Large-scale gene expression analysis reveals robust gene signatures for prognosis prediction in lung adenocarcinoma

Lung adenocarcinoma (LUAD) is the leading cause of cancer-related death worldwide. High mortality in LUAD motivates us to stratify the patients into high- and low-risk groups, which is beneficial for the clinicians to design a personalized therapeutic regimen. To robustly predict the risk, we identified a set of robust prognostic gene signatures and critical pathways based on ten gene expression datasets by the meta-analysis-based Cox regression model, 25 of which were selected as predictors of multivariable Cox regression model by MMPC algorithm. Gene set enrichment analysis (GSEA) identified the Aurora-A pathway, the Aurora-B pathway, and the FOXM1 transcription factor network as prognostic pathways in LUAD. Moreover, the three prognostic pathways were also the biological processes of G2-M transition, suggesting that hyperactive G2-M transition in cell cycle was an indicator of poor prognosis in LUAD. The validation in the independent datasets suggested that overall survival differences were observed not only in all LUAD patients, but also in those with a specific TNM stage, gender, and age group. The comprehensive analysis demonstrated that prognostic signatures and the prognostic model by the large-scale gene expression analysis were more robust than models built by single data based gene signatures in LUAD overall survival prediction.