Identification of plasma biomarkers in lipid metabolism for accurate prediction of Alzheimer's disease.

Metabolomics may reveal non-invasive biomarkers for early diagnosis in Alzheimer's disease (AD) and provide new insights into the disease mechanisms to develop effective treatments. Here, we comprehensively analyzed the blood plasma metabolomes from a Chinese cohort of 447 individuals, including 188 AD, 181 MCI (mild cognitive impairment), and 78 NC (normal control). Differential analysis identified altered metabolites, followed by forward feature selection to prioritize a panel of key metabolites, and construction of a diagnostic model using logistic regression. Key metabolite-enriched pathways were identified and quantified for comparison across different groups, which was then validated through external datasets. We observed extensive metabolic dysregulation in AD compared to age-matched NC, with 25% of the differential metabolites also significantly dysregulated in MCI in the same directions. A panel of 22 key metabolites was prioritized, where triglycerides (TG) and phosphatidylethanolamines (PE) ranked top in importance. With these key metabolites, we trained a diagnostic model that classified AD from NC accurately (Area Under the Curve [AUC] = 0.935 in the replication cohort). Pathway quantification analysis showed significant changes in lipid metabolism in AD, which were validated in two external cohorts. We presented a precise and robust blood metabolic diagnostic model for AD, which may help promote early diagnosis and deepen the understanding of AD mechanisms.