Abstract
BACKGROUND: Fingernail metabolomics provides a novel, non-invasive platform that captures long-term biochemical fluctuations for identifying reliable biomarkers for dementia and mild cognitive impairment (MCI) due to Alzheimer's disease (AD). METHODS: A total of 199 participants were enrolled and stratified according to Clinical Dementia Rating (CDR) scores (0, 0.5, 1, 2, and 3). Fingernail clippings were collected and analysed using gas chromatography-mass spectrometry (GC-MS) based metabolomic. Differentially expressed metabolites (DEMs) across cognitive groups were identified using clustering, ordinal logistic regression, and machine learning approaches. Pathway enrichment and correlation analyses were conducted to explore underlying disease mechanisms and clinical relevance. RESULTS: Thirty DEMs were identified across the five CDR categories. Among them, Dodecanoic Acid demonstrated a marked and progressive decline from cognitively normal individuals (CDR = 0) to those with advanced AD (CDR = 3). After adjustment for age, sex, education, body mass index, lifestyle factors, nutrition, and sleep quality, Dodecanoic Acid remained independently associated with disease severity (OR = 0.845, p = 0.019). Importantly, within each CDR category (0.5, 1, 2, and 3), Dodecanoic Acid levels showed no significant differences between individuals with and without 18F-AV45 PET-confirmed amyloid pathology (all p > 0.05). Correlation analysis revealed that lower levels of Dodecanoic Acid were linked to greater cognitive impairment (AVLT-IR: r = 0.29; ADAS-Cog: r = -0.32). Pathway enrichment analysis highlighted significant disruptions in fatty acid metabolism, suggesting deficits in energy regulation during AD progression. A machine learning model using 29 DEMs achieved an overall classification accuracy of 67.2 % for differentiating participants into normal cognition (NC), MCI, or AD dementia groups. The model yielded a micro-averaged AUC of 0.803, with one-vs-rest AUCs ranging from 0.71 to 0.87, indicating robust discriminatory performance. Notably, dodecanoic acid was the top contributor in the model, underscoring its potential as a diagnostic biomarker. CONCLUSIONS: Dodecanoic Acid emerges as a key biomarker reflecting disrupted fatty acid metabolism during AD progression. By enabling long-term metabolic profiling, fingernail metabolomics presents a promising, scalable, and non-invasive strategy for early diagnosis, staging, and monitoring of neurodegenerative diseases.