Comprehensive analysis of gene signature linking hypoxia and lactylation for predicting prognosis and immunotherapy response in patients with gastric cancer

Background: Gastric cancer (GC) is the fifth most commonly diagnosed cancer worldwide and ranks as the third leading cause of cancer-related deaths. In the tumor microenvironment (TME), malignant cells undergo hypoxic adaptation via metabolic reprogramming, which promotes cellular proliferation and lactate accumulation. However, the significance of hypoxia- and lactylation-related genes (HALRGs) in GC remains unclear. This study aimed to investigate the impact of hypoxia and lactylation on GC and to construct relevant prognostic models. Methods: We downloaded single-cell RNA sequencing (scRNA-seq), spatial transcriptome (ST) data, and clinical data. The least absolute shrinkage and selection operator (LASSO) algorithm was used to develop a prognostic model, which was validated in an external cohort. Immune cell infiltration was analyzed using the CIBERSORT algorithm, while the Tumor Immune Dysfunction and Exclusion (TIDE) score and ESTIMATE method were combined to evaluate the prognostic potential of immunotherapy response. Results: We analyzed 23,447 genes and 104,150 cells from scRNA-seq, classifying the cells into seven distinct types. The spatial distribution characteristics of these cell types were further examined using ST data. Inference of copy number variations (InferCNV) analysis was performed on epithelial cells to differentiate between malignant and normal cells, revealing that normal epithelial cells could be further categorized into five subtypes. We then developed a prognostic model based on HALRGs using The Cancer Genome Atlas (TCGA) data, which was validated in the Gene Expression Omnibus (GEO) dataset. Patients were divided into high- and low-risk groups according to HALRG scores. Interestingly, the high-risk group exhibited higher TIDE scores, suggesting a poorer response to immunotherapy. Conclusions: This study uncovered the cellular heterogeneity of GC by integrating scRNA-seq and ST data. The results demonstrated that the HALRG risk model may be effective in predicting the clinical prognosis and has the potential to guide personalized treatment in GC.