Establishing a glycolysis-linked multigene prognostic signature in lung adenocarcinoma: a multicenter integrative approach.

Background: Lung adenocarcinoma (LUAD) is the most common type of lung cancer, and has a high mortality rate. Surgical resection is the standard treatment for early stage patients; however, many patients experience recurrence or death within 5 years. Thus, reliable prognostic markers urgently need to be identified to guide clinical decisions and enhance patient outcomes. While molecular profiling has uncovered significant molecular irregularities in LUAD, existing prognostic signatures exhibit limited reproducibility and inconsistent performance across datasets. These limitations are mainly attributed to small sample sizes, technical biases, and inadequate biological context. This study aims to develop a function-derived gene signature for LUAD prognosis using a multicenter integrative analysis approach. Methods: We used gene expression and clinical data from 1,238 curatively resected LUAD patients across 11 public datasets. A meta-discovery dataset comprising 665 patients was analyzed to extract prognostic genes from a functional perspective. Using a single-sample gene set enrichment method in pairwise comparisons, we established a multigene signature. Specifically, we identified 14 glycolysis-related prognostic genes, which were then included in the 14-gene glycolysis metabolism prognostic signature (14GM-PS). Results: The 14GM-PS prognostic performance was validated in two independent datasets (n=299 and n=274), both demonstrating significant differences in 5-year survival rates (log-rank P<0.001 and P=0.004, respectively). The multivariate Cox analysis results confirmed that the 14GM-PS could be used to predict patient prognosis. Further, a novel nomogram incorporating the 14GM-PS and clinicopathological factors showed improved prognostic accuracy. The patients identified as high risk by the 14GM-PS had higher hypoxia, proliferation, and stemness scores, but lower immune scores, indicating a strong association between transcriptomic characteristics and clinical outcomes. The 14GM-PS showed consistent performance across different platforms and validation cohorts. Conclusions: This multicenter study showed the accuracy and stability of the 14GM-PS in predicting the prognosis of LUAD patients. This promising genomic tool could be used to guide individualized treatment decisions in LUAD patients. However, further prospective clinical validation is required before its clinical application.