A novel signature predicts prognosis in pancreatic cancer based on tumor membrane-associated genes.

Abstract

Background: Pancreatic cancer is one of the leading causes of tumor-related mortality, characterized by short patient survival times and limited treatment options. Some targeted therapies have not succeeded in improving patient prognosis. Tumor membranes possess potential target specificity, offering hope for enhancing the efficacy of immunotherapy and drug treatment.

Methods: In this study, we collected gene expression and survival data from two scRNA-seq projects and patient cohorts in TCGA, ICGC, and GEO. Differential analysis and dimensionality reduction clustering were employed to isolate tumor epithelial cells. High-expression membrane-associated genes in tumor epithelial cells were identified through PPI network analysis and functional enrichment. Subsequently, membrane-associated genes associated with patient prognosis were selected using LASSO and Cox regression to construct MaGPS, which was validated in external datasets. Potential therapeutic targets of the MaGPS signatures were identified and confirmed by integrating spatial transcriptomics, scRNA-seq, and protein expressions. In addition, drug sensitivity analysis was performed to explore potential targeted drugs associated with MaGPS.

Results: The results demonstrated the identification of a specific tumor epithelial cell cluster, c0. This cluster expressed 17 membrane-associated genes that are closely interconnected and play roles in extracellular interactions. The MaGPS model, developed based on the membrane-associated genes LMO7, APOL1, SLC2A1, C15orf48, FXYD3, and CLDN18, effectively predicted patient prognostic risk. Additionally, the expression of the six MaGPS signatures was observed to be elevated in tumors at both the protein expression and spatial transcriptomics levels. Furthermore, drug sensitivity analysis revealed that the MaGPS signature scores were significantly associated with the sensitivity to 38 different drugs, highlighting potential targeted therapies related to MaGPS.

Conclusion: The MaGPS model, based on bulk RNA-seq, scRNA-seq, and spatial transcriptomics data, effectively evaluated the prognosis of pancreatic cancer and provided valuable insights for better therapeutic targets.