An integrative analysis of transcriptome and single-cell RNA sequencing data identifies cytoskeleton-related prognostic genes in esophageal squamous cell carcinoma

Background Esophageal squamous cell carcinoma (ESCC) is an aggressive malignancy with dismal prognosis, primarily attributed to substantial tumor heterogeneity. Although the cytoskeleton is critically implicated in cancer progression, a comprehensive characterization of cytoskeleton-related prognostic genes in ESCC is still absent. Methods Bulk transcriptome data from public repositories were integrated with a single-cell RNA sequencing (scRNA-seq) dataset for comprehensive analysis. Cytoskeleton-related candidate genes were screened by intersecting differentially expressed genes with a curated cytoskeleton-related gene (CRG) list, and expression findings were experimentally corroborated in clinical specimens via real-time quantitative PCR (RT-qPCR). Independent prognostic genes were subsequently determined through univariate and multivariate Cox regression analyses and incorporated into a risk model, which was subsequently evaluated in two external cohorts. Additional analyses covered functional enrichment, immune infiltration profiling, drug sensitivity estimation, mutational landscape characterization, and single-cell-level investigations of inferred metabolic pathway activity, intercellular communication, and differentiation trajectories. Results A total of 222 cytoskeleton-related candidate genes were identified. Among these, three genes —GABARAPL2, PDLIM2, and PPP2R2A — were established as independent prognostic factors and incorporated into a robust risk-scoring model. This model reliably partitioned patients into high- and low-risk groups with markedly divergent survival outcomes, and a similar association between the risk score and survival was observed in two external cohorts. The high-risk group exhibited distinct estimated immune infiltration, immune checkpoint expression, somatic mutation, and predicted drug sensitivity profiles. scRNA-seq analysis characterized the distribution of the prognostic genes across cell types and supported the selection of T cells for further exploratory analyses of metabolism, intercellular communication, and differentiation trajectories. Conclusion This multi-omics integrative study identified and validated a CRG signature with potential prognostic value in ESCC and revealed its associations with immune microenvironment features and computationally predicted drug response patterns. These findings support the potential utility of the signature for risk stratification and provide testable hypotheses for further investigation of ESCC biology.