5.1 Data acquisition
The training cohort, TCGA-OV dataset containing 379 patients were downloaded from the Genomic Data Commons Data Portal (https://portal.gdc.cancer.gov/). The validation cohort, GSE9891 [20] and GSE26193 [21] dataset including 285 and 107 patients respectively were acquired from Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/). Perl software was used for data integration and extraction of lncRNAs expression data and corresponding clinical data.
5.2 Specimen collection
A total of 60 OC samples were collected at ShengJing Hospital of China Medical University (Shenyang, China) from January to December 2015. Clinical information of the OC samples is presented in Supplementary Table 1. The inclusion criterion of the samples was as follows: (1) High-grade serous OC diagnosed by postoperative pathology; (2) All patients underwent surgical treatment, and the lesion tissue was retained during the operation; (3) Complete prognostic information was available (4) Informed and consented participants in this study. The exclusion criteria of the samples was as follows: (1) Patients with cognitive dysfunction and autoimmune system diseases; (2) The patient had received hormonal or chemoradiotherapy prior to tissue collection; (3) Complicated with mental abnormalities, tumors in other parts or severe liver and kidney function abnormalities. This study was approved by the ethics committee of the ShengJing Hospital of China Medical University, and informed consent was obtained from all patients.
5.3 Identification of the m6A-related lncRNAs
The 21 m6A regulators including 8 writers (METTL3, ZC3H13, METTL14, RBM15B, CBLL1, WTAP, RBM15, and KIAA1429), 2 erasers (FTO and ALKBH5), and 11 readers (YTHDC1, YTHDC2, ELAVL1, YTHDF1, LRPPRC, YTHDF2, FMR1, YTHDF3, HNRNPC, HNRNPA2B1, and IGF2BP1) were extracted from the TCGA-OV dataset [22]. Pearson correlation analysis was performed to calculate the correlation coefficient between 21 m6A regulators and lncRNAs. LncRNAs meet the screening criteria p < 0.001 and |R|>0.4 were considered as m6A-related lncRNAs [23]. A lncRNAs- m6A regulators network was constructed and visualized by cytoscope software. Finally, function enrichment analysis was used to explore the functions of the m6A regulators in the network involed in OC through GenCLiP310 online website (http://ci.smu.edu.cn/genclip3/analysis.php) [24].
5.4 Construction of the m6A-related lncRNAs prognostic signature
The m6A-related lncRNAs were fitted into univariate Cox regression analysis to obtain the m6A-related lncRNAs with prognostic value according to p < 0.05. A least absolute shrinkage and selection operator (LASSO) Cox regression and forward stepwise method was conducted to further narrow the prognostic related lncRNAs. Multivariate Cox regression analysis was implemented to calculate the regression coefficients of the selected m6A-related lncRNAs. Each patient acquired a riskscore according to the formula:
Risk score =∑Coefi * xi (Coefi represents the regression coefficient, xi represents the expression level of m6A-related lncRNAs).
5.5 Real-time qPCR
Total RNA of OC samples was extracted using TriZol Reagent (Takara, Japan). cDNA synthesis was carried out using the AMV reverse transcriptase reagent box (Takara, Japan). Real-time PCR was performed using a 2×SYBR Green PCR Master Mix. Next, the 2-ΔΔCt method was used to calculate the relative gene expression with GAPDH serving as an internal reference. The sequences of primers used for RT-qPCR are presented in Supplementary Table 2.
5.6 Evaluation and validation of the m6A-related lncRNAs prognostic signature
Patients were divided into high- and low- risk groups based on the median riskscore. Kaplan-Meier (K-M) method was used to compare the differences in prognosis between groups. The receiver-operating characteristics (ROC) curve was performed to evaluate the effectiveness of the m6A-related lncRNAs prognostic signature and the area under the curve (AUC) was calculated. We then validated the results in GSE9891 and GSE26193 dataset. Finally, multivariate Cox regression analysis was used to investigate whether the riskscore was independent of the clinicopathological parameters as an independent prognostic factor in OC patients.
5.7 Construction of the nomogram model
To predict the survival rate of the OC patients, a nomogram model was conducted based on the expression level of the m6A-related lncRNAs prognostic signature using the “rms”package in R software. Calibration curves at 1-,3-,5- year were drawn to assess the consistency between actual and predicted survival rates [25].
5.8 Construction of the ceRNA network related to the m6A-related lncRNAs prognostic signature
CeRNA network plays an important role in the occurrence and progression of ovarian cancer. We constructed a ceRNA network related to the m6A-related lncRNAs prognostic signature and the corresponding m6A regulators. Firstly, we obtained the miRNAs interacted with the m6A-related lncRNAs prognostic signature from the miRDB online website (http://mirdb.org/custom.html) [26]. We then acquired the possible miRNAs interacted with the m6A regulators from miRWalk online website (http://mirwalk.umm.uni-heidelberg.de/) [27]. After intersecting the predicted miRNAs, a lncRNAs - miRNAs - mRNAs ceRNA network was conducted and visualized by cytoscope software.