Combined scRNA-seq and bulk transcriptome analysis establishes a stromal signature for predicting prognosis and therapeutic sensitivity in gastric cancer

Abstract Gastric cancer (GC) is a major disease threatening human health. Tumor stroma component, mainly including fibroblasts, pericytes and endothelial cells (ECs), play a prominent role in tumor progression. This study wanted to establish a stromal prognostic signature for GC patients. Published single-cell RNA sequencing (scRNA-seq) and bulk gene expression data were obtained from Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA). 572 stromal marker genes were identified from scRNA-seq analysis. We performed univariate cox regression and LASSO algorithm on TCGA training cohort (n=297) to construct stromal prognostic model, which were verified in validation cohorts of GSE15459 (n=182), GSE62254 (n=297) and GSE84437 (n=431). pRRophetic and Tumor Immune Dysfunction and Exclusion (TIDE) method was applied to predict chemosensitivity and immunotherapy response. Microenvironment Cell Populations-counter (MCP-counter) method was used to assess stromal and immune cell infiltration from bulk gene expression data. ScRNA-seq analysis and Immumohistochemical staining image derived from Human Protein Atlas (HPA) database was to explore gene expression location in GC tissue. We established a seven-gene prognostic signature from training cohort, which was further verified in other three validation cohorts. The seven genes (SERPINE1, SLCO2A1, GJA1, SDC2, BEX3, PLOD2, MARCKS) had different expression in GC stroma. The stromal signature could be a robust prognostic factor in GC, which provided evidence for patient stratification and stroma-targeted immunotherapeutic strategy.