Sex-Specific Prediction Models of Alzheimer's Disease: A Gene Expression Analysis

Alzheimer's disease (AD) exhibits sex-specific molecular signatures that may improve diagnostic precision. We aimed to identify and validate male- and female-specific blood and brain gene expression biomarkers for AD prediction. We analyzed four GEO datasets (blood- and brain-derived) using limma and Fisher's meta-analysis to identify sex-specific differentially expressed genes, assessed age associations via linear regression, and constructed 10-fold cross-validated logistic regression models. After performing a meta-analysis, 74 differentially expressed genes were identified in the female cohort and 89 DEGs were screened in the male cohort. ERH and MRPS33 were identified as the most relevant genes in the male cohort, and NDUFA1 and NDUFS5 were screened in the female cohort. The identified genes were downregulated in AD samples compared to controls. Both male-specific and female-specific prediction models achieved an AUC of above 0.7 in two external validation blood-derived datasets as well entorhinal cortex dataset. Paradoxically, qPCR showed significant upregulation of all four genes in the AD group compared to the control group.