Machine learning developed regulatory T cells-related signature for prognosis and immunotherapy benefit in oral squamous cell carcinoma

Abstract

Background

Oral squamous cell carcinoma (OSCC) is one of the most common malignancies with poor clinical outcome. Regulatory T cells (Tregs) have a dual role in maintaining immune homeostasis and suppressing anti-tumor immunity. The role of Tregs related genes (TRGs) in the prognosis of OSCC patients were rarely been reported.

Methods

Integrative analysis procedure containing 10 machine learning methods was used to develop a Tregs related gene signature (TRS) using datasets from TCGA, GSE41613, GSE65858, and GSE117973 cohort. Several predicting scores were used to investigate the performance of TRS in predicting immunotherapy benefit. The biological function of TNFAIP3 was explored by in vitro assay.

Results

The prognostic signature constructed using the LASSO algorithm was identified as the optimal TRS with the highest average C-index of 0.78. TRS served as a prognostic biomarker, with low TRS score correlating with favorable clinical outcome. Specifically, in TCGA dataset, the 1-, 3-, and 5-year AUC values were 0.796, 0.838 and 0.812, respectively. OSCC with low TRS score exhibited higher levels of immune-activated cells and immune-related functions, including CD8⁺ T, macrophage M1, and cytolytic activity. Low TRS indicated higher PD1&CTLA4 immunophenoscores, higher TMB, higher MSI, lower tumor immune dysfunction and exclusion score, decreased immune escape score, and lower intratumor heterogeneity scores. Additionally, OSCC cases with high TRS score showed elevated cancer-related hallmark score. Knock-down of TNFAIP3 inhibited OSCC cell proliferation.

Conclusion

This research established an optimal TRS for OSCC, which functions as a valuable indicator for forecasting clinical outcomes and assessing potential benefits from immunotherapy.