Background: Lung adenocarcinoma (LUAD) is one of the most common cancer types, threatening the human health around the world. However, the high heterogeneity and complexity of LUAD limit the benefits of targeted therapies. This study aimed to identify the key prognosis impacting genes and relevant subtypes for LUAD.
Methods: We recognized significant mutations and prognosis-relevant genes based on the omics data of 515 LUAD samples from The Cancer Genome Atlas. Mutation significance was estimated by MutSigCV. Prognosis analysis was based on the cox proportional hazards regression (Coxph) model. Specifically, the Coxph model was combined with a causal regulatory network to help reveal which genes play master roles among numerous prognosis impacting genes. Based on expressional profiles of the master genes, LUAD patients were clustered into different sub-types by a consensus clustering method and the importance of master genes were further evaluated by random forest.
Results: Significant mutations did not influence the prognosis directly. However, a collection of prognosis relevant genes were recognized, where 75 genes like GAPDH and GGA2 which are involved in mTOR signaling, lysosome or other key pathways are further identified as the master ones. Interestingly, the master gene expressions help separate LUAD patients into two sub-types displaying remarkable differences in expressional profiles, prognostic outcomes and genomic mutations in certain genes, like SMARCA4 and COL11A1. Meanwhile, the subtypes were re-discovered from two additional LUAD cohorts based on the top-10 important master genes.
Conclusions: This study can promote precision treatment of LUAD by providing a comprehensive description on the key prognosis-relevant genes and an alternative way to classify LUAD subtypes.
Keywords: Causal regulatory network; Lung adenocarcinoma; Prognosis; Prognosis master regulator; Subtypes.