Abstract: Objective:?This study aims to identify key immune cells and differentially expressed genes (DEGs) associated with oxidative stress (OS) in periodontitis by integrating bulk RNA sequencing (Bulk RNA-seq), single-cell RNA sequencing (scRNA-seq), and machine learning algorithms. Methods:?Periodontitis-related gene expression data were collected from the GEO database, to investigate the heterogeneity of OS across various cellular tiers following PD and treatment. Then, OS activity of different immune cell subgroups were evaluated using AUCELL, UCELL, SINGSCORE, and ADDMODULESCORE algorithms to identify key immune cells with high OS activity in periodontitis. Furthermore, machine learning models such as Random Forest, Least Absolute Shrinkage and Selection Operator regression, and Artificial Neural Network were employed to select the best feature genes and construct a risk prediction model. Results: Results from scRNA-seq dataset indicated that OS activity exhibited cellular heterogeneity following PD, with particularly heightened activity observed in monocytes. Next, 3 800 DEGs were identified, of which 111 were OS-related DEGs (OS-DEGs). Machine learning algorithms identified four key OS genes (NFE2L2, KDR, CXCL1, CYBB), which showed significant expression changes in periodontitis patients (P＜0.05) and were highly associated with periodontitis risk. Conclusions: This study unveils the cellular heterogeneity of OS activity and systematically identifies key oxidative stress genes in periodontitis. NFE2L2 and CYBB may exacerbate OS activity by regulating monocyte migration and infiltration, penetrating the periodontal barrier.

Key words: periodontitis, oxidative stress, scRNA-seq, risk prediction model