Description: Purpose: The National Lung Cancer Screening Trial has confirmed that lung cancer mortality can be reduced if tumors are diagnosed early, that is, at stage I. However, a substantial fraction of stage I lung cancer patients still develop metastatic disease within 5 years from surgery. Prognostic biomarkers are therefore needed to identify patients at risk of an adverse outcome, who might benefit from multimodality treatment. Experimental Design: We extensively validated a 10-gene prognostic signature in a cohort of 507 lung adenocarcinoma patients using formalin-fixed paraffin-embedded samples. Furthermore, we performed an integrated analysis of gene expression, methylation, somatic mutations, copy number variations, and proteomic profiles on an independent cohort of 468 patients from The Cancer Genome Atlas (TCGA). Results: Stage I lung cancer patients ( N = 351) identified as high-risk by the 10-gene signature displayed a 4-fold increased risk of death [HR = 3.98; 95% confidence interval (CI), 1.73–9.14], with a 3-year overall survival of 84.2% (95% CI, 78.7–89.7) compared with 95.6% (92.4–98.8) in low-risk patients. The analysis of TCGA cohort revealed that the 10-gene signature identifies a subgroup of stage I lung adenocarcinomas displaying distinct molecular characteristics and associated with aggressive behavior and poor outcome. Conclusions: We validated a 10-gene prognostic signature capable of identifying a molecular subtype of stage I lung adenocarcinoma with characteristics remarkably similar to those of advanced lung cancer. We propose that our signature might aid the identification of stage I patients who would benefit from multimodality treatment. Clin Cancer Res; 23(1); 62–72. ©2016 AACR .