MAPPING ALZHEIMER'S DISEASE PSEUDO-PROGRESSION WITH MULTIMODAL BIOMARKER TRAJECTORY EMBEDDINGS

Area of Science:

• Neuroscience
• Computational Biology
• Biomedical Data Science

Background:

• Alzheimer's Disease (AD) is a progressive neurodegenerative disorder impacting millions globally.
• Accurate mapping of AD progression is vital for early detection, intervention, and treatment development, but remains challenging.
• Understanding heterogeneous disease pathways is key to personalized medicine approaches.

Purpose of the Study:

• To develop and validate a novel analytical pipeline for modeling heterogeneous Alzheimer's Disease progression pathways.
• To leverage machine learning techniques from single-cell transcriptomics for analyzing longitudinal neuroimaging and biomarker data.
• To provide a more precise method for disease modeling and early diagnosis of Alzheimer's Disease.

Main Methods:

• Proposed an analytical pipeline integrating PHATE and Slingshot, machine learning methods from single-cell transcriptomics.
• Projected multimodal biomarker trajectories from longitudinal data into a low-dimensional space to generate pseudotime estimates.
• Applied the pipeline to the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset for aligning individual disease trajectories.

Main Results:

• Pseudotime estimates revealed distinct patterns of Alzheimer's Disease evolution and biomarker changes over time.
• Demonstrated the ability to align longitudinal data across individuals at various disease stages.
• Identified heterogeneous temporal dynamics in disease progression.

Conclusions:

• The developed approach offers a novel way to understand the temporal dynamics of Alzheimer's Disease.
• The findings highlight the potential for more precise disease modeling and earlier diagnosis in neurodegenerative diseases.
• This method provides deeper insights into individual patient trajectories and biomarker evolution in AD.