Abstract

Background Esophageal squamous cell cancer (ESCC) is a malignancy derived from the Esophagus, and dysregulation of the cGAS‐STING pathway contributes to ESCC progression. Method ESCC bulk‐seq dataset GSE38129 was acquired from GEO database and then underwent Limma and WGCNA analysis for the identification of shared DEGs, which were intersected with cGAS‐STING pathway gene list and underwent Cox regression analysis for recognized cGAS‐STING associated prognostic indicators. Next, Consensus clustering and machine learning combinations (Lasso + SurvivalSVM) were utilized for cGAS‐STING associated ESCC molecular subgroups and prognostic model construction in TCGA‐ESCC cohorts, and prognostic performance was validated in GSE53662 . Besides, hub prognostic variables were acquired from Lasso‐Cox regression, and their molecular and immune features were estimated via multiple bioinformatic approaches at TCGA‐ESCC cohort. In addition, heterogeneity of hub genes at single‐cell level for ESCC patients was also indicated in GSE188900 in spatial and temporal manners. Furthermore, Drug sensitivity and molecular docking analysis were performed for identification of optimal therapeutic agents targeting hub genes. Indeed, in vitro assays have been performed to assess the oncogenic potential of hub genes and efficacy of optimal therapeutic agents. Furthermore, implications of hub genes with cGAS‐STING pathway were estimated in single‐cell artificial intelligence (AI) driven‐virtual cell and bulk assays. Results By utilizing integrative AI and multi‐omic pipelines, we proved that the cGAS‐STING pathway can guide subgroup stratification and prognostic model construction for ESCC patients. PRKDC and SLC25A13 can be considered hub genes associated with ESCC pathogenesis and regulation of the cGAS‐STING pathway. BX‐912 and Navitoclax can be considered drug screening strategies for the treatment of ESCC patients by targeting PRKDC and SLC25A13. Conclusion cGAS‐STING pathway can gu