Scholarlink

Backgrounds <p>Immune cells play a crucial role in the tumor microenvironment (TME) by regulating the progression of cancer cells. However, the clinical relevance of immune cell infiltration-related mRNA in thyroid cancer (TC) remains uncertain. Current diagnostic methods, such as cytology and imaging, still face limitations in accurately assessing tumor behavior and prognosis, highlighting the need for more reliable molecular indicators.</p> Methods <p>Three cohorts (TCGA, GSE3678, and GSE33630) were included in the study to construct immune-related signatures for thyroid cancer (TC). The immune cell infiltration levels were quantified using single-sample gene set enrichment analysis (ssGSEA), followed by consensus clustering to identify immune cell-related molecular subtypes (IRMS). Subsequently, immune-related genes (IRGs) were selected via weighted gene co-expression network analysis (WGCNA). Based on these IRGs, we established two predictive models: an immune cell-related diagnostic signature (IRDS) was developed using a machine-learning framework with 113 combinations of 12 machine-learning algorithms, while an immune cell-related prognostic signature (IRPS) was constructed via LASSO regression algorithm. Finally, both signatures were systematically evaluated for their predictive performance.</p> Results <p>Through WGCNA analysis, key immune-related gene modules were identified in the TCGA-THCA cohort. From these modules, 28 immune-related genes were selected based on their expression patterns and univariate Cox regression results (<i>p</i> &lt; 0.05), and were subsequently used to construct both an immune-related diagnostic signature (IRDS) and an immune-related prognostic signature (IRPS). The IRDS demonstrated strong diagnostic performance (AUC = 0.958) and robustness across multiple validation cohorts (TCGA, GSE33630, and GSE3678). Similarly, the IRPS accurately predicted prognosis in the TCGA-THCA cohort (5-year AUC = 0.888), with higher IRPS scores associated with poorer survival outcomes (HR = 4.604, <i>P</i> = 2.24e-04). Moreover, the IRPS could serve as an independent prognostic factor, and patients with lower IRPS scores showed potential benefit from immunotherapy and chemotherapy. Finally, a candidate drug target CNR2 was identified for the high IRPS group patients.</p> Conclusions <p>The developed IRDS and IRPS show potential for improving thyroid cancer diagnosis and risk stratification. Furthermore, CNR2 represents a promising candidate therapeutic target for patients with high IRPS scores, highlighting a actionable path toward personalized treatment.</p> Graphical Abstract <p></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

Immune cell–related gene signatures for diagnostic and prognostic stratification in thyroid cancer using machine learning analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月18日

查看全文
  • Kun Fang,
  • Zhifang Zhang,
  • Meng Luo,
  • Xudong Niu

摘要

Backgrounds

Immune cells play a crucial role in the tumor microenvironment (TME) by regulating the progression of cancer cells. However, the clinical relevance of immune cell infiltration-related mRNA in thyroid cancer (TC) remains uncertain. Current diagnostic methods, such as cytology and imaging, still face limitations in accurately assessing tumor behavior and prognosis, highlighting the need for more reliable molecular indicators.

Methods

Three cohorts (TCGA, GSE3678, and GSE33630) were included in the study to construct immune-related signatures for thyroid cancer (TC). The immune cell infiltration levels were quantified using single-sample gene set enrichment analysis (ssGSEA), followed by consensus clustering to identify immune cell-related molecular subtypes (IRMS). Subsequently, immune-related genes (IRGs) were selected via weighted gene co-expression network analysis (WGCNA). Based on these IRGs, we established two predictive models: an immune cell-related diagnostic signature (IRDS) was developed using a machine-learning framework with 113 combinations of 12 machine-learning algorithms, while an immune cell-related prognostic signature (IRPS) was constructed via LASSO regression algorithm. Finally, both signatures were systematically evaluated for their predictive performance.

Results

Through WGCNA analysis, key immune-related gene modules were identified in the TCGA-THCA cohort. From these modules, 28 immune-related genes were selected based on their expression patterns and univariate Cox regression results (p < 0.05), and were subsequently used to construct both an immune-related diagnostic signature (IRDS) and an immune-related prognostic signature (IRPS). The IRDS demonstrated strong diagnostic performance (AUC = 0.958) and robustness across multiple validation cohorts (TCGA, GSE33630, and GSE3678). Similarly, the IRPS accurately predicted prognosis in the TCGA-THCA cohort (5-year AUC = 0.888), with higher IRPS scores associated with poorer survival outcomes (HR = 4.604, P = 2.24e-04). Moreover, the IRPS could serve as an independent prognostic factor, and patients with lower IRPS scores showed potential benefit from immunotherapy and chemotherapy. Finally, a candidate drug target CNR2 was identified for the high IRPS group patients.

Conclusions

The developed IRDS and IRPS show potential for improving thyroid cancer diagnosis and risk stratification. Furthermore, CNR2 represents a promising candidate therapeutic target for patients with high IRPS scores, highlighting a actionable path toward personalized treatment.

Graphical Abstract