This article is mentioned in:

Introduction

Ovarian carcinoma (OV) is one of the most lethal gynecological malignancies worldwide, and it has recently been reported that there are ~239,000 newly diagnosed cases of OV globally each year, of which 152,000 result in fatalities (1). Since OV is often asymptomatic until advanced stages, patients are frequently only diagnosed late during the course of the disease, making it more difficult to treat (2). The current widely accepted standard treatment for OV involves maximal cytoreductive surgical debulking, comprehensive staging once diagnosed during surgery according to current International Federation of Obstetrics and Gynecology (FIGO) recommendations, and six postoperative courses of platinum-based chemotherapy (3), although detailed treatment plans are tailored to individual conditions. In 2018, according to American data, the overall 5-year survival rate of OV was 47%, but for the majority of women who are diagnosed with advanced-stage disease the survival rate drops to 29% (4). In fact, >80% of these patients initially respond well to therapy, but most of them eventually relapse and ultimately develop chemotherapy-resistant disease (5). Currently, the widely accepted factors able to predict the prognosis of patients with OV include post-operative residual tumor, histological type, FIGO stage, patient age and the presence of ascites (6). Many researchers have been trying to identify biomarkers relating to the prognosis of patients with OV and so far many biomarkers have been identified for the early diagnosis and prediction of progression and prognosis of OV (7–9), however none have been suitable for clinical application.

The tumor microenvironment (TME) has been shown to have a critical influence on the initiation and spread of tumors by affecting gene expression in tumor tissues (10–12). The TME is the cellular milieu in which tumor is located and is composed of a complex network that includes immune cells, mesenchymal cells, endothelial cells, inflammatory mediators, extracellular matrix molecules, and other components (13). Immune cells and stromal cells are the two major non-tumor components of the TME and have been proposed to be valuable for the diagnostic and prognostic assessment of tumors (13). Tumor patients with different degree of infiltration of immune cells and stromal cells had different prognosis. The degree of infiltration of immune cells and stromal cells was inversely proportional to tumor purity. Patients with different tumor purity would therefore have different prognosis (13). In 2013, Yoshihara et al (14) published an algorithm called the Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data (ESTIMATE). This method uses gene expression signatures to infer the fraction of stromal and immune cells in tumor tissues and tumor purity using gene expression data. Combined with the large amount of available tumor data, it is very effective for drawing association between tumor tissue components and the prognosis of patients. Jia et al (15) reported a list of glioblastoma microenvironment-related genes using this method, and demonstrated that these genes could predict poor outcome in glioblastoma. However, to date, there have been no reports using ESTIMATE scores to study OV.

As the prognosis for patients with OV is worse than any other gynecologic cancer, in the present study, ESTIMATE was utilized to pull out useful information about OV from The Cancer Genome Atlas (TCGA) database in order to identify genes that are not only significant prognostic predictors, but also immune-therapy target markers for OV, creating a diagnostic nomogram for patients with OV.

Materials and methods

Source data of OV from TCGA

The data used in the present study were obtained from the TCGA database (https://tcga-data.nci.nih.gov/tcga/), including gene expression data and clinical data, such as sex, age, histological type, survival and outcome of patients with OV. The inclusion criteria comprised: i) Patients with complete gene expression data that could be used to calculate immune scores with ESTIMATE; ii) patients with complete clinical data; and iii) patients with complete prognosis data. In order to identify genes that were both potential prognostic predictors and immune-therapy target markers, the immune score was calculated by the ESTIMATE as the grouping indicator. The ESTIMATE algorithm applies single-sample gene set enrichment analysis to gene expression data and outputs the estimated levels of infiltrating stromal and immune cells and the estimated tumor purity. This step was performed with the R software (version 3.4.3).

Identification and analysis of differentially expressed genes (DEGs)

After extracting the immune score data (estimated levels of immune cells) obtained by ESTIMATE, the median was calculated to group the data into the low-score group (below the median) and high-score group (above the median). Log2|FC|>1 and adjusted P<0.05 were set as the cutoffs to screen for DEGs. A volcano plot was created to visualize the differences in gene expression levels between the two groups. This step was performed with the R software (version 3.4.3).

Functional enrichment analysis of genes of prognostic value

Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were used to explore the potential biological processes, cellular components, and molecular functions of DEGs. Significantly relevant signal pathways were determined with Database for Annotation, Visualization and Integrated Discovery (16,17) (https://david.ncifcrf.gov/; version 6.8).

