Integration of single-cell sequencing with machine learning and Mendelian randomization analysis identifies the NAP1L1 gene as a predictive biomarker for Alzheimer's disease

Background: The most effective approach to managing Alzheimer's disease (AD) lies in identifying reliable biomarkers for AD to forecast the disease in advance, followed by timely early intervention for patients. Methods: Transcriptomic data on peripheral blood mononuclear cells (PBMCs) from patients with AD and the control group were collected, and preliminary data processing was completed using standardized analytical methods. PBMCs were initially segmented into distinct subpopulations, and the divisions were progressively refined until the most significantly altered cell populations were identified. A combination of high-dimensional weighted gene co-expression analysis (hdWGCNA), cellular communication, pseudotime analysis, and single-cell regulatory network inference and clustering (SCENIC) analysis was used to conduct single-cell transcriptomics analysis and identify key gene modules from them. Genes were screened using machine learning (ML) in the key gene modules, and internal and external dataset validations were performed using multiple ML methods to test predictive performance. Finally, bidirectional Mendelian randomization (MR) analysis, regional linkage analysis, and the Steiger test were employed to analyze the key gene. Result: A significant decrease in non-classical monocytes was detected in PMBC of AD patients. Subsequent analyses revealed the inherent connection of non-classical monocytes to AD, and the NAP1L1 gene identified within its gene module appeared to exhibit some association with AD as well. Conclusion: The NAP1L1 gene is a potential predictive biomarker for AD.