Basic Science and Pathogenesis.

Area of Science:

• Neuroscience
• Genomics
• Biomarkers

Background:

• Alzheimer's disease (AD) presents with diverse clinical and pathological features, indicating potential patient subtypes.
• Molecular profiling is crucial for understanding AD heterogeneity, but current methods often overlook the importance of specific feature subsets.
• Identifying distinct AD subgroups is essential for targeted therapies and personalized medicine.

Purpose of the Study:

• To develop and validate a novel framework, AutoSGI, for identifying informative molecular feature subsets to characterize Alzheimer's disease patient subgroups.
• To leverage multi-omics data for a more nuanced understanding of AD heterogeneity and progression.

Main Methods:

• Developed AutoSGI, a framework that identifies informative feature subsets using pathway annotations or feature clustering for subgroup analysis.
• Employed the SGI (Subgroup Identification) toolbox for hierarchical sample clustering and clinical outcome evaluation at each branch point.
• Applied statistical adjustments to account for multiple feature subset analyses.

Main Results:

• Demonstrated AutoSGI's utility in two Alzheimer's disease case studies using metabolic and lipidomic data.
• In the first case, AutoSGI stratified postmortem brain metabolomics data by disease stage, revealing subgroups with significant differences in Braak and CERAD scores.
• In the second case, AutoSGI identified blood lipidomics-defined subgroups with varying disease progression, correlated with CSF tau and ADAS-Cog-13 scores.

Conclusions:

• AutoSGI effectively leverages multi-scale feature subsets for robust subgroup identification in Alzheimer's disease.
• The framework provides valuable insights into AD heterogeneity and progression, paving the way for more precise patient stratification.
• This approach enhances the understanding of molecular underpinnings of AD subtypes.