Overall survival curve

Kaplan-Meier (KM) plots were generated to illustrate the relationship between the expression levels of survival-related DEGs and the overall survival of patients with OV. The survival of patients with OV for each gene was compared and tested with the log-rank test. For PNPLA5, a weighted test, Tarone-Ware, was used for analysis. These steps were performed with the R software (version 3.4.3).

Identification and validation of survival-related DEGs

DEGs with P<0.05 were considered statistically significant and were included in subsequent analyses. Firstly, the expression levels of the identified DEGs were analyzed with a univariate Cox's proportional hazards regression model. Secondly, a multivariate Cox's proportional hazards regression model was performed to analyze the expression levels of these genes combined with age, tumor site, clinical stage and histologic grade. Then, the results of multivariate Cox's proportional hazards regression model were represented by forest plots. Finally, a lasso-penalized Cox's regression analysis was conducted to further narrow the range of DEGs with the greatest predictive performance using 10-fold cross validation based on the glmnet package from the R software. The expression levels of genes selected by lasso were then validated with the TCGA OV datasets, and genes whose expression levels were higher in OV tissues than normal tissues were used to generate the nomogram.

Establishment of the prognostic predictive nomogram

Two identified genes, and related clinical parameters were used in the establishment of a predictive nomogram through a step-by-step Cox's regression model to evaluate 3- and 5-year overall survival of patients with OV from the TCGA database. The risk score of every patient was calculated within CDC20B, PNPLA5 and age using the Survival program in the R software package, The risk score was calculated according to the following formula: Survival risk score=∑i=1NCoefi*Exi. The Coefi is the coefficient and Exi is the gene expression of two genes, and age. Patients were divided into six groups according to their age as follows: Group 1, ≤39 years; group 2, 40–49 years; group 3, 50–59 years; group 4, 60–69 years; group 5, 70–79 years; and group 6, ≥80 years. For each patient, the corresponding Coefi was multiplied by the corresponding age group number, to which were added the levels of CDC20B and the level of PNPLA5, in order to obtain the final risk score. These scores were used to build a nomogram predictive model using CDC20B, PNPLA5 and age. Finally, ROC curve analysis and calibration curves were performed to evaluate the effectiveness of the generated nomogram model.

External validation of the prognostic predictive nomogram

The GSE32062 dataset (18) was downloaded from Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE32062) in PubMed to validate the nomogram. As aforementioned, ROC and calibration curve analysis were performed to show the effectiveness of the nomogram applied in the GSE32062 external dataset. In addition, KM curves with these data were used to compare the prognosis between high-risk and low-risk groups according to the risk score.

Statistical analysis

Statistical analysis was carried out in the R software (version 3.4.3). Continuous variables were analyzed using Student's t-test for two independent samples. The Pearson correlation coefficient assessed the correlation between immune score and tumor purity. Univariate and multivariate Cox's regression analyses were also performed in R. In addition, the hazard ratio and 95% confidence interval was calculated to identify genes associated with overall survival. P<0.05 was considered to indicate a statistically significant difference.

Results

Obtaining the OV data from the TCGA dataset and calculating the immune score

The flowchart of the analysis procedure is shown in Fig. 1. The profiles were downloaded from the TCGA database, including gene expression and clinical information, of 379 patients with OV before 2017. Only cases with data for both complete gene expression and clinical information were included in the analysis, and the ESTIMATE algorithm produced immune scores for these cases ranging from −1756.05 to 6120.57 (Table SI). Results from the ESTIMATE algorithm found that in the OV data, the immune score was significantly negatively associated with tumor purity (Fig. 2), which is in line with a previous study that reported that the immune score is negatively associated with tumor purity (14). In order to explore the potential association between overall survival and the immune score, the 379 patients with OV were divided into high immune score and low immune score groups that had immune scores above and below the median score, respectively.

Figure 1. Flowchart describing the process of establishing a gene signature and prognostic nomogram for OV. DEGs, differentially expressed genes; ESTIMATE, Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data; OV, ovarian carcinoma; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas.

Figure 2. Correlation analysis between immune score and tumor purity in OV.

Identification of DEGs and bioinformatics analysis

When comparing genes from the low immune score group to genes from the high immune score group, 890 genes were identified as upregulated, whereas 1069 genes were downregulated, as shown in the volcano plot (Log2|FC|>1; P<0.05; Fig. 3).

Figure 3. Volcano plot showing 890 upregulated genes (in red) and 1,069 genes downregulated genes (in green) in ovarian carcinoma samples obtained from The Cancer Genome Atlas database.

