Advancing Alzheimer's Diagnosis: MRI-Based Predictive Modeling of Alzheimer’s Disease Molecular Subtypes

PROJECT SUMMARY Alzheimer's Disease (AD) is a progressive neurodegenerative disorder that impacts cognition and memory, imposing a substantial burden on the aging population. While individuals generally need to display core clinical and biological features to meet diagnostic criteria for AD, there is substantial variability among patients

regarding onset and the course of the disease. This poses significant challenges to early diagnosis of AD and impedes the development of effective therapeutic strategies. Molecular subtyping of AD is of high importance due to its potential to identify disease mechanisms and new therapeutic targets. Recent studies have uncovered

evidence indicating that molecular subgroups of AD display distinct disease characteristics aligning with key clinical features. The identification of distinct AD subtypes has spurred research regarding their potential to be identified using non-invasive neuroimaging techniques, which holds promise for early detection. However,

progress towards this overarching goal has been impeded by studies lacking comprehensive data collected from the same individuals. The goal of this proposal is to enhance our understanding of the biological signatures that distinguish subgroups of patients with AD. We will use state-of-the-art computational approaches to jointly

analyze multiple levels of data collected from the same individuals, encompassing a genetic, epigenetic, and neuroimaging information, constituting an unprecedented cohort in which to investigate AD subtypes. The proposed investigation will address the following gaps in the current scientific knowledge:

1) We will determine whether AD subtypes can be identified solely using neuroimaging-derived features. To address this question, we will use using state-of-the-art machine learning algorithms to predict molecular subtype of AD from magnetic resonance imaging (MRI)-derived phenotypes. 2) AD is a multifaceted disorder in which epigenetic regulatory mechanisms play a crucial role. However,

prior studies on AD subtypes demonstrated limited or no integration of epigenetic information, thereby restricting knowledge of variability between AD subtypes. To address this critical knowledge gap, we will integrate epigenetic profiles previously associated with AD, encompassing expression levels of small non-coding RNAs

(miRNAs) and DNA methylation (DNAm) markers, to determine their contribution to AD subtypes. This multi- omic approach has the potential to uncover new molecular signatures of distinct AD subgroups.