Sage Journals: Discover world-class research

Abstract

BACKGROUND:

Lung squamous cell carcinoma (LUSC) is malignant disease with poor therapeutic response and unfavourable prognosis.

OBJECTIVE:

This study aims to develop a long non-coding RNA (lncRNA) signature for survival prediction in patients with LUSC.

METHODS:

We obtained lncRNA expression profiles of 493 LUSC cases from The Cancer Genome Atlas, and randomly divided the samples into a training set ( $n=$ 296) and a testing set ( $n=$ 197). Univariate Cox regression and random survival forest algorithm were performed to select optimum survival-related lncRNAs.

RESULTS:

A lncRNA-focused risk score model was then constructed for prognosis prediction in the training set and further validated in the testing set and the entire set. Finally, bioinformatics analysis was carried out to explore the potential signaling pathways associated with the prognostic lncRNAs. A set of 9 lncRNAs were found to be strongly correlated with overall survival of LUSC patients. These 9 lncRNAs were integrated into a prognostic signature, which could separate patients into high- and low-risk groups with significantly different survival times in the training set (median: 30.5 vs. 80.5 months, log-rank $P<$ 0.001). This signature was also confirmed in the testing set and the entire set. Besides, the prognostic value of the 9-lncRNA signature was independent of clinical features and maintained stable in stratified analyses. Functional enrichment study suggested that the 9 lncRNAs may be mainly involved in metabolism-related pathways, phosphatidylinositol signaling system, p53 signaling pathway, and notch signaling pathway.

CONCLUSIONS:

Our study demonstrated the potential clinical implication of the 9-lncRNA signature for survival prediction of LUSC patients.

Keywords

Biomarker long non-coding RNA lung squamous cell carcinoma prognosis

1. Introduction

Lung cancer is one of the most commonly diagnosed cancers around the world, contributing to over 18% of all cancer-related deaths in both male and female [1]. The major pathological type of lung cancer was non-small cell lung cancer, which mainly includes lung adenocarcinoma and lung squamous cell carcinoma (LUSC). Patients with LUSC are often diagnosed in advanced stage when currently available therapy cannot be timely administered [2]. To the treatment by chemotherapy or radiotherapy, LUSC patients are also not as sensitive as small cell cancer patients. In addition, the prognosis of LUSC is poor, with an estimated 5-year survival rate of $<$ 15% [3]. So, there is an imperative need to identify effective prognostic biomarkers for patients with LUSC.

It is generally accepted that protein-coding genes (mRNAs) only constitute a small fraction of human genome, while more than 98% of the genome may be processed into non-coding RNAs (ncRNAs) [4]. Long non-coding RNAs (lncRNAs) are a newly discovered major subtype of ncRNAs with $>$ 200 nucleotides in length [5]. Accelerating evidence showed that lncRNAs can function importantly in a wide range of biological processes through multiple mechanisms, such as chromatin modification, transcriptional modulation, and post-transcriptional regulation [6, 7, 8]. Aberrant lncRNAs have also been frequently detected in many cancers and associated with tumorigenesis and tumor progression. Some well-described lncRNAs, such as HOTAIR, MALAT1 and NEAT1 were identified to obviously up-regulate in breast carcinoma, gastric cancer and hepatocellular carcinoma [9, 10, 11], whereas GAS5, ZFAS1 and CASC2 might serve as tumor suppressors [12, 13, 14]. Recent studies have also demonstrated the emerging role of lncRNAs in LUSC. For example, lncRNA SNHG1 was reported to facilitate cell metastasis and invasion in LUSC by regulating ZEB1 signaling pathway [15]; and, SFTA1P could promote tumor apoptosis and increase cisplatin chemosensitivity in LUSC by modulating the hnRNP-U-GADD45A axis [16]. However, the prognostic significance of lncRNAs in LUSC remains under-investigated and needs to be systematically evaluated.

