Longitudinal Genotype-Phenotype Association Study via Temporal Structure Auto-Learning Predictive Model.

Area of Science:

• Neuroscience
• Genetics
• Biomedical Engineering

Background:

• Imaging genetics combines neuroimaging and genetic data to study complex brain disorders like Alzheimer's Disease (AD).
• Understanding the relationship between Single Nucleotide Polymorphisms (SNPs) and longitudinal neuroimaging phenotypes is crucial for AD research.
• Existing machine learning models often struggle with automatically learning interrelations among longitudinal prediction tasks.

Purpose of the Study:

• To propose a novel temporal structure auto-learning model for uncovering longitudinal genotype-phenotype interrelations.
• To enhance phenotype prediction by utilizing automatically learned longitudinal structures.
• To investigate the genetic basis and biological mechanisms underlying AD progression.

Main Methods:

• Developed a temporal structure auto-learning model to automatically identify interrelations among longitudinal genotype-phenotype tasks.
• Applied the model to the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort.
• Utilized 3,123 SNPs and two types of biomarkers: Voxel-Based Morphometry (VBM) and FreeSurfer.

Main Results:

• The proposed model demonstrated superior performance compared to existing methods in longitudinal phenotype prediction.
• Empirical results validated the model's ability to uncover meaningful genotype-phenotype interrelations.
• Top selected SNPs showed consistency with available literature, supporting the model's biological relevance.

Conclusions:

• The novel temporal structure auto-learning model effectively captures longitudinal genotype-phenotype interrelations for enhanced AD prediction.
• This approach offers a promising tool for advancing imaging genetics research in neurodegenerative diseases.
• The findings contribute to a deeper understanding of the genetic underpinnings of brain structure and function changes in AD.