Cellular senescence is a state in which cells permanently stop dividing but remain alive and metabolically active. Senescent cells accumulate with aging and in response to stress such as DNA damage or cancer-causing mutations. Rather than dying normally, these cells persist and release a complex mix of inflammatory signals called the senescence-associated secretory phenotype (SASP).
In cancer, senescence plays a paradoxical role. On one hand, it can act as a tumor-suppressive mechanism by preventing damaged cells from proliferating. On the other hand, the SASP released by senescent cells can promote tumor growth, suppress immune function, and create conditions that help cancer spread. Understanding the net effect of senescence in a specific cancer type requires detailed molecular analysis.
This study developed a Senescence-Related Scoring Model (SRSM) for clear cell renal cell carcinoma (ccRCC) using a comprehensive machine learning approach. The model uses gene expression patterns to assign each patient's tumor a senescence score, which predicts prognosis and immune microenvironment characteristics.
The research team tested 101 combinations of machine learning algorithms to identify the most accurate and robust method for constructing the senescence scoring model, ensuring that the final result reflects a genuinely optimized approach rather than an arbitrary algorithmic choice.
Clear cell RCC is the most common subtype of kidney cancer, accounting for approximately 75% of cases. It is characterized by mutations in the VHL gene, which normally helps suppress tumor development. Advanced ccRCC responds to targeted therapies and immunotherapy, but resistance is common and prognosis for metastatic disease remains poor.
Research has shown that the tumor microenvironment in ccRCC is highly immunologically complex, with substantial variation in immune cell infiltration and activity between patients. Senescent cells and their SASP products can significantly influence this immune landscape, potentially explaining why some patients respond better to immunotherapy than others.
Previous studies of senescence in cancer have identified individual senescence-related genes as prognostic markers, but none had systematically evaluated which combination of senescence genes and machine learning algorithms best predicts outcomes in ccRCC. This study fills that gap by applying a rigorous, multi-algorithm optimization strategy.
The researchers used data from multiple independent ccRCC cohorts to train and validate the model, including the TCGA-KIRC dataset and several GEO (Gene Expression Omnibus) datasets. Using multiple independent datasets helps ensure that the model captures genuine biological signals rather than patterns specific to a single study population.
The team began with a comprehensive set of senescence-related genes curated from the CellAge database and published literature. These genes were then narrowed down using differential expression analysis comparing tumor and normal kidney tissue, retaining only those showing significant changes in expression in ccRCC.
To identify the optimal predictive signature, the researchers tested 101 different combinations of machine learning algorithms including LASSO regression, elastic net, stepwise regression, random forest, gradient boosting, support vector machines, and several others. Each combination was evaluated on training data and independently validated across multiple test cohorts.
The algorithm combination that consistently achieved the highest performance across all validation cohorts was selected as the basis for the final SRSM. This approach avoids the risk of overfitting to a specific dataset and provides greater confidence that the model will generalize to new patients in different clinical settings.
The final SRSM incorporated nine genes as its core signature. Patients were stratified into high-SRSM and low-SRSM groups based on their score, and the clinical, immunological, and therapeutic implications of each group were systematically analyzed.
To validate biological relevance, the researchers performed in vitro experiments using ccRCC cell lines. They manipulated the expression of NME2, one of the most prominent signature genes, and observed the effects on cell behavior including proliferation, migration, and senescence markers, directly testing the functional role of this gene in kidney cancer cells.
Patients with a high SRSM score had significantly worse overall survival compared to low-SRSM patients across all cohorts tested. The model successfully stratified patients into distinct risk groups, and this survival difference remained significant even after adjusting for conventional clinical factors such as age, stage, and grade in multivariate analyses.
High-SRSM tumors displayed characteristics of an immunosuppressive microenvironment. They had greater infiltration by regulatory T cells and M2-polarized macrophages, which actively suppress anti-tumor immune responses, and lower levels of cytotoxic CD8+ T cells, which are the primary immune cells capable of killing cancer. This immune landscape is associated with poorer responses to checkpoint inhibitor immunotherapy.
Pathway analysis revealed that high-SRSM tumors also showed enhanced activity of oxidative phosphorylation pathways, the metabolic process by which cells generate energy using oxygen in the mitochondria. Enhanced oxidative phosphorylation is a metabolic feature associated with drug resistance in several cancer types and may represent a therapeutic vulnerability in high-SRSM ccRCC.
Drug sensitivity predictions using the GDSC database suggested that high-SRSM patients might respond differently to several agents used in kidney cancer treatment, including mTOR inhibitors and specific chemotherapy compounds. These predictions provide hypotheses for future clinical investigation of treatment selection based on senescence score.
NME2 (also known as NM23-H2) is a gene encoding a protein involved in cellular energy metabolism and genome stability. It appeared consistently among the top-ranked genes across multiple algorithm combinations, suggesting it plays a particularly important role in the senescence landscape of ccRCC.
In vitro experiments using ccRCC cell lines confirmed that NME2 is functionally relevant. When NME2 expression was reduced using gene silencing techniques, cells showed changes in proliferation and migration rates, as well as alterations in markers of cellular senescence. This functional validation establishes that NME2 is not merely a statistical association but an actively participating gene in kidney cancer cell biology.
NME2 has been studied in other cancer types where it has shown both tumor-suppressive and tumor-promoting roles depending on the cancer context. In ccRCC, its role within the senescence network appears to be connected to the immunosuppressive microenvironment, making it a candidate for further investigation as a prognostic biomarker and potential therapeutic target.
The identification and functional validation of NME2 demonstrates the value of the comprehensive multi-algorithm approach. By screening 101 algorithm combinations and selecting genes that were consistently identified across methods, the team improved confidence that NME2 is a genuine driver rather than a noise artifact from a single analysis pipeline.
The SRSM offers a potential tool for stratifying ccRCC patients at diagnosis or before treatment decisions. Patients with high SRSM scores may be less likely to respond to immunotherapy and might benefit from alternative or combination approaches that address their metabolic and immunosuppressive tumor characteristics.
For oncologists, a validated senescence score could complement existing biomarkers such as PD-L1 expression when deciding which patients to treat with checkpoint inhibitors. The current limitations of PD-L1 as a predictive biomarker in ccRCC mean that additional biomarkers are urgently needed, and the SRSM may fill part of that gap.
The metabolic findings, specifically the link to enhanced oxidative phosphorylation in high-SRSM tumors, suggest a potential treatment angle. Drugs that target mitochondrial metabolism are under investigation in several cancer types, and ccRCC patients with high SRSM scores might be particularly good candidates for trials combining metabolic inhibitors with standard systemic therapy.
For patients and families, this research contributes to a growing toolkit for personalized kidney cancer care. While the SRSM is not yet ready for direct clinical application, it represents the kind of molecular insight that is progressively moving kidney cancer treatment away from one-size-fits-all approaches toward individualized therapy guided by tumor biology.
This study establishes cellular senescence as a biologically significant and clinically relevant dimension of ccRCC behavior. The SRSM, built through systematic optimization across 101 machine learning algorithm combinations, provides a robust and reproducible tool for quantifying senescence activity in patient tumors.
The integration of genomic analysis, immune profiling, metabolic characterization, and in vitro validation gives the findings a multi-layered biological foundation that strengthens their credibility. The convergent evidence from computational and experimental approaches is particularly compelling for NME2 as a candidate biomarker and therapeutic target.
Future research should focus on prospective clinical validation of the SRSM in real-world patient cohorts, exploration of senescence-targeting therapies known as senolytics in preclinical ccRCC models, and investigation of whether the SRSM score changes in response to treatment and whether those changes correlate with clinical outcomes.