GO and KEGG pathway enrichment analyses were used to identify the functions of the identified DEGs. GO analysis showed that DEGs were significantly enriched in biological processes related to cell proliferation and differentiation, G-protein coupled receptor (GPCR) signaling pathways and neurogenesis (Fig. 4A-C). These findings were consistent with previous research on the role of endocrine GPCRs in OV (19), showing that GPCRs are involved in many aspects of tumorigenesis, including the promotion of aberrant growth, increased cell viability, angiogenesis and metastasis (20,21). In addition, results from Fig. 4C demonstrated that DEGs were enriched in integral component of membrane, plasma membrane and postsynaptic membrane, which were regarded as important in biological information transmission. In addition, KEGG analysis showed that the identified DEGs were significantly enriched in biological processes related to the interactions between neuroactive ligands and receptors (Fig. 4D). We performed KM survival analysis on all up-regulated and down-regulated genes, and obtained the overall survival of 38 genes that were statistically different (Table SII).

Figure 4. Functional enrichment analysis of DEGs. (A) GO terms for molecular functions. (B) GO terms for biological processes. (C) GO terms for cellular components. (D) Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis of DEGs (P<0.05, low immune group vs. high immune score group). GO, Gene Ontology; DEGs, differentially expressed genes.

Evaluation of prognostic factors in OV

Prognostic factors associated with overall survival of OV were identified by univariate Cox's regression analyses. The results demonstrated that age and 14 DEGs were significantly associated with the overall survival (Table SIII). From these identified 14 DEGS, multivariate Cox's regression analysis showed that six genes and clinical features, including age, were significantly associated with OV prognosis (Fig. 5). In order to identify the DEGs that could better reflect the association with OV prognosis, lasso-penalized Cox's regression was performed. Results showed that these six genes, CDC20B, UDP glucuronosyltransferase family 1 member A6 (UGT1A6), PNPLA5, apolipoprotein A5 (APOA5), spermidine/spermine N1-acetyl transferase like 1 (SATL1), and zinc finger and SCAN domain containing 4 (ZSCAN4), age and clinical stage were significantly related with the prognosis of OV (Fig. 6 and Table SIV). When these results were combined with the multivariate Cox's regression analysis results, clinical stage was excluded, as it did not show a statistically significant difference in the multivariate Cox's regression analysis.

Figure 5. Forest plot showing the clinical factors and genes significantly associated with ovarian carcinoma prognosis. *P<0.05; **P<0.01; ***P<0.001. CI, confidence interval.

Figure 6. Identification of prognostic genes in patients with ovarian carcinoma. (A) Lasso coefficients. Curve 1 to 10 represented age, tumor site, clinical stage, histologic grade, CDC20B, UGT1A6, PNPLA5, APOA5, SATL1 and ZSCAN4 respectively. (B) Plots of the cross-validation error rates. The dashed lines signify the value of the minimal error and the greater λ value.

In order to explore the potential prognostic signature genes, KM survival curves of the genes identified in the lasso-penalized Cox's regression were calculated (Fig. 7). High expression of CDC20B and PNPLA5 was associated with a low survival rate, whereas high expression of the other four genes showed an opposite effect. These results suggest that CDC20B and PNPLA5 might be oncogenes, whereas the remaining genes may be tumor suppressor genes. After validation of the expression levels of the six genes, only the expression levels of CDC20B and PNPLA5 were in line with higher in OV tissues (Fig. 8), and only CDC20B, PNPLA5 and age were included in the subsequent analysis.

Figure 7. Kaplan-Meier curves of the six identified ovarian carcinoma prognostic genes. (A) APOA5. (B) CDC20B. (C) PNPLA5. (D) SATL1. (E) UGT1A6. (F) ZSCAN4. High expression of CDC20B and PNPLA5 was significantly associated with a lower survival rate, whereas high expression of APOA5, SATL1, UGT1A6 and ZSCAN4 were associated with a higher survival rate. APOA5, apolipoprotein A5; CDC20B, cell division cycle 20B; PNPLA5, patatin-like phospholipase domain containing 5; SATL1, spermidine/spermine N1-acetyl transferase like 1; UGT1A6, UDP glucuronosyltransferase family 1 member A6; ZSCAN4, zinc finger and SCAN domain containing 4.

