Prediction of Alzheimer's Disease Using Patterns of Methylation Levels in Key Immunologic-Related Genes.

Abstract

DNA methylation is expected to become a kind of new diagnosis and treatment method of Alzheimer's disease (AD). Neuroinflammation- and immune-related pathways represent one of the major genetic risk factors for AD. We aimed to investigate DNA methylation levels of 7 key immunologic-related genes in peripheral blood and appraise their applicability in the diagnosis of AD. Methylation levels were obtained from 222 participants (101 AD, 72 MCI, 49 non-cognitively impaired controls). Logistic regression models for diagnosing AD were established after least absolute shrinkage and selection operator (LASSO) and best subset selection (BSS), evaluated by respondent working curve and decision curve analysis for sensitivity. Six differentially methylated positions (DMPs) in the MCI group and 64 in the AD group were found, respectively. Among them, there were 2 DMPs in the MCI group and 30 DMPs in the AD group independent of age, gender, and APOE4 carriers (p < 0.05). AD diagnostic prediction models differentiated AD from normal controls both in a training dataset (LASSO: 8 markers, including methylation levels at ABCA7 1040077, CNR1 88166293, CX3CR1 39322324, LRRK2 40618505, LRRK2 40618493, NGFR 49496745, TARDBP 11070956, TARDBP 11070840 area under the curve [AUC] = 0.81; BSS: 2 markers, including methylation levels at ABCA7 1040077 and CX3CR1 39322324, AUC = 0.80) and a testing dataset (AUC = 0.84, AUC = 0.82, respectively). Our work indicated that methylation levels of 7 key immunologic-related genes (ABCA7, CNR1, CX3CR1, CSF1R, LRRK2, NGFR, and TARDBP) in peripheral blood was altered in AD and the models including methylation of immunologic-related genes biomarkers improved prediction of AD.