BACKGROUND: Obstructive sleep apnea (OSA) is the most prevalent sleep disturbance that affects approximately 936 million people worldwide and leads to extensively increased incidence of cardiovascular disease, metabolic syndrome, neurological disorders, and traffic accidents. Severe OSA patients suffer a significantly higher risk of complications and worse comorbidity outcomes. Notwithstanding, with inadequate access to contact diagnosis based on polysomnography (PSG), numerous patients with severe sleep apnea have not been diagnosed, especially during the pandemic. Moreover, how the T cell immunity is impaired in OSA remains largely unknown. METHODS: We primarily investigated the T cell receptor (TCR) repertoires of 50 patients with severe OSA, 23 patients with mild-to-moderate OSA, 23 patients without OSA, and 157 healthy individuals, from their peripheral blood. Firstly, we compared the clinical characteristics, blood cell counts, the ratio of neutrophil-to-lymphocyte (NLR), platelet-to-lymphocyte (PLR), and CD4+/CD8+T cell count between groups. Then, we compared the diversity, clonotypes, unique VJ alleles in patients with different disease severity. Furthermore, by identifying a series of disease-associated amino acid sequences, we employed a repeated hold-out machine learning strategy to explore the optimal algorithm for calculating the TCR repertoire characteristic Index (OSA-TCI). We further confirmed its relation with clinical features by linear regression analysis. Moreover, in followup of severe OSA patients who accepted adherent non-invasive ventilation, we assessed the changes of TCR repertoires, OSA-TCI, ESS, NLR, PLR, and CD4+/CD8+T after therapy. RESULTS: We found an unexpected increase in diversity and clonotypes in the TCR repertoire of OSA patients. Furthermore, we successfully developed a novel indicator termed OSA-TCI to summarize the unique repertoire alteration, which provided 90% of sensitivity and 87% of specificity in distinguishing severe OSA. In rationalization, OSA-TCI was found correlated to AHI, BMI, hemoglobin, N1, N2 percentage of sleep, snoring, smoking and lowest oxygen saturation, but only independently related to AHI (R = 0.603) and smoking (R = 0.22). Finally, we observed OSA-TCI in the eight severe patients decreased significantly after home noninvasive ventilation for three months during follow-up, consistently in line with the TCR repertoire improvement. In contrast, NLR, PLR, and the ratio of CD4+/CD8+T cell count were found useless to diagnose and therapeutic surveillance of severe OSA. CONCLUSIONS: Our study is the first to unveil the TCR repertoire alteration in OSA, indicates possible insidious autoimmune mechanisms underlying OSA, and suggests that TCR repertoires serve as a convenient peripheral blood biomarker for OSA assessment without long-time contact and facility/instrument occupation. It may shed light on future diagnostic, immunological, pathophysiological, and prognostic research on OSA.