In this study, by integrating lncRNA expression profiles and related clinical information of LUSC patients from The Cancer Genome Atlas database (TCGA, https://cancergenome.nih.gov), we obtained a set of 9 prognostic lncRNA biomarkers associated with overall survival (OS) of patients with LUSC. These lncRNAs were then combined to construct a 9-lncRNA risk score model that could effectively predict patient’s OS. The significant prognostic power of the 9-lncRNA signature was further confirmed in the testing dataset and the whole set.

2. Materials and methods

2.1 LUSC patient data

The level-3 RNA sequencing data (HTSeq-FPKM) and relevant clinical details of TCGA-LUSC cohort were downloaded from TCGA database. After removal of LUSC samples with incomplete survival data and normal samples, a total of 493 patients with LUSC were enrolled in this study. LncRNAs and mRNAs were recognized through mapping gene identification to the annotations from the GENCODE (https://www. gencodegenes.org, release 22). The expression profiles of lncRNAs were normalized with log ${}_{2}$ (x $+$ 1) transformation. Those lncRNAs with zero expression value in more than 10% samples were discarded, resulting in 4494 lncRNAs for subsequent analysis. The whole dataset was then randomly divided into a training set ( $n=$ 296) and a testing set ( $n=$ 197). The clinical characteristics of patients in the datasets were summarized in Table 1.

Table 1
Clinical characteristics of patients with LUSC

Variables	Training set	Testing set	Entire set
	( $n=$ 296)	( $n=$ 197)	( $n=$ 493)
Age, $n$ (%)
$<$ 65 yrs	101 (34.6)	68 (34.7)	169 (34.6)
$\geqslant$ 65 yrs	191 (65.4)	128 (65.3)	319 (65.4)
Gender, $n$ (%)
Female	79 (26.7)	49 (24.9)	128 (26.0)
Male	217 (73.3)	148 (75.1)	365 (74.0)
Current smoker, $n$ (%)
No	209 (72.3)	140 (72.9)	349 (72.6)
Yes	80 (27.7)	52 (27.1)	132 (27.4)
Tumor stage, $n$ (%)
I/II	247 (84.0)	152 (77.9)	399 (81.6)
III/IV	47 (16.0)	43 (22.1)	90 (18.4)
Residual tumor, $n$ (%)
R0	242 (94.2)	151 (87.3)	393 (91.4)
R1/2/x	15 (5.8)	22 (12.7)	37 (8.6)

LUSC, lung squamous cell carcinoma.

2.2 Construction of a prognostic lncRNA signature

The association between lncRNA expression and patient’s OS was evaluated using univariate Cox regression analysis. With a parametric test ( $P<$ 0.01), we obtained a set of candidate lncRNAs that could strongly predict patient’s OS. To minimize the number of genes in the predictive model, we carried out the random survival forests variable hunting (RSFVH) algorithm using R “randomForestSRC” package [17]. We followed previous reports for parameter setting in this algorithm [18]. Briefly, the number of Monte Carlo iterations (n ${}_{\text{rep}}$ ) was set to be n ${}_{\text{rep}}=$ 100, and the step size used in the forward process (n ${}_{\text{step}}$ ) was set as n ${}_{\text{step}}=$ 5. Then, the lncRNAs selected by the RSFVH algorithm were subjected to multivariate Cox regression analysis, and a risk score was developed by linear combination of the expression levels of selected lncRNAs weighted by their Cox regression coefficients.

Table 2
The nine lncRNAs significantly associated with overall survival in the training set

