Development of a diagnostic and risk prediction model for Alzheimer's disease through integration of single-cell and bulk transcriptomic analysis of glutamine metabolism.

Background: In this study, we present a novel system for quantifying glutamine metabolism (GM) to enhance the effectiveness of Alzheimer's disease (AD) diagnosis and risk prediction. Methods: Single-cell RNA sequencing (scRNA-seq) analysis was utilized to comprehensively assess the expression patterns of GM. The WGCNA algorithm was applied to investigate the most significant genes related to GM. Subsequently, three machine learning algorithms (Boruta, LASSO, and SVM-RFE) were employed to identify GM-associated characteristic genes and develop a risk model. Patients were divided into high- and low-risk groups based on this model. Moreover, we explored biological properties, distinct signaling pathways, and immunological characteristics of AD patients at different risk levels. Finally, Results: Both scRNA-seq and bulk transcriptomic analyses revealed increased GM activity in AD patients, specifically in certain cell subsets (pDC, Tem/Effector helper T cells (LTB), and plasma cells). Cells with higher GM scores demonstrated more significant numbers and strengths of interactions with other cell types. The WGCNA algorithm identified 360 genes related to GM, and a risk score was constructed based on nine characteristic genes (ATP13A4, PIK3C2A, CD164, PHF1, CES2, PDGFB, LCOR, TMEM30A, and PLXNA1) identified through multiple machine learning algorithms displayed reliable diagnostic efficacy for AD onset. Nomograms, calibration curves, and decision curve analysis (DCA) based on these characteristic genes provided significant clinical benefits for AD patients. High-risk AD patients exhibited higher levels of immune-related functions and pathways, increased immune cell infiltration, and elevated expressions of immune modulators. RT-qPCR analysis revealed that the majority of the nine characteristic genes were differentially expressed in AD-induced rat neurons. Knocking down PHF1 could protect against neurite loss and alleviate cell injury in AD neurons. Conclusion: This comprehensive identification of gene expression patterns contributes to a deeper understanding of the underlying pathological mechanisms driving AD pathogenesis. Furthermore, the risk model based on the nine-gene signature offers a promising theoretical foundation for developing individualized treatments for AD patients.