Deep learning-based polygenic risk analysis for Alzheimer's disease prediction.

Area of Science:

• Genetics
• Computational Biology
• Neuroscience

Background:

• Alzheimer's disease (AD) is polygenic, meaning multiple genetic variants contribute to susceptibility.
• Current statistical methods struggle to fully capture complex polygenic risk for AD, limiting prediction accuracy.
• Deep learning offers potential for more accurate AD risk prediction by modeling nonlinear relationships in high-dimensional genomic data.

Purpose of the Study:

• To develop and evaluate deep learning neural network models for modeling Alzheimer's disease (AD) polygenic risk.
• To compare the performance of deep learning models against traditional statistical approaches for AD risk prediction.
• To explore the utility of deep learning in uncovering AD etiology and identifying distinct pathological mechanisms.

Main Methods:

• Construction of neural network models to assess AD polygenic risk.
• Comparison of deep learning models with weighted polygenic risk score and lasso models.
• Robust linear regression to analyze the association between deep learning-derived polygenic risk and AD endophenotypes (plasma biomarkers, cognitive performance).
• Unsupervised clustering applied to neural network hidden layer outputs for individual stratification.

Main Results:

• Deep learning models demonstrated superior performance in modeling AD risk compared to conventional statistical models.
• The polygenic risk scores derived from deep learning facilitated the identification of AD-associated biological pathways.
• Deep learning enabled effective stratification of individuals based on distinct pathological mechanisms.

Conclusions:

• Deep learning methods are highly effective for modeling and classifying genetic risks in Alzheimer's disease (AD).
• These methods enhance our understanding of AD etiology by uncovering disease-specific mechanisms.
• Deep learning holds promise for improving risk prediction and stratification in AD and other complex diseases.