Ensemble id	Gene symbol	Genomic location	HR ${}^{\text{a}}$	$P$ -value ${}^{\text{a}}$	Coefficient ${}^{\text{a}}$
ENSG00000268970.1	CTD-3099C6.11	Chr19: 52,597,699–52,598,887 ( $-$ )	1.500	9.02E-03	0.405
ENSG00000272459.1	RP11-1277A3.3	Chr5: 177,554,824–177,555,364 ( $+$ )	0.362	5.75E-03	$-$ 1.017
ENSG00000225548.4	AC098973.2	Chr3: 27,802,762–27,891,301 ( $+$ )	0.809	1.57E-03	$-$ 0.212
ENSG00000226330.1	RP11-739N20.2	Chr1: 204,377,850–204,435,846 ( $+$ )	2.085	1.01E-03	0.735
ENSG00000249001.4	RP11-742B18.1	Chr4: 87,568,035–87,733,956 ( $-$ )	0.696	8.24E-03	$-$ 0.362
ENSG00000272010.1	CTD-3025N20.3	Chr8: 65,591,850–65,592,472 ( $-$ )	0.551	1.56E-03	$-$ 0.596
ENSG00000229167.1	RP11-73M7.1	Chr1: 31,571,585–31,575,573 ( $-$ )	2.000	2.73E-03	0.693
ENSG00000237813.3	AC002066.1	Chr7: 116,238,260–116,499,465 ( $-$ )	4.296	2.40E-03	1.458
ENSG00000224848.1	RP11-535M15.1	Chr9: 96687056–96721121 ( $+$ )	1.438	5.17E-04	0.363

${}^{\text{a}}$ Derived from univariate Cox regression model. HR, hazard ratio.

Figure 1.

Random survival forest algorithm for lncRNA selection. A, error rate for the data as a function of trees; B, out-of-bag importance values for lncRNA predictors.

Figure 2.

Identification and evaluation of the 9-lncRNA signature for OS prediction of LUSC patients. A, the five-year ROC curve for OS prediction by the 9-lncRNA signature in the training set. B–D, Kaplan-Meier analysis for OS prediction by the 9-lncRNA signature in the training, the testing, and the entire sets.

2.3 Statistical methods

To assess the prognostic performance of the lncRNA risk score signature, we conducted time-dependent receiver operating characteristic (ROC) analyses using R “survivalROC” package [19]. LUSC Patients were then assigned into high- and low-risk groups based on the optimal cut-off value of risk score from the 5-year ROC curve. Survival analysis was performed using Kaplan-Meier method, and log-rank test was used to evaluate the OS difference between high- and low-risk groups. Univariate and multivariate Cox regression analyses were implemented to examine whether the prognostic value of lncRNA signature was independent of clinical features. The statistical significance was based on $P<$ 0.05.

2.4 Functional enrichment analysis

Pearson correlation coefficients were calculated to identify the co-expression relationship between lncRNAs and mRNAs. To reduce the potential of false positives, we selected only the lncRNA-mRNA pairs with correlation coefficients greater than 0.35. Kyoto encyclopedia of genes and genomes (KEGG) pathway for the mRNAs co-expressed with the prognostic lncRNAs were explored using KOBAS web server (version 3.0, http://kobas.cbi.pku.edu.cn) [20]. The pathways with $P<$ 0.05 were considered as significantly enriched.

3. Results

3.1 Identification of a prognostic lncRNA signature in the training set

A flowchart for data processing and analysis is shown in Supplementary Fig. 1. By using univariate Cox regression, we identified a total of 43 candidate lncRNAs that were strongly correlated with patients’ OS in the training set. Based on the RSFVH algorithm, 9 lncRNAs were further selected as prognostic biomarkers. Table 2 showed the genomic details of these 9 lncRNAs with their statistics from univariate Cox regression. In Fig. 1, we can see that AC002066.1 had the greatest importance value than other predictors. Besides, the positive coefficients of 5 lncRNAs indicated that their higher expressions were correlated with shorter survival, and the negative coefficients of the other 4 lncRNAs suggested that their higher expressions were associated with longer survival. By linear combination of the expression levels of the 9 lncRNAs weighted by their Cox regression coefficients, a 9-lncRNA risk-score signature was constructed as follows: risk score $=$ (0.4610 $\times$ CTD-3099C6.11) $+$ ( $-$ 1.0418 $\times$ RP11-1277A3.3) $+$ ( $-$ 0.1145 $\times$ AC098973.2) $+$ (0.3835 $\times$ RP11-739N20.2) $+$ ( $-$ 0.0373 $\times$ RP11-742B18.1) $+$ ( $-$ 0.6747 $\times$ CTD-3025N20.3) $+$ (0.5935 $\times$ RP11-73M7.1) $+$ (0.7215 $\times$ AC002066.1) $+$ (0.1502 $\times$ RP11-535M15.1). To assess the prognostic performance of the risk score, a 5-year ROC curve analysis was performed. As depicted in Fig. 2A, the area under the curve (AUC) was 0.757, indicating a good prognostic ability of the 9-lncRNA signature in the training set. The patients were then recruited into high- ( $n=$ 69) and low-risk groups ( $n=$ 135) according to the best cut-off value (0.877) derived from the 5-year ROC curve. LUSC patients in the high-risk group had an obviously shorter OS than those in the low-risk group (median: 30.5 vs. 80.5 months, log-rank $P<$ 0.001; Fig. 2B).

