Abstract
Objectives: The aim of the present study was to construct a polygenic risk score (PRS) for poor survival among patients with stomach adenocarcinoma (STAD) based on expression of malignant cell markers. Methods: Integrated analyses of bulk and single-cell RNA sequencing (scRNA-seq) of STAD and normal stomach tissues were conducted to identify malignant and non-malignant markers. Analyses of the scRNA-seq profile from early STAD were used to explore intratumoral heterogeneity (ITH) of the malignant cell subpopulations. Dimension reduction, cell clustering, pseudotime, and gene set enrichment analyses were performed. The marker genes of each malignant tissue and cell clusters were screened to create a PRS using Cox regression analyses. Combined with the PRS and routine clinicopathological characteristics, a nomogram tool was generated to predict prognosis of patients with STAD. The prognostic power of the PRS was validated in two independent external datasets. Results: The malignant and non-malignant cells were identified according to 50 malignant and non-malignant cell markers. The malignant cells were divided into nine clusters with different marker genes and biological characteristics. Pseudotime analysis showed the potential differentiation trajectory of these nine malignant cell clusters and identified genes that affect cell differentiation. Ten malignant cell markers were selected to generate a PRS: RGS1, AADAC, NPC2, COL10A1, PRKCSH, RAMP1, PRR15L, TUBA1A, CXCR6, and UPP1. The PRS was associated with both overall and progression-free survival (PFS) and proved to be a prognostic factor independent of routine clinicopathological characteristics. PRS could successfully divide patients with STAD in three datasets into high- or low-risk groups. In addition, we combined PRS and the tumor clinicopathological characteristics into a nomogram tool to help predict the survival of patients with STAD. Conclusion: We revealed limited but significant intratumoral heterogeneity in STAD and proposed a malignant cell subset marker-based PRS through integrated analysis of bulk sequencing and scRNA-seq data.