Abstract
Esophageal squamous cell carcinoma (ESCC) exhibits substantial molecular heterogeneity and unfavorable clinical outcomes. Current transcriptomic advances are shifting the focus from static gene expression profiles to the dynamic architecture of gene interaction networks. However, gene interaction perturbation signatures specific to ESCC remain poorly understood. This study aimed to develop a network-informed prognostic index derived from malignant epithelial cell signatures. In-house single-cell RNA sequencing data from 15 ESCC samples from the First Affiliated Hospital of Zhengzhou University were analyzed to identify dysregulated genes in malignant squamous epithelial cells. Then, a gene interaction perturbation network index (GIPNI) was constructed by systematically evaluating 75 combinations of machine-learning methods and validated across 3 independent cohorts. Associations between the GIPNI and genomic alterations, immune-related characteristics, and therapeutic response were also evaluated. Results showed ESCCs with high-GIPNI scores were associated with advanced clinicopathological features and overactivated mitotic cell cycle and epithelial cell differentiation pathways. Immune profiling suggested that low-GIPNI tumors had a more immune-infiltrated microenvironment. Notably, high-GIPNI ESCCs were associated with higher sensitivity to some common chemotherapeutic agents. Overall, the GIPNI provides a network-informed and malignant squamous cell-oriented framework for prognostic assessment in ESCC. This integrative approach may facilitate risk stratification and provide insights into individualized therapeutic strategies.