Table 3
Univariate and multivariate Cox regression analysis for overall survival

Variables	HR	95% CI	$P$ -value	HR	95% CI	$P$ -value
	Univariate Cox regression			Multivariate Cox regression
Training set ( $n=$ 296)
Age	1.019	0.998–1.041	0.074	1.041	1.014–1.067	2.31E-03
Gender (male vs. female)	1.540	1.006–2.358	4.69E-02	1.818	1.120–2.952	1.57E-02
Current smoker (yes vs. no)	1.557	1.073–2.258	1.96E-02	1.691	1.119–2.556	1.26E-02
Tumor stage (III/IV vs. I/II)	1.383	0.898–2.219	0.141	1.506	0.935–2.424	0.092
Residual tumor (R1/2/x vs. R0)	1.848	0.852–4.007	0.120	1.470	0.635–3.403	0.369
9-lncRNA risk score (high vs. low)	2.887	1.939–4.298	1.76E-07	3.020	1.935–4.715	1.15E-06
Testing set ( $n=$ 197)
Age	1.015	0.988–1.043	0.276	1.030	0.999–1.063	0.058
Gender (male vs. female)	0.819	0.495–1.356	0.438	0.754	0.429–1.327	0.327
Current smoker (yes vs. no)	1.376	0.858–2.207	0.186	1.841	1.098–3.086	2.06E-02
Tumor stage (III/IV vs. I/II)	2.061	1.261–3.371	3.94E-03	2.134	1.209–3.767	8.96E-03
Residual tumor (R1/2/x vs. R0)	1.910	0.936–3.897	0.076	1.682	0.750–3.768	0.207
9-lncRNA risk score (high vs. low)	1.965	1.204–3.206	6.88E-03	2.005	1.166–3.448	1.19E-02
Entire set ( $n=$ 493)
Age	1.016	0.999–1.033	0.056	1.035	1.015–1.055	4.71E-04
Gender (male vs. female)	1.209	0.875–1.671	0.250	1.303	0.911–1.863	0.148
Current smoker (yes vs. no)	1.488	1.112–1.992	7.49E-03	1.699	1.238–2.332	1.03E-03
Tumor stage (III/IV vs. I/II)	1.590	1.153–2.192	4.68E-03	1.630	1.154–2.301	5.51E-03
Residual tumor (R1/2/x vs. R0)	1.905	1.133–3.204	1.51E-02	1.813	1.059–3.105	3.01E-02
9-lncRNA risk score (high vs. low)	2.285	1.694–3.082	6.30E-08	2.344	1.692–3.245	2.94E-07

CI, confidence interval; HR, hazard ratio.

Figure 3.

Kaplan-Meier analysis for OS prediction of LUSC using the 9-lncRNA signature in the subanalyses by age (A and B), smoking status (C and D), and tumor stage (E and F).

3.2 Evaluation of the lncRNA signature in the testing set and the entire set

To assess the robustness of the 9-lncRNA signature in OS prediction of LUSC patients, we further examined it in the testing set and the entire set. The patients were divided into high- and low-risk groups using the same risk score model and cut-off point derived from the training set. The AUC of 5-year ROC curve in the testing and the whole sets was 0.613 and 0.705, respectively, implying a good prognostic capacity of the 9-lncRNA signature. Accordingly, patients in the low-risk group ( $n=$ 74) exhibited a longer OS as compared to those in the high-risk group ( $n=$ 123) (median: 76.1 vs. 38.7 months, log-rank $P=$ 0.006; Fig. 2C). A similar result was present in the entire set, where high-risk patients also had a significantly shorter OS than low-risk patients (median: 34.8 vs. 76.8 months, log-rank $P<$ 0.001; Fig. 2D).

