Tuesday, April 1, 2025
4/1/2025
Autopsy-informed integrated clinical and imaging models for prediction of non-AD co-pathologies in AD - PROJECT SUMMARY The overall goal of this project is to use gold-standard autopsy-confirmed datasets to build computational models with which widely available neuroimaging and cognitive assessments can be used to predict presence of non- Alzheimer’s disease neuropathological changes (non-ADNC) in individuals with ADNC. Far from a rare problem, estimates of rates of neurodegenerative co-pathology in patients with ADNC range from 50% to 100%. These co-pathologies have been shown to influence the clinical progression of AD symptoms. Yet, biomarkers remain mostly unavailable for non-ADNC, including Lewy body disease (LBD), transactive response DNA-binding protein 43 (TDP-43) proteinopathy, and cerebral amyloid angiopathy (CAA). Without biomarkers to recognize the presence and severity of these co-pathologies in individuals with AD, every clinical trial designed to show the effectiveness of an AD treatment will continue to systematically underestimate the maximum therapeutic effect of the drug. Better biomarkers to detect co-pathologies in ADNC cases are an essential precursor to optimally treating AD. We will pursue the following specific aims: Aim 1: Develop precise machine-learning based computational models that predict the presence of three major forms of non-ADNCs (LBD, TDP-43, CAA) in individuals with ADNC using a single timepoint of in vivo multimodal neuroimaging and clinical data on a Discovery Cohort of N=530 from the University of California San Francisco (UCSF) ADRC, the Alzheimer’s Disease Neuroimaging Initiative (ADNI), and Religious Orders Study / Rush Memory and Aging Project (ROS/MAP) by leveraging effort of the Alzheimer's Disease Sequencing Project Phenotype Harmonization Consortium (ADSP-PHC). Aim 2: Increase model precision by adding clinical progression features from Discovery Cohort longitudinal neuroimaging and clinical assessments to predict non-ADNC in individuals with ADNC. The models will be optimized for real-world implementation by clinical researchers via performance testing against a subset of NACC participants (N=365) and a non-autopsy cohort of 2000+ ADNI and DIAN participants (Validation Cohort). We will examine to what extent clinical stage, timing of data collection, and missing data affect the accuracy of model predictions, and create and disseminate tools based on these models for all researchers to use. These models will rely heavily on the macrostructural brain changes seen on MRI, and will integrate multimodal MRI sequences (T1, FLAIR) and FDG-PET, cross-sectional and longitudinal timepoints, and clinical data, including neuropsychological testing and neuropsychiatry. Because biomarker development requires validation against ground-truth data, key to this proposal is the harmonization of three highly detailed and comprehensive autopsy-confirmed cohorts (Discovery Cohort) with enriched neuropathology, neuroimaging, and clinical data. At the conclusion of this study, we will have for the first time established clinically translatable computational models to predict the presence of TDP-43, LBD, and CAA co-pathologies at the single case level in individuals with ADNC from imaging and clinical data.
HHS’ Tracking Accountability in Government Grants System (TAGGS) website provides access to detailed descriptions of grants, loans, aggregated direct payments and other types of financial assistance awarded by the HHS Operating Divisions (OPDIVs) and the Office of the Secretary Staff Divisions (STAFFDIVs).