Theses and Dissertations

ORCID

https://orcid.org/0009-0008-5133-0006

Advisor

Yarahmadian, Shantia

Committee Member

Qian, Chuanxi

Committee Member

Sivaraman, Vaidyanathan

Committee Member

Barlow, Jonathan

Date of Degree

5-15-2026

Original embargo terms

Visible MSU Only 6 months

Document Type

Dissertation - Campus Access Only

Major

Computational Engineering

Degree Name

Doctor of Philosophy (Ph.D.)

College

James Worth Bagley College of Engineering

Department

Computational Engineering Program

Abstract

Alzheimer’s disease (AD) is a highly complex, yet very pressing public health problem that benefits from having both predictive models and mechanistic insights. Predictive models are needed to better identify patients at risk for Alzheimer’s disease earlier, and mechanistic models are needed to better understand the biological mechanisms behind AD progression. To this end, this thesis takes a two-fold approach. In the first half of this thesis, a series of classical and modern machine learning methods are used to predict AD based on clinical and demographic data collected from the National Alzheimer’s Coordinating Center (NACC). In the second half of the thesis, a mechanistic model was created that describes some of the important interactions between metal ions, amyloid-beta, reactive oxygen species (ROS), antioxidants, and tau. The specific machine learning algorithms used in the first half include ensemble methods, neural networks, and the attention-based TabNet model. Models were created using hundreds of combinations of hyperparameters and feature selectors, and the performance of these models was benchmarked against each other. This model, using carefully structured clinical data, was able to predict cognitive status and mild cognitive impairment with accuracies and discrimination metrics in the range of the best available published estimates. In the second half of the thesis, a mechanistic model was created. The model contains 24 coupled ordinary differential equations that model metal ion dyshomeostasis, amyloid-beta aggregation, ROS generation, antioxidants, and tau phosphorylation. The interactions in the model lead to nonlinear and complex behaviors that can be seen analytically and numerically, which may help to better understand AD progression. The model is able to rationalize how factors like oxidative stress and metal-ion competition can increase the speed of neuronal damage. This mechanistic model can also be used to better explore possible therapeutic avenues by targeting oxidative and metabolic processes.

Download

Share

COinS