An eight-miRNA signature as a potential biomarker for predicting survival in lung adenocarcinoma

Abstract

Background: Lung adenocarcinoma is a heterogernous disease that creates challenges for classification and management. The purpose of this study is to identify specific miRNA markers closely associated with the survival of LUAD patients from a large dataset of significantly altered miRNAs, and to assess the prognostic value of this miRNA expression profile for OS in patients with LUAD.

Methods: We obtained miRNA expression profiles and corresponding clinical information for 372 LUAD patients from The Cancer Genome Atlas (TCGA), and identified the most significantly altered miRNAs between tumor and normal samples. Using survival analysis and supervised principal components method, we identified an eight-miRNA signature for the prediction of overall survival (OS) of LUAD patients. The relationship between OS and the identified miRNA signature was self-validated in the TCGA cohort (randomly classified into two subgroups: n = 186 for the training set and n = 186 for the testing set). Survival receiver operating characteristic (ROC) analysis was used to assess the performance of survival prediction. The biological relevance of putative miRNA targets was also analyzed using bioinformatics.

Results: Sixteen of the 111 most significantly altered miRNAs were associated with OS across different clinical subclasses of the TCGA-derived LUAD cohort. A linear prognostic model of eight miRNAs (miR-31, miR-196b, miR-766, miR-519a-1, miR-375, miR-187, miR-331 and miR-101-1) was constructed and weighted by the importance scores from the supervised principal component method to divide patients into high- and low-risk groups. Patients assigned to the high-risk group exhibited poor OS compared with patients in the low-risk group (hazard ratio [HR] = 1.99, P <0.001). The eight-miRNA signature is an independent prognostic marker of OS of LUAD patients and demonstrates good performance for predicting 5-year OS (Area Under the respective ROC Curves [AUC] = 0.626, P = 0.003), especially for non-smokers (AUC = 0.686, P = 0.023).

Conclusions: We identified an eight-miRNA signature that is prognostic of LUAD. The miRNA signature, if validated in other prospective studies, may have important implications in clinical practice, in particular identifying a subgroup of patients with LUAD who are at high risk of mortality.

Figure 1

MiRNAs associated with prognosis in different clinical subclasses of TCGA lung adenocarcinoma cohort. The matrix visualizes the significant HRs for the 16 miRNAs in the TCGA LUAD cohort. Numbers in the rectangles indicate the HRs for expression with significant univariate Cox regression (P < 0.1). Red rectangles indicate HRs >1 and blue rectangles indicate HRs <1.

Figure 2

MicroRNA predictor-score analysis of 372 LUAD patients in TCGA cohort. (A) MicroRNA predictor-score distribution. (B) Patients’ survival status. (C) Color-gram of miRNA expression profiles of LUAD patients. The blue dotted line represents the median miRNA signature cutoff dividing patients into low-risk and high-risk groups.

Figure 3

Kaplan–Meier and ROC curves for the 8-miRNA signature in TCGA LUAD cohort. (A) The Kaplan–Meier curves for LUAD risk groups obtained from the TCGA cohort (n = 372) divided by the median cutoff point. Patients with high scores had poor outcome in terms of OS (Median OS: 39.0 months vs. 59.3 months, P < 0.001). (B) The ROC curve had an AUC of 0.626 (P = 0.003). The permutation P value was obtained from 1,000 permutations for testing the null hypothesis (AUC = 0.5).

Figure 4

Kaplan–Meier and ROC curves for the 8-miRNA signature in TCGA non-smoking LUAD cohort. (A) The Kaplan–Meier curves for LUAD risk groups obtained from the TCGA non-smoking LUAD cohort (n = 145) divided by the median cutoff point. Non-smoking patients with high scores had poor outcome in terms of OS (Median OS: 46.0 months vs. 59.3months, P = 0.027). (B) The ROC curve had an AUC of 0.686 (P = 0.023). The permutation P value was obtained from 1,000 permutations for testing the null hypothesis (AUC = 0.5).