Multi-task deep autoencoder to predict Alzheimer's disease progression using temporal DNA methylation data in...

Area of Science:

• Computational Biology
• Neuroscience
• Genomics

Background:

• Alzheimer's disease (AD) diagnosis relies on invasive and costly methods like brain imaging and cerebrospinal fluid analysis.
• Peripheral tissue biomarkers offer a non-invasive alternative, but their utility for AD progression prediction using DNA methylation is underexplored.
• Predicting AD progression from complex, high-dimensional longitudinal DNA methylation data presents significant modeling challenges.

Purpose of the Study:

• To develop and evaluate novel multi-task deep autoencoder models for predicting Alzheimer's disease progression.
• To leverage peripheral blood DNA methylation data for non-invasive AD diagnostics.
• To assess the efficacy of deep learning approaches in handling high-dimensional longitudinal genomic data.

Main Methods:

• Development of two multi-task deep autoencoders: convolutional autoencoder and long short-term memory autoencoder.
• Jointly minimizing reconstruction error and maximizing prediction accuracy for feature representation learning.
• Benchmarking against state-of-the-art machine learning methods using longitudinal DNA methylation data from the Alzheimer's Disease Neuroimaging Initiative (ADNI).

Main Results:

• The proposed multi-task deep autoencoders significantly outperformed existing machine learning approaches in predicting AD progression.
• The models demonstrated high accuracy in reconstructing temporal DNA methylation profiles.
• Accurate prediction of AD progression was achieved using historical DNA methylation data, with further improvement when incorporating all temporal data.

Conclusions:

• Multi-task deep autoencoders provide an effective and non-invasive method for predicting Alzheimer's disease progression using peripheral blood DNA methylation.
• These models offer a promising advancement over traditional diagnostic methods.
• The developed approach facilitates accurate AD progression prediction from longitudinal genomic data.