3.3 Independence of the lncRNA signature for survival prediction

Univariate and multivariate Cox regression analyses were conducted to test whether the prognostic ability of the 9-lncRNA signature was independent of clinical data. For each dataset, we included age, gender, smoking status, tumor stage, residual tumor and lncRNA risk score as explanatory variables. The results showed that the 9-lncRNA risk score was an independent predictor of OS (high- vs. low-risk group: HR $=$ 3.020, 95% CI $=$ 1.935–4.715 for the training set; HR $=$ 2.005, 95% CI $=$ 1.166–3.448 for the testing set; HR $=$ 2.344, 95% CI $=$ 1.692–3.245 for the entire set; Table 3). In addition, it was found that other clinical factors including age, smoking status and tumor stage could also affect the OS of LUSC patients. Thus, we performed additional analysis by stratifying the whole dataset into 2 stratums based on these clinical features. The prognostic power of the 9-lncRNA signature was consistently observed across LUSC patients with different age (Fig. 3A and B), smoking status (Fig. 3C and D), and tumor stage (Fig. 3E and 3F).

3.4 Potential biological roles of the lncRNA signature

To investigate the potential functional roles involving the 9 prognostic lncRNAs, the co-expressed relationships between the lncRNAs and mRNAs were evaluated using Pearson correlation coefficients in the entire set. The expressions of 621 mRNAs were found to be highly correlated with at least one of the prognostic lncRNAs. KEGG pathways for these genes were then analyzed, using the whole human genome as the background. The results demonstrated that the prognostic lncRNAs were mainly enriched in pathways such as metabolism-related pathways, phosphatidylinositol signaling system, p53 signaling pathway, and notch signaling pathway (Supplementary Fig. 2).

4. Discussion

LUSC is a subtype of non-small cell lung cancer and accounts for nearly 40% of all lung cancer. There is a problem in not only the early detection but also the prognostic assessment of LUSC, which lead to a poor 5-year survival rate of the patients [1]. Although recent studies have improved the prognosis prediction for LUSC patients, most of them focused on biomarkers such as mRNAs or miRNAs. For example, Li et al. constructed a 4-mRNA signature to predict the outcomes of patients with LUSC [21]. Also, Gao et al. have identified a 7-miRNA signature that may have clinical implications in the outcome prediction of LUSC [22]. LncRNAs are a major subclass of ncRNAs expressed in a more tissue- and cancer-specific manner than mRNAs, implying their intrinsic utility in diagnosis and prognosis [23]. Indeed, several lncRNAs, such as VPS9D1-AS1 and MALAT-1, have been reported to be correlated with the survival of LUSC patients [24, 25]. However, to date, the prognostic lncRNA signatures for prognostic prediction of LUSC are scare and need to be systematically investigated.

In the present study, we analyzed the lncRNA expression profiles and clinical information of 493 LUSC patients from TCGA database. A set of 9 lncRNAs were identified to be strongly associated with patient’s OS by using univariate Cox regression and the subsequent RSFVH algorithm. These prognostic lncRNAs were then integrated into a 9-lncRNA risk score signature by combining their expressions and relative contributions in the multivariate Cox regression model. This signature showed a good performance in prognosis prediction, which could assign patients into low-risk and high-risk groups with significantly different OS in the training set. Similar predictive ability for OS of the 9-lncRNA signature was also documented in the testing set and the whole set. The above results suggested that the 9-lncRNA signature is reliable and robust in prognosis prediction of LUSC patients. In addition, by using the TCGA-LUSC dataset, a recent study has proposed a 6-lncRNA signature for outcome prediction of LUSC [26]. In that study, however, the 5-year AUC of ROC curve was only 0.672, indicating an inferior performance as compared to our 9-lncRNA signature.

