Exploring the immune landscape and immunotherapy relevance of ER stress-related lncRNAs in melanoma through multi-omics and explainable machine learning

BACKGROUND: Melanoma is a highly aggressive malignancy with poor clinical outcomes, and endoplasmic reticulum (ER) stress plays an important role in tumor progression, immune regulation, and therapeutic resistance. However, the prognostic significance and biological functions of ER stress-related long noncoding RNAs (ERlncRNAs) in melanoma remain largely unclear. METHODS: We performed an integrative analysis to systematically investigate the prognostic significance and biological relevance of ERlncRNAs in melanoma by combining bulk transcriptomic analysis, single-cell RNA sequencing, explainable machine learning, and experimental validation. Transcriptomic and clinical data were obtained from the TCGA, GTEx, and GEO databases. A 13-ERlncRNA prognostic signature was established and externally validated. Survival analysis, receiver operating characteristic analysis, and nomogram evaluation were conducted to assess model performance. SHapley Additive exPlanations (SHAP) analysis was incorporated to improve the interpretability of ERlncRNA-based risk-group classification. Immune infiltration, tumor mutational burden, TIDE scores, immune checkpoint gene expression, and predicted sensitivity to selected small-molecule targeted agents were further analyzed. Functional experiments were performed to evaluate the roles of selected ERlncRNAs in melanoma cell proliferation, colony formation, migration, and apoptosis, and Western blot analysis was used to assess the expression of ER stress-related proteins following gene silencing. RESULTS: A 13-ERlncRNA signature with stable prognostic performance was established and externally validated in melanoma cohorts. The model effectively stratified patients into high- and low-risk groups with significantly different overall survival and served as an independent prognostic predictor. SHAP analysis identified major contributors to ERlncRNA-based risk-group classification and improved model interpretability. Significant differences in immune cell infiltration, tumor mutational burden, TIDE scores, immune checkpoint gene expression, and predicted sensitivity to selected small-molecule targeted agents were observed between the two risk groups. Single-cell RNA sequencing further revealed marked heterogeneity and dynamic alterations of ER stress-related features across distinct cell populations in melanoma. Functional experiments showed that knockdown of representative ERlncRNAs inhibited melanoma cell proliferation, colony formation, and migration, while promoting apoptosis. Moreover, gene silencing was accompanied by increased CHOP expression without an obvious concomitant increase in GRP78, suggesting a shift of the ER stress response toward a pro-apoptotic state in melanoma cells. CONCLUSIONS: This study established an interpretable ERlncRNA-based prognostic framework for melanoma and highlighted the biological and clinical relevance of ER stress-related lncRNAs. These findings provide insights into tumor heterogeneity, immune-related characteristics, and potential therapeutic stratification in melanoma, and suggest that selected ERlncRNAs may contribute to melanoma progression partly through modulation of ER stress-associated apoptotic signaling.