Gene expression-based clinical predictions in lung adenocarcinoma

Abstract

Mining disease-related genes contributes momentously to handling lung adenocarcinoma (LUAD). But genetic complexity and tumor heterogeneity severely get in the way. Fortunately, new light has been shed by dramatic progress of bioinformatic technology in the past decades. In this research, we investigated relationships between gene expression and clinical features of LUAD via integrative bioinformatic analysis. First, we applied limma and DESeq2 packages to analyze differentially expressed genes (DEGs) of LUAD from GEO database and TCGA project (tumor tissues versus normal tissues), and acquired 180 down-regulated DEGs and 52 up-regulated DEGs. Then, we investigated genetic and biological assignment of theses DEGs by Bioconductor packages and STRING database. We found these DEGs were distributed dispersedly among chromosomes, enriched observably in extracellular matrix-related processes, and weighted hierarchically in interaction network. Finally, we established DEGs-based statistical models for evaluating TNM stage and survival status of LUAD. And these models (logistic regression models for TNM parameter and Cox regression models for survival probability) all possessed fine predictive efficacy (C-indexes: T, 0.740; N, 0.687; M, 0.823; overall survival, 0.678; progression-free survival, 0.611). In summary, we have successfully established gene expression-based models for assessing clinical characteristics of LUAD, which will assist its pathogenesis investigation and clinical intervention.