To examine the predictive independence of the 9-lncRNA signature, we performed multivariate Cox regression analyses using the lncRNA risk score and available clinical data as explanatory variables. In the training, the testing and the entire sets, the 9-lncRNA signature was consistently identified to be an independent predictor for the OS of LUSC patients. Moreover, some clinical features, such as age, smoking status and tumor stage, were also significantly correlated with patient’s prognosis. Thus, to further confirm the utility of the 9-lncRNA signature in OS prediction, we carried out subanalyses by these clinical variables. We found that the prognostic power of the 9-lncRNA risk score was not modified by age, smoking status, and tumor stage. In summary, these findings indicated that the predictive value of the 9-lncRNA signature for OS of LUSC is independent of other clinical variables.

Although the number of documented lncRNAs has increased over the past decade, relatively few of them have been well-characterized. For example, only 184 human lncRNAs with known functions are included in the lncRNAdb v2.0 database [27]. It has been shown that lncRNAs exert biological functions through interacting with their partners such as mRNAs [28]. Therefore, the functions of the 9 prognostic lncRNAs were predicted by analyzing their co-expressed mRNAs. With functional enrichment analysis, we found that these lncRNAs mainly participate in regulation of metabolism-related pathways, phosphatidylinositol signaling system, p53 signaling pathway, and notch signaling pathway. Metabolism pathways, such as retinol metabolism, xenobiotic metabolism by cytochrome P450 and drug metabolism, have been reported to act as key enriched pathways in lung cancer patients [29, 30]. Recent studies have also highlighted the functional importance of phosphatidylinositol signaling system in LUSC [31, 32]. In addition to the well-known p53 signaling, alteration of the notch signaling pathway was increasingly reported as a pathological feature of lung cancer cells [33]. These results showed that the 9 prognostic lncRNAs may be involved in critical biological pathways associated with tumorigenesis and tumor progression of LUSC.

Some limitations should be acknowledged in this study. First of all, the 9-lncRNA signature was only identified and validated using RNA-Seq data from TCGA database. Therefore, this signature required additional confirmation in future large cohorts of LUSC patients. Second, although we used bioinformatics approach to explore the functional role of the 9 prognostic lncRNAs, the exact molecular mechanisms of them still remain unclear and need to be elucidated by experimental researches. Third, due to the lack of data, we cannot evaluate the impact of some confounders on the prognosis of LUSC patients, such as treatment strategies or medications.

Taken together, by analyzing the lncRNA expression profiles and clinical data of LUSC patients from TCGA database, we identified a total of 9 lncRNA biomarkers that were correlated with the OS of LUSC patients. The developed 9-lncRNA risk score model could effectively predict patient’s OS in the training set, which was further validated in the testing dataset and the entire set. Therefore, the 9-lncRNA signature may provide novel insights to predict the prognosis of LUSC patients.

Footnotes

Conflict of interest

The authors declare no conflict of interest.

Supplementary

Supplementary Fig. 1.

A flowchart for data processing and analysis.

Supplementary Fig. 2.

KEGG enriched by the 9-lncRNA signature using bioinformatics analysis.

References

Bray

Ferlay

Soerjomataram

Siegel

R.L.

Torre

L.A.

and Jemal

, Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries, CA Cancer J Clin 68 (2018), 394–424.

Schwartz

A.M.

and Rezaei

M.K.

, Diagnostic surgical pathology in lung cancer: diagnosis and management of lung cancer, 3rd ed: American College of Chest Physicians evidence-based clinical practice guidelines, Chest 143 (2013), e251S–e262S.

Wang

B.Y.

Huang

J.Y.

Cheng

C.Y.

Lin

C.H.

and Liaw

Y.P.

, Lung cancer and prognosis in taiwan: a population-based cancer registry, J Thorac Oncol 8 (2013), 1128–1135.

