Employ machine learning to identify NAD+ metabolism-related diagnostic markers for ischemic stroke and develop a diagnostic model

Abstract

Ischemic stroke (IS) is a severe condition regulated by complex molecular alterations. This study aimed to identify potential nicotinamide adenine dinucleotide (NAD+) metabolism-associated diagnostic markers of IS and explore their associations with immune dynamics. Weighted Gene Co-expression Network Analysis and single-sample gene set enrichment analysis (ssGSEA) were employed to identify key gene modules on the GEO dataset (GSE16561). LASSO regression was used to identify diagnostic genes. A diagnostic model was then developed using the training dataset, and its performance was assessed using a validation dataset (GSE22255 dataset). Associations between hub genes and immune cells, immune response genes, and human leukocyte antigen (HLA) genes were assessed by ssGSEA. A regulatory network was constructed using mirBase and TRRUST databases. A total of 20 NAD+ metabolic genes exhibited noteworthy expression variations. Within the module notably associated with NAD+ metabolism, 19 specific genes were included in the diagnostic model, which was validated on the GSE22255 dataset (AUC: 0.733). There were significant disparities in immune cell populations, immune response genes, and HLA gene expression, all of which were associated with the hub genes. A regulatory network composed of 153 edges and 103 nodes was constructed. This study advances our understanding of IS by providing insights into NAD+ metabolism and gene interactions, contributing to potential diagnostic innovations in IS.