A Pathway Based Classification Method for Analyzing Gene Expression for Alzheimer's Disease Diagnosis.

Area of Science:

• Genomics
• Biomarkers
• Neurodegenerative Diseases

Background:

• Blood gene expression may differentiate Alzheimer's disease (AD) and mild cognitive impairment (MCI) from controls.
• Replicability challenges exist at the single gene marker level.
• Pathway-based gene expression analysis may offer a more robust approach.

Purpose of the Study:

• To compare the robustness and predictive performance of pathway-level versus gene-level classification models for AD/MCI.
• To evaluate models using two independent gene expression datasets: AddNeuroMed (ANM) and Dementia Case Registry (DCR).

Main Methods:

• Gene expression data collected using Illumina Human HT-12 Expression BeadChips from blood samples.
• Random forest modeling with recursive feature elimination employed for case/control prediction.
• Age and APOE ɛ4 status included as covariates in all analyses.

Main Results:

• Both gene-level and pathway-level models demonstrated similar performance.
• Model performance was comparable to a model based solely on demographic information.

Conclusions:

• The tested pathway-level approach did not enhance predictive ability over gene-level models in these harmonized datasets.
• Further investigation is needed into alternative pathway scoring methods, particularly those incorporating pathway topology.
• Future research should explore endophenotype-based approaches for improved biomarker discovery.