Published in the Alzheimer’s & Dementia: The Journal of the Alzheimer’s Association. Here is a link to the article.
Regenstrief Institute authors: Kun Huang, PhD
Alzheimer’s disease (AD) progression is highly variable across individuals. Some patients exhibit cognitive decline that does not match the severity of their tauopathy, a condition not well studied due to limitations in accessing brain samples in living patients. This study addresses that gap by leveraging matched transcriptomic data from both brain and blood in the ROSMAP cohort to identify atypical AD subtypes. Using a supervised transfer learning approach, researchers classified three AD subgroups—Asymptomatic AD, Low-NFT AD and Typical AD—alongside a normal control group. These subgroups were established based on clinical data related to tauopathy severity and disease progression.
The trained model was applied to blood RNA-seq data from two external cohorts, ADNI and ANMerge, using an optimal transport method to transfer labels. Consistently expressed genes across all three cohorts were then identified to determine subgroup-specific molecular signatures. Additionally, diffusion pseudo-time analysis was employed to uncover gene expression dynamics within each subgroup.
The study successfully identified distinct, consistently expressed genes for each AD subgroup. These gene signatures also demonstrated variation based on sex, age of onset (early vs. late) and progression pattern (sudden vs. gradual), highlighting their relevance to the disease’s heterogeneity.
Overall, this research provides a novel blood-based method for identifying atypical AD subtypes in living patients. The findings enhance understanding of AD pathophysiology and open avenues for early prognosis and personalized treatment strategies, with potential applications in diagnosing and managing related tauopathies.
