Date of Award


Document Type

Thesis campus only


Computer Science

First Advisor

Matthew Hibbs

Second Advisor

Seth Fogarty

Third Advisor

Kwan Cheng


The phenomenon of phase separation in mixed lipid systems is one of significant biological importance in cell membrane construction and has been extensively studied both experimentally and theoretically. Understanding the behavior of the phase-separated lipid nanodomains through simulation data can help researchers better understand the mechanisms of these lipid phases/domains, which can in turn provide a useful approach to study membrane damage through lipid phases disruptions, and ultimately, guide potential drug development associated with membrane-targeted diseases, such as the pathogenesis of Alzheimer's. In this project, we tested two unsupervised learning algorithms to determine lipid phases: Non-negative Matrix Factorization (NMFk) and a customized form of KMeans. We then compared these learning algorithms with a brute-force neighbor-searching method. The results of lipid phase classification from these three different methods are compared and contrasted by measuring the thickness and area-per-lipid of those groups. The project provides three different pipelines to study the evolution of lipid phases and segregation, and characterizes the membrane damage induced by the appearance of amyloidogenic proteins.