A Five Autophagy-Related Long Non-Coding RNA Prognostic Model for Patients with Lung Adenocarcinoma

Lung adenocarcinoma is the most common pathological type among non-small cell lung cancer. Although huge progress has been made in terms of early diagnosis and precision treatment in recent years, the overall 5-year survival rate of a patient remains low. In our study, we try to construct an autophagy-related lncRNA prognostic signature that may guide clinical practice.

The mRNA and lncRNA expression matrix of lung adenocarcinoma patients were retrieved from the TCGA database. Next, we constructed a co-expression network of lncRNAs and autophagy-related genes. Lasso regression and multivariate Cox regression were then applied to establish a prognostic risk model. Subsequently, a risk score was generated to differentiate the high and low risk groups and a ROC curve and nomogram to visualize the predictive ability of the current signature. Finally, gene ontology and pathway enrichment analysis were executed via GSEA.

A total of 1,703 autophagy-related lncRNAs were screened and five autophagy-related lncRNAs (LINC01137, AL691432.2, LINC01116, AL606489.1, and HLA-DQB1-AS1) were finally included in our signature. Judging from univariate (HR=1.075, 95% CI=1.046-1.104) and multivariate (HR=1.088, 95% CI=1.057-1.120) Cox regression analysis, the risk score is an independent factor for LUAD patients. Further, the AUC value based on the risk score for 1-year, 3-years, and 5-years, was 0.735, 0.672, and 0.662, respectively, indicating a reliable model. Drug sensitivity analysis revealed low risk patients were more sensitive to Gemcitabine and Gefitinib, while high risk patients had a better response to Paclitaxel and Erlotinib. Moreover, the lncRNAs included in our signature were primarily enriched in the autophagy process, metabolism, p53 pathway, and JAK/STAT pathway. Finally, a multi-omics analysis of correlated genes showed CFLAR overexpressed in the tumor sample, while GAPDH and MLST8 had a slightly higher expression in the normal sample.

Overall, our study indicated that the prognostic model we generated had certain predictability for LUAD patients' prognosis and the related genes might be potential biomarkers and therapeutic targets.