Building predictive models of cognitive decline in population at risk of LOAD
by using machine learning tools.
 
   Lead Investigator:    Jaime Ramos-Cejudo
   Institution      :    New York University School of Medicine
   E-Mail           :    jaime.ramoscejudo@nyumc.org
   Proposal ID      :    1099



   Proposal Description:
   No tools are available yet to predict cognitive decline and LOAD in aging population at risk of Alzheimer's disease. While some patients show clear declining trajectories in cognition with time, others recover or show patterns that are difficult to explain. Being a complex and multidimensional disease, from a biomarkers point of view I believe that single features extracted from a subject at risk is (e.g. CSF data) is not consistent enough to be used as a predictor. For this reason big data integration and machine learning tools are key in Alzheimer's research and that is why recent big data approaches have identified vascular factors as one of the earliest events in LOAD, even before amyloid deposition, structural or metabolism alterations are evident in images (Iturria-Medina Y, 2016).    I work as a research scientist at New York University in close collaboration with Silvia Fossati (director of Biomarker's core), Mony de Leon (director of the Center for Brain Health) and Nathanael Fillmore (Associate Director for Machine Learning and Predictive Analytics at MAVERIC and the Department of Biostatistics and Computational Biology at Dana-Farber Cancer Institute). My work is focused on building predictive models of dementia in population at risk by using machine learning (ML) tools. Because one of the main limitations in ML is the availability of data, I believe using the NACC database would proper the present project. Importantly, our team has access to the MAVERIC (VA database) including the million veteran's program at the VA and we believe using both cohorts as a cross validation will help validate our working hypothesis.    Using longitudinal data from the NACC, I propose to integrate genomic variants in healthy subjects at risk and prepare a report that relates these counts to future cognitive status, neuropathology, CSF biomarkers, demographics, and other data elements of interest. With a focus on cardiovascular risk factors, I will build a predictive model o