A comprehensive analysis of LC-MS metabolomics of plasma samples from a Chinese cohort for Alzheimer's disease.
a. This study covered a total of 477 individuals, which were divided into discovery cohort (n=257) and replication cohort (n=220). b. Untargeted metabolomics approach based on LC–MS was applied to plasma samples to get metabolic MS profiles. c. The metabolic profiles of Alzheimer's disease (AD) cases, mild cognitive impairment (MCI) cases and normal controls (NC) were compared in the discover cohort to obtain the differential metabolites. Then a key panel of metabolites were further determined as the metabolic biomarker for AD using the forward feature selection algorithm. d. Based on the biomarker, a logistic regression classifier for AD diagnosis was trained in the discovery cohort with 5-fold cross validation, and then evaluated in the replication cohort. e. Metabolic pathways were quantified by summarizing the abundance of the measured metabolites within the pathway to represent metabolic changes at the pathway level. f. Significantly altered pathways were validated in external cohorts by principal components analysis (PCA) and differential tests of pathway abundance.