A machine learning-defined cellular senescence signature systematically enhances prognostication and guides immunotherapy strategies for the treatment of gliomas.

Gliomas are the most common and heterogeneous primary brain tumors, which leads to poor prognosis in many cases. Cellular senescence plays a key role in tumor progression and drug resistance, yet the prognostic value of senescence in gliomas remains unclear. Here, we identified key senescence-related genes through consensus clustering and weighted gene co-expression network analysis (WGCNA), and developed a cellular senescence-related gene prognostic signature (CSRGPS) using ten machine learning algorithms. The CSRGPS demonstrated strong predictive power, outperforming traditional clinical and molecular models. It stratified patients into distinct prognostic groups exhibiting differences in survival, clinical features, biological functions, and the tumor microenvironment. Single-cell analysis revealed a transition from low to high CSRGPS states. Furthermore, clinical data indicated an association between low CSRGPS and better outcomes following anti-PD-1 therapy. We also developed a nomogram integrating CSRGPS and clinical data, which further improved individualized prognosis prediction. Overall, CSRGPS offers a robust, clinically applicable tool for glioma prognosis and immunotherapy guidance, with potential utility in other cancers.