Abstract

As a crucial type of post-translational modification, glycosylation plays a fundamental role in maintaining cellular homeostasis and is closely associated with the progression of Alzheimer's Disease (AD). Given the central involvement of the hippocampus in AD pathogenesis, elucidating the mechanisms of glycosylation in this brain region may provide critical insights and facilitate the development of precision medicine strategies for AD. We employed an integrated bioinformatics framework to identify glycosylation-related diagnostic biomarkers for AD. Limma and WGCNA were conducted on a hippocampal gene expression microarray dataset to detect glycosylation-associated DEGs, which were intersected with a glycosylation gene set obtained from GeneCards. Key diagnostic genes were selected using three machine learning algorithms in an independent cohort. The diagnostic model was subsequently validated in two additional independent microarrays datasets. Functional exploration and hippocampal heterogeneity were assessed at both bulk-tissue and single-cell levels. PPI analysis and NMF clustering further stratified AD patients into two subtypes. Finally, qRT-PCR validated the expression of biomarkers, and molecular docking based on the CTD suggested potential therapeutic candidates. Our analysis identified CKMT1B and AP1S1 as key downregulated glycosylation-related genes in AD. These genes were predominantly and highly enriched in hippocampal microglia at both bulk and single-cell levels and demonstrated strong diagnostic potential. PPI network analysis and NMF revealed that these hub genes could stratify AD patients into two distinct molecular subgroups. Furthermore, quercetin was identified as a potential multi-target therapeutic agent through database screening and CTD molecular docking studies. Collectively, this study bridges fundamental discovery with clinical translation by providing a diagnostic model, patient stratification subtypes, and a repositioned therapeutic can