Development and validation of a 13-gene signature associated with immune function for the detection of Alzheimer's disease

Abstract

Current knowledge of Alzheimer's disease (AD) etiology and effective therapy remains limited. Thus, the identification of biomarkers is crucial to improve the detection and treatment of patients with AD. Using robust rank aggregation method to analyze the microarray data from Gene Expression Omnibus database, we identified 1138 differentially expressed genes in AD. We then explored 13 hub genes by weighted gene co-expression network analysis, least absolute shrinkage, and selection operator, and logistic regression in the training dataset. The detection model, which composed of CD163, CDC42SE1, CECR6, CSF1R, CYP27A1, EIF4E3, H2AFJ, IFIT2, IL10RA, KIAA1324, PSTPIP1, SLA, and TBC1D2 genes, along with APOE gene, showed that the area under the curve for detecting AD was 0.821 (95% confidence interval [CI] = 0.782–0.861) and the model was validated in ADNI dataset (area under the curve = 0.776; 95%CI = 0.686–0.865). Notably, the 13 genes in the model were highly enriched in immune function. These findings have implications for the detection and therapeutic target of AD.