Identifying underlying patterns in Alzheimer's disease trajectory: a deep learning approach and Mendelian...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biostatistics

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder with variable deterioration rates.
• Understanding AD progression is crucial for effective early interventions and treatment strategies.
• Identifying distinct patterns in AD trajectory can inform personalized medicine approaches.

Purpose of the Study:

• To analyze Alzheimer's disease (AD) progression and identify distinct patterns in its trajectory.
• To develop and validate a deep learning model for predicting time-to-conversion and clustering AD subgroups.
• To investigate the causality between identified patterns and disease progression using Mendelian randomization.

Main Methods:

• Proposed a deep learning model for joint prediction of time-to-conversion and clustering of AD progression patterns.
• Validated the model on the ADNI dataset (1370 participants) and externally on the AIBL dataset (233 participants).
• Employed Mendelian randomization (MR) analysis to infer causality between patterns and time-to-conversion.

Main Results:

• The model identified distinct patterns with significantly different biomarkers and progression rates.
• Identified patterns demonstrated strong predictive ability for conversion from cognitively normal (CN) to mild cognitive impairment (MCI) and to AD dementia (e.g., AUC for 5-year prediction: CN→MCI 0.876, MCI→AD 0.957).
• External validation showed competitive performance (CN→MCI C-Index 0.693, MCI→AD C-Index 0.752); MR analysis suggested a causal link for MCI→AD patterns.

Conclusions:

• The developed deep learning model effectively identifies biologically and clinically meaningful patterns in AD progression.
• These patterns offer valuable insights into disease trajectory, aiding in early detection and personalized treatment strategies.
• The findings support improved clinical trial design and decision-making in Alzheimer's disease management.