Mapping amyloid, tau, and neurodegeneration markers with neuropsychiatric phenotypes and genotypes in Alzheimer's disease continuum using machine learning

Washington University, Saint Louis MO

Investigators

Abstract

Project Abstract / Summary Alzheimer’s disease (AD) is the leading cause of dementia and has significant medical and public health concerns. Despite extensive efforts, the exact relationships between in-vivo amyloid-beta (A), tau (T), and neurodegeneration (N), neuropsychiatric symptoms (NPS), vascular risk factors, and genetic risks for NPS in AD are poorly understood. Patients with AD pathology exhibit neurobiological and NPS heterogeneity characteristics. The central goals have been to better understand AD’s underlying neurobiological heterogeneity mechanisms and improve outcomes. This study plans to leverage in-vivo ATN markers from amyloid PET, tau PET, and MRI to investigate the multivariate relationships between these markers and neuropsychiatric phenotypes and genotypes. This project will utilize large-sample data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), Washington University’s Knight Alzheimer’s Disease Research Center (Knight ADRC), Baltimore Longitudinal Study of Aging (BLSA), and National Alzheimer’s Coordinating Center (NACC) cohorts. This study will be the first to systematically investigate the multivariate relationships between ATN biomarkers, NPS, vascular risks, and genetic risks for NPS via machine learning predictive modeling and heterogeneity analytics in AD. In all cohorts, the project will consistently and optimally quantify PET as regional distribution volume ratio (DVR) and standard uptake value ratio (SUVR), as well as MRI as regional volume measures. We will integrate these cohorts and generate harmonized data resources for machine learning method developments and their applications. Aim 1 will develop and optimize machine learning/artificial intelligence methods in the context of regional ATN biomarkers and identify their multivariate predictive relationships with global NPS and highly prevalent domain-specific NPS features. This aim will identify dominant ATN regional hubs collectively involved in NPS processes. Aim 2 will identify the regional heterogeneity of ATN biomarkers via semi-supervised machine learning and reproducibility analytical methods. This aim will correlate the ATN profiles and NPS prevalence and severity between identified subgroups of patients or controls and each subgroup of patients to test the hypothesis of whether ATN and NPS phenotypes differ between subgroups of patients. Aim 3 will examine the relationships of individualized ATN heterogeneity signatures with baseline NPS and longitudinal trajectories in NPS. Aim 4 will study vascular risk and genetics for NPS to test whether ATN and NPS heterogeneity signatures have differential genetic etiologies. In summary, this innovative project will provide critical information on neuropsychiatric aspects of AD mechanisms and significantly contribute to precision medicine in the diagnosis and treatment of AD patients with NPS.