Multicenter integrative analysis identifies a multigene signature for lung adenocarcinoma prognosis via glycolysis metabolism discoordination

Abstract

Abstract Background To explore reliable and reproducible prognostic signatures to aid in guiding clinical decision-making, the present study proposed an integrative analysis method to identify a function-derived gene signature for lung adenocarcinoma (LUAD) prognosis. Methods Total 1238 LUAD patients treated with curative resection alone were sourced from public datasets. Using three cohorts of 665 patients in the meta-discovery dataset, we first utilized an integrative analysis method to extract prognostic genes, and identified the essential prognostic genes from a function-derived perspective. Thereafter, we proposed pairwise comparison of single-sample gene set enrichment method to establish a multigene signature for LUAD prognosis based on the function-derived prognostic genes. Results Based on integrative analysis, we identified 14 metabolic-related prognostic genes involved in glycolysis metabolism, and established a function-derived signature consisting of these genes for LUAD prognosis (14GM-PS). The prognostic performance of the signature was rigorously validated in two multiple cross-platform independent datasets comprising 299 (log-rank P = 2.78E-06) and 274 (log-rank P = 0.0042) patients, respectively, with significantly different 5-year survival rate. Multivariate Cox analysis demonstrated that the function-derived signature was an independent prognostic factor for LUAD prognosis. Furthermore, the novel proposed nomogram significantly improved the prognostic performance (concordance index) of clinicopathological factors. The molecular and immune characteristics analysis showed that high-risk patients identified by 14GM-PS were characterized by higher hypoxia, proliferation and stemness scores, and lower immune score, providing evidence that could reflect transcriptomic characteristics that are strongly associated with clinical outcomes in the molecular mechanism. Conclusion This multicenter study illustrates the accuracy and stability of the function-derived signature for LUAD prognosis, and might become a promising genomic tool to guide individualized application and decision-making of LUAD in clinical practice, with further prospective validation in clinical trials.