Multi-omics integration identifies a gut-microbiota-associated plasma metabolite signature for alzheimer disease diagnosis and metabolic pathway targeting

Abstract

Background Alzheimer’s disease (AD) is the leading cause of dementia, with early diagnosis remaining constrained by the invasiveness and limited accessibility of current biomarkers. The gut–brain axis has emerged as a critical communication network linking intestinal homeostasis with central nervous system function, yet the diagnostic utility of gut-microbiota-associated circulating metabolites in AD remains largely unexplored. Methods In this case-control study, 85 participants (37 AD patients and 48 cognitively normal controls) were enrolled. Targeted plasma metabolomics using four complementary UPLC-MS/MS assays covering 375 endogenous metabolites was integrated with fecal metagenomic sequencing in 72 participants with paired samples. Differential metabolites were screened using the Mann–Whitney U test, orthogonal partial least squares discriminant analysis, and fold-change thresholds. Gut-microbiota-associated plasma metabolites were defined through Spearman correlation with linear discriminant analysis effect size-identified differential bacterial species. A diagnostic panel was constructed using least absolute shrinkage and selection operator logistic regression with stability selection, and its performance was evaluated by support vector machine with repeated cross-validation, with age and sex evaluated as covariates. Methods We identified 27 differential plasma metabolites enriched in tricarboxylic acid cycle and amino acid metabolic pathways. Correlation analysis identified 10 gut-microbiota-associated plasma metabolites significantly associated with gut microbial alterations. A three-metabolite panel (citric acid, 3-hydroxyoctanoic acid, and glyoxylic acid) achieved a cross-validated area under the curve of 0.865, with a modest sensitivity of 62.2% and specificity of 100.0%, positioning it as a confirmatory auxiliary tool rather than a broad screening instrument. The panel score correlated negatively with Mini-Mental State Examination scores (r = −0.544, P < 0.001), and integration with clinical cognitive assessment yielded an area under the curve of 0.978. Conclusion These findings potentially advance the translational understanding of the gut–brain metabolic axis in neurodegeneration and support the exploratory clinical utility of microbiota-linked peripheral metabolic signatures as non-invasive auxiliary diagnostic tools for AD, awaiting external validation in larger, multi-center cohorts.