Introduction: The aim of this study was to (1) replicate previous associations between six blood lipids and Alzheimer's disease (AD) (Proitsi et al 2015) and (2) identify novel associations between lipids, clinical AD diagnosis, disease progression and brain atrophy (left/right hippocampus/entorhinal cortex).
Methods: We performed untargeted lipidomic analysis on 148 AD and 152 elderly control plasma samples and used univariate and multivariate analysis methods.
Results: We replicated our previous lipids associations and reported novel associations between lipids molecules and all phenotypes. A combination of 24 molecules classified AD patients with >70% accuracy in a test and a validation data set, and we identified lipid signatures that predicted disease progression (R2 = 0.10, test data set) and brain atrophy (R2 ≥ 0.14, all test data sets except left entorhinal cortex). We putatively identified a number of metabolic features including cholesteryl esters/triglycerides and phosphatidylcholines.
Discussion: Blood lipids are promising AD biomarkers that may lead to new treatment strategies.
Keywords: Alzheimer's disease; Biomarkers; Brain atrophy; Classification; Dementia; Lipidomics; Machine learning; Metabolomics; Multivariate; Random forest; Rate of cognitive decline; sMRI.
Copyright © 2016 The Authors. Published by Elsevier Inc. All rights reserved.