• Select the area you would like to search.
  • ACTIVE INVESTIGATIONS Search for current projects using the investigator's name, institution, or keywords.
  • EXPERTS KNOWLEDGE BASE Enter keywords to search a list of questions and answers received and processed by the ADNI team.
  • ADNI PDFS Search any ADNI publication pdf by author, keyword, or PMID. Use an asterisk only to view all pdfs.
Principal Investigator  
Principal Investigator's Name: Yuhan Cui
Institution: University of Pennsylvania
Department: Radiology
Proposed Analysis: I am participating in an AI4AD project which aims to leverage the power of AI to extract, prioritize and synthesize the features from extensive genomic, biomarker, and cognitive data, and lead to more precise characterization and early detection of AD. Aim 1. Discover new genetic signatures for AD through genome tiling. a. We will use data “tiling” to parse genomic sequences into small segments (“tiles”). This lossless representational strategy is particularly suitable for machine learning. b. We will compute personalized genomic risk and hazard scores for AD, using mathematically novel feature discovery approaches. c. We will test predictive features in non-European participants. Hypothesis: Our methods will yield personalized AD risk metrics, which will improve risk prediction in independent samples, compared to traditional genotyping-based risk score methods.
Aim 2. Use novel AI methods to identify neuroimaging markers of AD risk. a. We will develop advanced image harmonization and AI methods to 1) integrate diverse imaging datasets; 2) dissect patterns of neuropathology (amyloid and tau), neurodegeneration, and cerebrovascular disease and 3) construct imaging endophenotypes of AD and its preclinical stages. b. We will use AI to cluster and determine disease subtypes from clinical and imaging data and identify multivariate genomic predictors of biomarker subtype and rate of progression. H: AI-derived multivariate imaging patterns will serve as better phenotypes for discovery of genomic features than case/control status or simple region of interest (ROI) variables.
Aim 3. Identify new genetic effects on AD imaging and cognitive markers and build a predictive model of cognitive decline and AD progression using multimodal imaging, whole genomes and deep learning. a. Determine the relative predictive value of whole-genome features (including combinations of low frequency variants) versus GWAS-derived polygenic risks on clinical progression, indexed by decline on harmonized cognitive tests. b. Identify genetic factors affecting AD imaging markers, and integrate multimodal imaging, whole genomes, and demographics via deep learning to build a predictive model of cognitive decline and AD progression. H: Machine learning and deep learning on whole genomes and multimodal imaging markers will detect new AD genes and predict subtle changes in cognitive performance during the earliest stages of disease. These methods have distinct advantages in power over classical statistical methods or any single imaging data modality.
Aim 4. Target prioritization and drug re-purposing. We will prioritize genomic variants from Aims 1 and 3 by integrating with the identified results from Aims 1-3 and determine network-based signatures for repositioning existing drugs in AD therapeutics. We will also provide a platform for bridging academia to the pharmaceutical industry.
Aim 5. Disseminate data and AI methods to the AD research community. We will provide curated NGS, Imaging, and Cognitive Data for ADSP via NIAGADS, as well as prioritized and annotated lists of novel therapeutic targets. We will host training workshops to showcase our cloud-based informatics tools that link WGS, imaging, and cognitive phenotypes along the AD continuum.
Additional Investigators