Integrated Machine Learning and Single-Sample Gene Set Enrichment Analysis Identifies a TGF-Beta Signaling Pathway Derived Score in Headneck Squamous Cell Carcinoma.

Background:The TGF-β signaling pathway is clinically predictive of pan-cancer. Nevertheless, its clinical prognosis and regulation of immune microenvironment (TME) characteristics as well as the prediction of immunotherapy efficacy need to be further elucidated in head and neck squamous cell carcinoma. Method:At first, we summarized TGF-β related genes from previous published articles, used ssGSEA to establish the TGF-β risk score. Considering the complexity of its clinical application, we improved it with the LASSO-COX algorithm to construct the model. In addition, we explored the predictive efficacy of TGF-β risk score in the observation of TME phenotype and immunotherapy effect. Finally, the potency of TGF-β risk score in adjusting precise treatment of HNSC was evaluated. Results:We systematically established TGF-β risk score with multi-level predictive ability. TGF-β risk score was employed to predict the tumor microenvironment status, which was negatively associated with NK cells but positively related to macrophages and fibroblasts. It reveals that patients with high TGF-β risk score predict "cold" TME status. In addition, higher risk scores indicate higher sensitivity to immunotherapy. Conclusion:We first construct and validate TGF-β characteristics that can predict immune microenvironment phenotypes and immunotherapeutic effect in multiple datasets. Noteworthy, TGF-β risk score is helpful for individualized precise treatment of patients with the head and neck squamous cell carcinoma.