Discovery of a Metabolic Signature Predisposing High Risk Patients with Mild Cognitive Impairment to Converting to Alzheimer’s Disease

Assessing dementia conversion in patients with mild cognitive impairment (MCI) remains challenging owing to pathological heterogeneity. While many MCI patients ultimately proceed to Alzheimer's disease (AD), a subset of patients remain stable for various times. Our aim was to characterize the plasma metabolites of nineteen MCI patients proceeding to AD (P-MCI) and twenty-nine stable MCI (S-MCI) patients by untargeted metabolomics profiling. Alterations in the plasma metabolites between the P-MCI and S-MCI groups, as well as between the P-MCI and AD groups, were compared over the observation period. With the help of machine learning-based stratification, a 20-metabolite signature panel was identified that was associated with the presence and progression of AD. Furthermore, when the metabolic signature panel was used for classification of the three patient groups, this gave an accuracy of 73.5% using the panel. Moreover, when specifically classifying the P-MCI and S-MCI subjects, a fivefold cross-validation accuracy of 80.3% was obtained using the random forest model. Importantly, indole-3-propionic acid, a bacteria-generated metabolite from tryptophan, was identified as a predictor of AD progression, suggesting a role for gut microbiota in AD pathophysiology. Our study establishes a metabolite panel to assist in the stratification of MCI patients and to predict conversion to AD.