiently large datasets of those instances of glycation become available.A brief summary of the key findings of the dissertation is provided in Chapter 4. Additionally, this chapter discusses some limitations of the current research and proposes solutions for how to address and follow-up on these in the future.
Read
- In Collections
-
Electronic Theses & Dissertations
- Copyright Status
- Attribution-NonCommercial-ShareAlike 4.0 International
- Material Type
-
Theses
- Authors
-
Turkette, Thomas
- Thesis Advisors
-
Root-Bernstein, Robert
- Committee Members
-
Wehrwein, Erica
Dillon, Patrick
Chan, Christina
Leishman, Derek
- Date
- 2024
- Subjects
-
Physiology
- Program of Study
-
Physiology - Doctor of Philosophy
- Degree Level
-
Doctoral
- Language
-
English
- Pages
- 143 pages
- Permalink
- https://doi.org/doi:10.25335/nrd5-1m13