Scholarlink

Background <p>Regulated cell death programs influence melanoma progression and antitumor immunity, yet a robust prognostic model integrating multiple cell-death modalities remains limited.</p> Methods <p>Transcriptomic and clinical data from TCGA and independent GEO cohorts (GSE19234, GSE22153, and GSE65904) were analyzed. Activity of diverse cell-death programs was quantified using ssGSEA/GSVA. Prognostic genes were screened by univariate Cox regression and used to construct a cell death-related signature (CDS) through systematic benchmarking of multiple machine-learning survival pipelines. Model performance was evaluated using C-index, and prognostic value was assessed by Kaplan-Meier survival analysis, time-dependent ROC, and PCA-based separation. CDS was further compared with published prognostic signatures. Tumor microenvironment (TME) features were inferred by ESTIMATE, and immune functional states were examined using cancer-immunity cycle-related GSVA modules. A key CDS component gene was prioritized and experimentally validated.</p> Results <p>Cell death programs showed heterogeneous activation across samples. A machine-learning-derived CDS built from prognostic cell death-related genes achieved stable discrimination across cohorts and significantly stratified patients into high- and low-risk groups. Time-dependent ROC analyses supported the predictive accuracy of CDS, and PCA indicated distinct transcriptomic patterns between risk groups. Compared with published signatures, CDS demonstrated competitive or superior performance in multiple datasets. Higher CDS risk was associated with reduced immune/stromal infiltration and increased tumor purity, consistent with an immune-cold phenotype and broadly attenuated cancer-immunity cycle activity. MAPK7 emerged as the most positively risk-associated gene; its high expression correlated with adverse molecular features, and MAPK7 knockdown suppressed clonogenic growth and migration in melanoma cell lines.</p> Conclusions <p>CDS is a robust multi-cell death mode prognostic signature for melanoma and reflects TME immune suppression. MAPK7 represents a potential functional driver and therapeutic target linked to the high-risk state.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A machine-learning-derived multi-cell death mode signature predicts melanoma prognosis and reveals an immune-cold tumor microenvironment

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月3日

查看全文
  • Dan Wu,
  • Yuze Zhou,
  • Jiahong Fang,
  • Jiaheng Xie,
  • Jie Ren

摘要

Background

Regulated cell death programs influence melanoma progression and antitumor immunity, yet a robust prognostic model integrating multiple cell-death modalities remains limited.

Methods

Transcriptomic and clinical data from TCGA and independent GEO cohorts (GSE19234, GSE22153, and GSE65904) were analyzed. Activity of diverse cell-death programs was quantified using ssGSEA/GSVA. Prognostic genes were screened by univariate Cox regression and used to construct a cell death-related signature (CDS) through systematic benchmarking of multiple machine-learning survival pipelines. Model performance was evaluated using C-index, and prognostic value was assessed by Kaplan-Meier survival analysis, time-dependent ROC, and PCA-based separation. CDS was further compared with published prognostic signatures. Tumor microenvironment (TME) features were inferred by ESTIMATE, and immune functional states were examined using cancer-immunity cycle-related GSVA modules. A key CDS component gene was prioritized and experimentally validated.

Results

Cell death programs showed heterogeneous activation across samples. A machine-learning-derived CDS built from prognostic cell death-related genes achieved stable discrimination across cohorts and significantly stratified patients into high- and low-risk groups. Time-dependent ROC analyses supported the predictive accuracy of CDS, and PCA indicated distinct transcriptomic patterns between risk groups. Compared with published signatures, CDS demonstrated competitive or superior performance in multiple datasets. Higher CDS risk was associated with reduced immune/stromal infiltration and increased tumor purity, consistent with an immune-cold phenotype and broadly attenuated cancer-immunity cycle activity. MAPK7 emerged as the most positively risk-associated gene; its high expression correlated with adverse molecular features, and MAPK7 knockdown suppressed clonogenic growth and migration in melanoma cell lines.

Conclusions

CDS is a robust multi-cell death mode prognostic signature for melanoma and reflects TME immune suppression. MAPK7 represents a potential functional driver and therapeutic target linked to the high-risk state.