Prognostic role of stress granule-related Gene signatures in pancreatic ductal adenocarcinoma: insights from 101-combination machine learning and single-cell sequencing

Lei Wang; Linhao Zong; Yuanyuan Wang; Yanting Jiang; Miao Guan

doi:10.1097/JS9.0000000000004458

Prognostic role of stress granule-related Gene signatures in pancreatic ductal adenocarcinoma: insights from 101-combination machine learning and single-cell sequencing

Int J Surg. 2025 Dec 16. doi: 10.1097/JS9.0000000000004458. Online ahead of print.

Authors

Lei Wang¹, Linhao Zong², Yuanyuan Wang¹, Yanting Jiang¹, Miao Guan²

Affiliations

¹ Department of Oncology, Lianyungang Clinical College of Nanjing Medical University (The First People' s Hospital of Lianyungang), Lianyungang, Jiangsu, China.
² Jiangsu Key Laboratory for Biodiversity and Biotechnology, College of Life Sciences, Nanjing Normal University, Nanjing, Jiangsu, China.

PMID: 41403347
DOI: 10.1097/JS9.0000000000004458

Abstract

Background: Pancreatic ductal adenocarcinoma (PDAC) is an exceptionally aggressive malignancy of the digestive system, characterized by a fibrotic microenvironment that serves as an ideal model for studying stress granules (SGs). This study aimed to investigate SG-related mechanisms in PDAC, with particular focus on risk stratification and therapeutic strategies.

Methods: PDAC-related datasets were retrieved from TCGA and GEO databases. Differential expression analysis, univariate Cox regression, and 101 algorithmic combinations from 10 machine learning methods were employed to identify prognostic SG-related genes (SGRGs) and construct a risk model. Prognostic analyses were further extended through independent prognostic evaluation, nomogram development, immune microenvironment profiling, drug sensitivity testing, and enrichment analysis. Additionally, GSE197177 was examined to identify key cell types and perform pseudo-time and cell communication analyses.

Results: A risk model based on four prognostic SGRGs (LAMA3, ITGA6, COL17A1, and TOP2A) was developed, demonstrating superior predictive accuracy for PDAC prognosis. A nomogram incorporating age, N stage, and risk score was constructed, showing robust prognostic capacity. Further analyses revealed that immune cells, such as M0 macrophages and CD8 T cells, as well as drug sensitivities to ERK inhibitors and trametinib, were associated with risk stratification in PDAC patients. ITGA6 was notably enriched in the "regulation of glycolytic process" pathway. Pseudo-time analysis indicated a significant correlation between the expression of prognostic SGRGs and the differentiation status of key ductal cells, while cell communication analysis highlighted strong interactions between ductal cells and fibroblasts.

Conclusion: This study highlights the pivotal role of SGs in PDAC progression. A novel prognostic signature based on SGRGs was developed and validated, offering substantial potential for predicting patient outcomes in PDAC.

Keywords: machine learning; pancreatic ductal adenocarcinoma; prognostic signature; single-cell RNA sequencing; stress granules.