Szymanski

and Barciszewski

, RNA regulation in mammals, Ann N Y Acad Sci 1067 (2006), 461–468.

Hangauer

M.J.

Vaughn

I.W.

and McManus

M.T.

, Pervasive transcription of the human genome produces thousands of previously unidentified long intergenic noncoding RNAs, PLoS Genet 9 (2013), e1003569.

Mercer

T.R.

Dinger

M.E.

and Mattick

J.S.

, Long non-coding RNAs: insights into functions, Nat Rev Genet 10 (2009), 155–159.

Moran

V.A.

Perera

R.J.

and Khalil

A.M.

, Emerging functional and mechanistic paradigms of mammalian long non-coding RNAs, Nucleic Acids Res 40 (2012), 6391–6400.

Tay

Rinn

and Pandolfi

P.P.

, The multilayered complexity of ceRNA crosstalk and competition, Nature 505 (2014), 344–352.

Avazpour

Hajjari

and Tahmasebi Birgani

, HOTAIR: a promising long non-coding RNA with potential role in breast invasive carcinoma, Front Genet 8 (2017), 170.

10.

Lee

N.K.

Lee

J.H.

Ivan

Ling

Zhang

Park

C.H.

Calin

G.A.

and Lee

S.K.

, MALAT1 promoted invasiveness of gastric adenocarcinoma, BMC Cancer 17 (2017), 46.

11.

Zheng

Chan

M.T.

and Wu

W.K.

, NEAT1: a novel cancer-related long non-coding RNA, Cell Prolif 50 (2017).

12.

Gao

Z.Q.

Wang

J.F.

Chen

D.H.

X.S.

Tang

and Dang

X.W.

, Long non-coding RNA GAS5 suppresses pancreatic cancer metastasis through modulating miR-32-5p/PTEN axis, Cell Biosci 7 (2017), 66.

13.

Dong

Zhao

and Sun

, ZFAS1: a novel tumor-related long non-coding RNA, Cancer Cell Int 18 (2018), 125.

14.

Chen

Bao

Zhang

Liao

Liang

and Cheng

, The role of lncRNA CASC2 on prognosis of malignant tumors: a meta-analysis and bioinformatics, Onco Targets Ther 11 (2018), 4355–4365.

15.

Zhang

H.Y.

Yang

Zheng

F.S.

Wang

Y.B.

and Lu

J.B.

, Long non-coding RNA SNHG1 regulates zinc finger E-box binding homeobox 1 expression by interacting with TAp63 and promotes cell metastasis and invasion in Lung squamous cell carcinoma, Biomed Pharmacother 90 (2017), 650–658.

16.

Yin

J.Y.

F.Z.

Huang

M.S.

Zhu

Gao

Y.F.

Chen

Y.X.

Zhou

D.B.

Chen

Sun

L.Q.

Zhang

Zhou

H.H.

and Liu

Z.Q.

, Long noncoding RNA SFTA1P promoted apoptosis and increased cisplatin chemosensitivity via regulating the hnRNP-U-GADD45A axis in lung squamous cell carcinoma, Oncotarget 8 (2017), 97476–97489.

17.

Ishwaran

Kogalur

U.B.

Blackstone

E.H.

and Lauer

M.S.

, Random survival forests, Ann Appl Statist 2 (2008), 841–860.

18.

Meng

Zhang

Yang

and Fu

, A four-long non-coding RNA signature in predicting breast cancer survival, J Exp Clin Cancer Res 33 (2014), 84.

19.

Heagerty

P.J.

Lumley

and Pepe

M.S.

, Time-dependent ROC curves for censored survival data and a diagnostic marker, Biometrics 56 (2000), 337–344.

20.

Mao

Cai

Luo

and Wei

, KOBAS server: a web-based platform for automated annotation and pathway identification, Nucleic Acids Res 34 (2006), W720–724.

21.

Wang

Chen

Yang

and Chen

, A prognostic 4-gene expression signature for squamous cell lung carcinoma, J Cell Physiol 232 (2017), 3702–3713.

22.

