Obstructive sleep apnea (OSA) is caused by repeated blockage of the upper respiratory airways during sleep. The traditional evaluation methods for OSA severity are yet limited. This study aimed to screen gene signatures to effectively evaluate OSA severity. Expression profiles of peripheral blood mononuclear cells in the different severities of OSA patients were accessed from Gene Expression Omnibus (GEO) database. A total of 446 differentially expressed genes (DEGs) were screened among the varying severities of OSA samples by analysis of variance (ANOVA) test. A total of 1,152 DEGs were screened between the pre- and post-treatment OSA samples by using t   test. Overlap of the two groups of DEGs was selected (88 DEGs) for Metascape enrichment analysis. Afterwards, Mfuzz package was used to perform soft clustering analysis on these 88 genes, by which 6 clusters were obtained. It was observed that the gene expression condition of the cluster 3 was positively associated with OSA severity degree; also, the gene expression condition in cluster 4 was negatively correlated with OSA severity. A total of 10 gene markers related to OSA progression were selected from cluster 3 and cluster 4. Their expression levels and correlation were analyzed. The marker genes in cluster 3 and cluster 4 were examined, finding that most genes were significantly correlated with apnea hypopnea index (AHI). An accurate and objective assessment of the severity of OSA is of great significance for formulating follow-up treatment strategies for patients with OSA. In this paper, a set of marker genes that can detect the severity of OSA were screened by bioinformatics methods, which could be jointly used with the traditional OSA diagnostic index to achieve a more reliable OSA severity evaluation.