Comprehensive molecular characterization of high-stemness gastric cancer cells using single-cell transcriptomics,...

Gastric cancer (GC) remains a global clinical challenge due to late diagnosis, high heterogeneity, and poor prognosis. Tumor stemness has emerged as a key factor driving tumor aggressiveness and therapeutic resistance. However, the systematic characterization of high-stemness GC cells and their molecular features remains limited. We integrated single-cell RNA sequencing (scRNA-seq), spatial transcriptomics, and bulk RNA-seq data to identify and characterize high-stemness GC cells. Stemness scores were calculated using CytoTRACE, and malignant cells were classified into high stemness (top 25% CytoTRACE-scored cells, HighStem), dynamic transition stemness (DTStem), and low stemness (LowStem) subpopulations based on the quartile method cutoff. ScPagwas and cell-cell communication profiling were used to explore genomic instability, genetic susceptibility, and microenvironmental interactions. HighStem-specific co-expression modules were identified via high-dimensional WGCNA (hdWGCNA), and features were screened using six machine learning algorithms. A benchmark model was constructed for HighStem prediction and interpreted using SHAP analysis. HighStem GC cells exhibited enhanced intercellular signaling, metabolic reprogramming, and stemness-related pathway activity. Five genes-APMAP, MAPRE1, GLB1, TSPAN6, and CDKN2A-were identified as robust HighStem features. Spatial and bulk transcriptomic validation confirmed their tumor-specific expression and prognostic relevance. The Support Vector Machine (SVM) model incorporating these genes achieved high accuracy (AUC = 0.973) in distinguishing HighStem cells, demonstrating strong clinical utility at the scRNA-seq level. In addition, experimental validation through knockdown of core genes (APMAP, CDKN2A, TSPAN6, MAPRE1, and GLB1) in SGC7901 and HGC-27 gastric cancer cell lines revealed a significant reduction in JAK1-STAT3 pathway activity, supporting their functional involvement in tumor stemness regulation. Furthermore, knockdown of these genes increased the sensitivity of GC cells to chemotherapeutic agents like 5-FU and cisplatin, indicating their potential role in chemoresistance. This study provides a comprehensive molecular and functional characterization of high-stemness GC cells. The identified signature genes and predictive models offer novel insights into GC stemness biology and could guide personalized therapeutic strategies. Furthermore, our findings suggest that the core genes identified in this study may serve as potential biomarkers for predicting treatment outcomes and monitoring therapeutic resistance in GC.