Gao

and Li

, Identification of a seven-miRNA signature as prognostic biomarker for lung squamous cell carcinoma, Oncotarget 7 (2016), 81670–81679.

23.

Cheetham

S.W.

Gruhl

Mattick

J.S.

and Dinger

M.E.

, Long noncoding RNAs and the genetics of cancer, Br J Cancer 108 (2013), 2419–2425.

24.

Tan

and Yang

, Long noncoding RNA VPS9D1-AS1 overexpression predicts a poor prognosis in non-small cell lung cancer, Biomed Pharmacother 106 (2018), 1600–1606.

25.

Schmidt

L.H.

Spieker

Koschmieder

Schaffers

Humberg

Jungen

Bulk

Hascher

Wittmer

Marra

Hillejan

Wiebe

Berdel

W.E.

Wiewrodt

and Muller-Tidow

, The long noncoding MALAT-1 RNA indicates a poor prognosis in non-small cell lung cancer and induces migration and tumor growth, J Thorac Oncol 6 (2011), 1984–1992.

26.

Yang

Y.E.

Jin

Zhang

M.Y.

Liu

X.X.

Yin

Y.H.

and Qu

Y.Q.

, Identification of lncRNA biomarkers in lung squamous cell carcinoma using comprehensive analysis of lncRNA mediated ceRNA network, Artif Cells Nanomed Biotechnol 47 (2019), 3246–3258.

27.

Quek

X.C.

Thomson

D.W.

Maag

J.L.

Bartonicek

Signal

Clark

M.B.

Gloss

B.S.

and Dinger

M.E.

, lncRNAdb v2.0: expanding the reference database for functional long noncoding RNAs, Nucleic Acids Res 43 (2015), D168–173.

28.

Liao

Liu

Yuan

Kang

Miao

Xiao

Zhao

Luo

Zhao

Skogerbo

and Zhao

, Large-scale prediction of long non-coding RNA functions in a coding-non-coding gene co-expression network, Nucleic Acids Res 39 (2011), 3864–3878.

29.

Moulas

A.N.

Gerogianni

I.C.

Papadopoulos

and Gourgoulianis

K.I.

, Serum retinoic acid, retinol and retinyl palmitate levels in patients with lung cancer, Respirology 11 (2006), 169–174.

30.

Xie

Z.C.

T.T.

Gan

B.L.

Gao

Chen

and Hu

X.H.

, Investigation of miR-136-5p key target genes and pathways in lung squamous cell cancer based on TCGA database and bioinformatics analysis, Pathol Res Pract 214 (2018), 644–654.

31.

N. Cancer Genome Atlas Research, Comprehensive genomic characterization of squamous cell lung cancers, Nature 489 (2012), 519–525.

32.

Beck

J.T.

Ismail

and Tolomeo

, Targeting the phosphatidylinositol 3-kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR) pathway: an emerging treatment strategy for squamous cell lung carcinoma, Cancer Treat Rev 40 (2014), 980–989.

33.

Wael

Yoshida

Kudoh

Hasegawa

Niimori-Kita

and Ito

, Notch1 signaling controls cell proliferation, apoptosis and differentiation in lung carcinoma, Lung Cancer 85 (2014), 131–140.

A nine-long non-coding RNA signature for prognosis prediction of patients with lung squamous cell carcinoma

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

1. Introduction

2. Materials and methods

2.1 LUSC patient data

Table 1 Clinical characteristics of patients with LUSC

Table 2 The nine lncRNAs significantly associated with overall survival in the training set

2.4 Functional enrichment analysis

3. Results

3.1 Identification of a prognostic lncRNA signature in the training set

Table 3 Univariate and multivariate Cox regression analysis for overall survival

3.3 Independence of the lncRNA signature for survival prediction

3.4 Potential biological roles of the lncRNA signature

4. Discussion

Footnotes

Conflict of interest

Supplementary

References

Table 1
Clinical characteristics of patients with LUSC

Table 2
The nine lncRNAs significantly associated with overall survival in the training set

Table 3
Univariate and multivariate Cox regression analysis for overall survival