Alzheimer's disease (AD) is the leading cause of dementia in the United States but has no disease-modifying treatments. AD is also a substantially genetically based disease but the heritability of AD declines with older ages-at-onset suggesting increased environmental contributions at older age. We propose to study the AD epigenome since it is influenced by both genetic and environmental factors. Here, we will study DNA methylation at the fifth position of cytosine, which could modulate transcriptional activity. We propose an innovative machine learning computational approach to discover new AD-associated methylation sites beyond what can be interrogated by array-based approaches. We propose to use machine learning classification models trained on existing AD epigenetic studies and publicly available, multi-omics data to generate an AD-specific ?risk score? for each methylation site in the genome by leveraging the tremendous investment in AD research, particularly the Accelerating Medicines Partnership-Alzheimer's Disease (AMP-AD), and the Molecular Mechanisms of the Vascular Etiology of Alzheimer's Disease (M2OVE-AD). We predict that high scoring methylation sites will associate with AD, and we will test this prediction with a 2-stage (discovery/replication) study using a unique dataset of human postmortem brains from the Rush Memory and Aging Project (MAP) and Religious Orders Study (ROS) and an autopsy dataset from Emory University's Alzheimer's Disease Research Center (ADRC). Simultaneously, we propose functional studies in model systems to determine whether the selective identified loci could influence AD pathogenesis.
In Aim 1, we propose to build machine learning models and using the models to obtain, AD-specific ?risk scores? for each of the 28 million methylation sites in the genome In Aim 2, we propose to test our models using a unique dataset of 1695 human postmortem brains from the Rush Memory and Aging Project (MAP) and Religious Orders Study (ROS) and 432 postmortem human brains from the Emory ADRC. Both ROS and MAP are prospective longitudinal studies annually collect data cognitive, mental, and physical health, and has the requisite available tissue and data from genomic, transcriptomic, and proteomic studies of the dorsolateral prefrontal cortex (dPFC). We propose a 2-stage genetic study to validate whether predicted methylation sites are associated with AD.
In Aim 3, we will examine the functional consequences of top AD-associated ?risk scores? in relevant model systems, including fly, mammalian cell culture and mouse models of AD. This project can potentially identify important molecular contributors of AD that might not be apparent through other approaches, leading to new insights into mechanisms and treatment targets for AD and thereby have an important and sustained impact on public health.
Alzheimer's disease (AD) affects 5.4 million people in the U.S and is the 6th leading cause of death. AD is a substantially genetically based disease, and our proposal aims to find new genetic causes of AD by employing cutting-edge computational approaches, followed by experimental validation. Discovery of new genetic causes of AD will: 1) shed light on new biological mechanisms, 2) facilitate development of new diagnostic tests, and 3) provide new targets for AD treatment.
