Machine learning to detect Alzheimer’s disease from circulating non-coding RNAs
AbstractBackgroundTo develop therapeutics for Alzheimer’s disease, early detection of patients awakes new hope. Circulating small non-coding RNAs are among the prominent candidates for a blood-based diagnosis, requiring however growing cohort sizes.MethodsWe determined abundance levels of 21 known circulating microRNAs in 465 individuals encompassing Alzheimer’s patients and controls recruited in US and Germany. We computed models to assess the relation between microRNA-expression and phenotypes, gender, age and disease severity (Mini-Mental State Examination MMSE).Results20 of 21 miRNAs were consistently dys-regulated in the US and Germany. 18 were significantly correlated to neurodegeneration (adjusted p<0.05) with highest significance for miR-532-5p (adjusted p=4.8×10−30). Ten miRNAs were significantly correlated with MMSE, in particular miR-26a/26b-5p (adjusted p=0.0002). Machine learning models reached an AUC value of 87.6% in differentiating AD patients from controls.ConclusionsOur data provide strong evidence for the relevance of circulating non-coding RNAs to detect Alzheimer’s from a blood sample.