RESUMO
Ovarian cancer (OC) is a malignancy associated with poor prognosis and has been linked to regulatory T cells (Tregs) in the immune microenvironment. Nevertheless, the association between Tregs-related genes (TRGs) and OC prognosis remains incompletely understood. The xCell algorithm was used to analyze Tregs scores across multiple cohorts. Weighted gene co-expression network analysis (WGCNA) was utilized to identify potential TRGs and molecular subtypes. Furthermore, we used nine machine learning algorithms to create risk models with prognostic indicators for patients. Reverse transcription-quantitative polymerase chain reaction and immunofluorescence staining were used to demonstrate the immunosuppressive ability of Tregs and the expression of key TRGs in clinical samples. Our study found that higher Tregs scores were significantly correlated with poorer overall survival. Recurrent patients exhibited increased Tregs infiltration and reduced CD8+ T cell. Moreover, molecular subtyping using seven key TRGs revealed that subtype B exhibited higher enrichment of multiple oncogenic pathways and had a worse prognosis. Notably, subtype B exhibited high Tregs levels, suggesting immune suppression. In addition, we validated machine learning-derived prognostic models across multiple platform cohorts to better distinguish patient survival and predict immunotherapy efficacy. Finally, the differential expression of key TRGs was validated using clinical samples. Our study provides novel insights into the role of Tregs in the immune microenvironment of OC. We identified potential therapeutic targets derived from Tregs (CD24, FHL2, GPM6A, HOXD8, NAP1L5, REN, and TOX3) for personalized treatment and created a machining learning-based prognostic model for OC patients, which could be useful in clinical practice.