Integrative learning in developing an immunologic lncRNA signature as a consensus risk-stratification tool for lung adenocarcinoma

[摘要] Background: In the tumor immune microenvironment, the contribution of innate and adaptive immune cells to tumor progression has been consistently demonstrated. However, reliable prognostic biomarkers for lung adenocarcinoma (LUAD) have not yet been identified. We thus developed and validated an immunologic long noncoding RNA (lncRNA) signature (ILLS) to facilitate the classification of patients with high and low risk and provide potential “made-to-measure” treatment choices. Methods: The LUAD data sets were obtained and processed from public databases of The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO). The abundance of immune infiltration and its related pathways were calculated through consensus clustering, weighted gene coexpression network analysis (WGCNA), and an integrated ImmLnc to identify immune-related lncRNAs and extract immune-related prognostic lncRNAs. Based on the integrative procedure, the best algorithm composition was least absolute shrinkage and selection operator (LASSO) and stepwise Cox regression in both directions to develop the ILLS in the TCGA-LUAD data set and validate the predictive power of 4 independent data sets, GSE31210, GSE37745, GSE30219, and GSE50081 through survival analysis, receiver operating characteristic (ROC) analysis, and multivariate Cox regression. The concordance index (C-index) analysis was transversely compared with 49 published signatures in the above 5 data sets to further confirm its stability and superiority. Finally, drug sensitivity analysis was conducted to explore potential therapeutic agents. Results: Patients from the high-risk groups consistently had worse overall survival (OS) compared to the low-risk groups. ILLS proved to be an independent prognostic factor with favorable sensitivity and specificity. Among the 4 GEO data sets, compared to those reported in the other literature, ILLS maintained stable prediction ability and was more suitable as a consensus risk-stratification tool. However, The Cancer Immunome Atlas and IMvigor210 data sets demonstrated practical utility in recognizing target populations with effective immunotherapy, while the high-risk group exhibited potential targets for certain chemotherapy drugs, such as carmustine, etoposide, arsenic trioxide, and alectinib. Conclusions: ILLS demonstrated superior and stable prognostic prediction ability and thus has potential as a tool for assisting in risk classification and clinical decision-making in patients with LUAD.

[学科分类] 呼吸医学

[关键词] Machine learning;long noncoding RNA (lncRNA);immune;lung adenocarcinoma (LUAD);prognosis