Basic Science and Pathogenesis.

Area of Science:

• Neuroscience
• Genomics
• Biotechnology

Background:

• Alzheimer's disease (AD) presents complex, heterogeneous patient profiles, hindering diagnosis and personalized treatment development.
• Omics technologies (transcriptomics, metabolomics, proteomics) provide comprehensive molecular data for understanding AD.
• Integrating multi-omics data can reveal disease mechanisms and identify patient subgroups for personalized medicine.

Purpose of the Study:

• To propose a multi-omics data integration and disease subtyping approach for Alzheimer's disease.
• To leverage data-driven multi-scale network structures from the AD Atlas for enhanced analysis.
• To enable personalized medicine by identifying distinct Alzheimer's disease subtypes.

Main Methods:

• Extracted network module features from transcriptomics, proteomics, and metabolomics data using WGCNA and PCA.
• Applied deep graph representation learning and unsupervised clustering to identify 25 multi-omics clusters.
• Utilized an autoencoder model for integrated multi-omics expression profiling and disease subtyping.

Main Results:

• The autoencoder effectively learned low-dimensional multi-omics representations.
• Identified molecular subtypes of AD not apparent through traditional analyses.
• Demonstrated sensitivity of the brain multi-ome to AD neuropathology, clinical status, and other variables.

Conclusions:

• The multi-omics integration methodology offers a robust framework for AD patient stratification.
• Integration of multi-omics data and the AD Atlas network identified patient groups with shared clinical parameters.
• Advanced understanding of Alzheimer's disease molecular heterogeneity and subtype-specific manifestations.