Soluble amyloid-beta isoforms predict downstream Alzheimer's disease pathology

Area of Science:

• Neuroscience
• Biochemistry
• Computational Biology

Background:

• Soluble amyloid-beta (Aβ) levels in cerebrospinal fluid (CSF) are early indicators of Alzheimer's disease (AD).
• Predictive value of Aβ isoforms for downstream tau pathology (T+) and neurodegeneration (N+) in cognitively unimpaired (CU) individuals remains unclear.

Purpose of the Study:

• To investigate if combined soluble Aβ isoforms can predict T+ and N+ positivity in CU individuals.
• To explore biological processes altered in individuals with predicted T+ and N+ using proteomics.

Main Methods:

• Machine learning (ML) models were trained using CSF Aβ_1-38, Aβ_1-40, and Aβ_1-42 measurements from 318 CU individuals.
• 2046 predictive ML models were built using nested cross-validation.
• Proteomics data analyzed for functional enrichment of differentially expressed proteins (DEPs) in misclassified individuals.

Main Results:

• Aβ isoforms accurately predicted T+ (AUC=0.929) and N+ (AUC=0.936).
• Proteomics identified 17 DEPs in misclassified individuals.
• Enrichment analysis revealed altered myelinization, glucose metabolism, amino acid biosynthesis, and cell adhesion pathways.

Conclusions:

• Refined ML analysis demonstrates high predictive accuracy of Aβ isoforms for T+ and N+.
• CSF proteomics identified potential biomarkers for improving T+ and N+ prediction in early AD stages.