Scholarlink

Background <p>The progression of colorectal cancer (CRC) is profoundly shaped by immune and metabolic dysregulation within the tumor microenvironment (TME). This study aimed to develop an immune–metabolism-related gene (IMRG)-based prognostic model and experimentally validate its key molecular drivers.</p> Methods <p>Transcriptomic and clinical data from the TCGA and GEO cohorts were analyzed to identify differentially expressed IMRGs. Molecular subtypes were defined using non-negative matrix factorization (NMF) clustering, and a machine learning–based prognostic model integrating 101 algorithmic combinations was constructed and validated. Gene Set Enrichment Analysis (GSEA), immune infiltration profiling, and drug sensitivity analyses were performed to elucidate biological and immunological differences. IL20RB, identified as a hub IMRG, was further validated through qRT–PCR, Western blotting, and IHC. Its functional role was assessed using CCK-8, wound-healing, and Transwell assays.</p> Results <p>NMF clustering revealed two distinct molecular subtypes of CRC with divergent survival outcomes and immune characteristics. The 4-gene IMRG-based model demonstrated robust prognostic performance (C-index = 0.657) and was externally validated in a GEO cohort (AUC up to 0.824 for 1-, 3-, and 5-year survival). GSEA showed that the high-risk group was enriched for complement/coagulation cascades and extracellular matrix remodeling, whereas the low-risk group was enriched for metabolic programs including butanoate metabolism and the TCA cycle. Immune profiling indicated increased CD8⁺ T-cell infiltration in the high-risk group and a distinct immune checkpoint landscape, with most checkpoints elevated in the low-risk group, while ADORA2A, TNFRSF25 and CD276 were relatively higher in the high-risk group. Experimental analyses confirmed that IL20RB was significantly upregulated in CRC tissues and cell lines, and its silencing markedly inhibited CRC cell proliferation, migration, and invasion, consistent with its predicted oncogenic role.</p> Conclusion <p>This integrative transcriptomic and computational analysis combined with machine learning and experimental validation identified IMRGs—particularly IL20RB—as critical mediators of CRC progression. The proposed IMRG-based prognostic model offers a robust framework for patient risk stratification, immune landscape characterization, and therapeutic targeting in CRC.</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

Integrative machine learning models reveal immune and metabolic signatures predictive of colorectal cancer prognosis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月3日

查看全文
  • Zhanyuan Sun,
  • Zijing Wang,
  • Qingchen Lv,
  • Shaohua Hou,
  • Guo Li,
  • Tao Jiang,
  • Hai Li

摘要

Background

The progression of colorectal cancer (CRC) is profoundly shaped by immune and metabolic dysregulation within the tumor microenvironment (TME). This study aimed to develop an immune–metabolism-related gene (IMRG)-based prognostic model and experimentally validate its key molecular drivers.

Methods

Transcriptomic and clinical data from the TCGA and GEO cohorts were analyzed to identify differentially expressed IMRGs. Molecular subtypes were defined using non-negative matrix factorization (NMF) clustering, and a machine learning–based prognostic model integrating 101 algorithmic combinations was constructed and validated. Gene Set Enrichment Analysis (GSEA), immune infiltration profiling, and drug sensitivity analyses were performed to elucidate biological and immunological differences. IL20RB, identified as a hub IMRG, was further validated through qRT–PCR, Western blotting, and IHC. Its functional role was assessed using CCK-8, wound-healing, and Transwell assays.

Results

NMF clustering revealed two distinct molecular subtypes of CRC with divergent survival outcomes and immune characteristics. The 4-gene IMRG-based model demonstrated robust prognostic performance (C-index = 0.657) and was externally validated in a GEO cohort (AUC up to 0.824 for 1-, 3-, and 5-year survival). GSEA showed that the high-risk group was enriched for complement/coagulation cascades and extracellular matrix remodeling, whereas the low-risk group was enriched for metabolic programs including butanoate metabolism and the TCA cycle. Immune profiling indicated increased CD8⁺ T-cell infiltration in the high-risk group and a distinct immune checkpoint landscape, with most checkpoints elevated in the low-risk group, while ADORA2A, TNFRSF25 and CD276 were relatively higher in the high-risk group. Experimental analyses confirmed that IL20RB was significantly upregulated in CRC tissues and cell lines, and its silencing markedly inhibited CRC cell proliferation, migration, and invasion, consistent with its predicted oncogenic role.

Conclusion

This integrative transcriptomic and computational analysis combined with machine learning and experimental validation identified IMRGs—particularly IL20RB—as critical mediators of CRC progression. The proposed IMRG-based prognostic model offers a robust framework for patient risk stratification, immune landscape characterization, and therapeutic targeting in CRC.