Document resource
Background Immune checkpoint inhibitors (ICIs) offer significant benefits for advanced non-small cell lung cancer (NSCLC) but yield objective response rates of only 10-30% in unselected patients. Circular RNAs (circRNAs), implicated in cancer RNA dysregulation, may serve as biomarkers for ICI response.Methods Based on TCCIA database, we analyzed circRNAs expression profiles from 891 advanced NSCLC patients in the OAK and POPLAR ICI cohorts, identified circRNAs associated with the efficacy of immunotherapy in NSCLC patients. Using the selected circRNAs, we establish predictive models for immunotherapy efficacy using multiple methods and conduct performance verification. Finally, we performed Gene Set Enrichment Analysis (GSEA) using the fifty hallmark gene sets collection from The Molecular Signatures Database as a reference to compare patients stratified by high versus low model scores. Concurrently, we extracted characteristic gene sets for 22 tumor-infiltrating immune cell types (LM22) from CIBERSORT and employed the Gene Set Variation Analysis (GSVA) algorithm to calculate enrichment scores for each of the cell types.Results In result, we identified a total of 81662 distinct circRNAs in the 699 patients of the OAK cohort. Served the median expression level of each circRNA as the threshold, patients were stratified into high- and low-expression groups. Univariate analysis and LASSO regression was subsequently employed to identify the most significant circRNAs, ultimately pinpointing 11 circRNAs, circRNA-Sig. Based on circRNA-Sig, we developed prediction models using multiple distinct approaches, and the predictive performance of each model was subsequently evaluated in both internal and external validation cohorts, and corresponding survival curves were generated. Compare the performance models and select the final prediction model, which predicted atezolizumab efficacy with an area under the curve (AUC) of 0.71 in OAK and 0.67 in POPLAR. Survival analysis in OAK showed patients with low circRNA-Sig scores benefited more from ICI than chemotherapy (hazard ratio [HR] = 1.347 [1.049-1.730]; p = 0.019), whereas those with high scores showed no significant difference (HR = 1.020 [0.796-1.307]; p = 0.876). Enrichment analysis revealed that low-scoring patients exhibit an activated tumor immune microenvironment, with upregulated pathways in IFN-γ and IL-2/STAT5, which can activate immune cells such as CD8+T cells and NK cells, suggesting mechanistic links to ICI sensitivity.Conclusions In this study, we systematically characterized the differential expression profiles of circRNAs in NSCLC patients treated with ICIs and established circRNA-Sig model, validated across two cohorts, offers a novel, clinically actionable tool for stratifying NSCLC patients for ICI therapy, potentially enhancing personalized treatment strategies.