Figure 8. Expression of the (A) CDC20B, (B) PNPLA5, (C) APOA5, (D) ZSCAN4, (E) UGT1A6, and (F) SATL1 genes in the TCGA OV database. Red represents ovarian tumor tissue (n=426) and grey represents normal ovary tissue (n=88). TCGA, The Cancer Genome Atlas; APOA5, apolipoprotein A5; CDC20B, cell division cycle 20B; OV, ovarian carcinoma; PNPLA5, patatin like phospholipase domain containing 5; SATL1, spermidine/spermine N1-acetyl transferase like 1; UGT1A6, UDP glucuronosyltransferase family 1 member A6; ZSCAN4, zinc finger and SCAN domain containing 4. Red, Number (T)=426; Gray, Number(N)=88.

Prognosis analysis with CDC20B and PNPLA5

Risk score was calculated based on CDC20B, PNPLA5 and age, and results showed that patients with high risk score had a worse long-term survival rate (Fig. 9A). Time-dependent ROC and calibration curves were created to determine the prognostic values of the predictive model based on the two genes combined with age (Fig. 9B and C). The areas under the curve (AUCs) for 3- and 5-year overall survival predictions for the risk scores were 0.678, and 0.620, respectively.

Figure 9. Prognosis analysis of the two genes in the The Cancer Genome Atlas ovarian carcinoma database. (A) Risk score curve showing that patients with high risk score had lower survival rate compared with patients with low risk score (P<0.0001). (B) Receiver operating characteristic curve. The AUC of 3-year survival was 0.678 and the AUC of 5-year survival was 0.620. (C) Nomogram calibration curve of cell division cycle 20B and patatin-like phospholipase domain containing 5 genes combined with age. AUC, area under the curve. Dashed curves, 95% confidence interval.

Building a predictive nomogram model

All of the parameters, including CDC20B, PNPLA5, age and the AUCs of the risk scores from the prognosis analysis were used to build a prognostic predictive nomogram evaluating 3- and 5-year overall survival, based on the stepwise Cox's regression model (Fig. 10). We displayed the two genes into different points according to their own expression levels, and each point corresponds to a score. Subsequently, we performed the same way to deal with the age. Three scores were obtained and were added to get the total score. The total score was used to determine the 3- and 5-year overall survival rates.

Figure 10. Nomogram model of ovarian carcinoma. CDC20B and PNPLA5 gene expression and clinical parameters can be calculated for their respective scores. In clinical practice, the patient's total score can be used to predict the patient's 3- and 5-year overall survival.

External validation of the predictive nomogram model

To validate the generated prognostic nomogram, the GSE32062 dataset downloaded from GEO was utilized as an external sample. The AUCs of the 3- and 5-year overall survival predictions for the nomogram using the GSE32062 dataset were 0.699 and 0.643, respectively (Fig. 11A and B). A survival curve was also calculated with these data (Fig. 11C), which showed that a high-risk score was associated with worse long-term prognosis, whereas a low risk score was associated with better long-term prognosis. The calibration curve further showed that the originated nomogram was a very effective predictive model (Fig. 11D).

Figure 11. External validation of the generated nomogram model. ROC curve of (A) 3-year and (B) 5-year survival. The AUC of 3- and 5-year survival was 0.699 and 0.643, respectively. (C) Survival curve. GSE32062 database from Gene Expression Omnibus was utilized to perform a survival curve, which also suggested that patients with high risk score had lower survival rates compared with those with low risk score (P<0.001). (D) Nomogram calibration curve for the external dataset. AUC, area under the curve; ROC, receiver operating characteristic.

Discussion

OV is a gynecologic tumor with high mortality rates, and as aforementioned, the prognosis for OV remains poor, with only a 47% overall 5-year survival rate (4). In general, the prognosis of patients with OV is closely associated with the stage at diagnosis. The 5-year survival rate is >70% for stage I and II, but decreases to 40 and 20% for stage III and IV, respectively (22). FIGO staging is currently the most valid tool for the prognostic prediction of OV (23), and all of the guidelines for OV diagnosis and treatment still rely on the postoperative pathology analysis for accurate staging (24). Therefore, researchers are looking for molecular biomarkers of OV to identify OV in earlier stages and thus begin treatment promptly. Some biomarkers might also work as prognostic predictors allowing clinicians to use them to give better treatment suggestions to patients and provide more individualized treatment plans, including more precise surgery, chemotherapy, radiotherapy, targeted molecular therapy and immunotherapy (25).

These molecular biomarkers can not only be used as a beneficial supplement to FIGO staging but can also be used to predict the progression of OV and can serve as new therapeutic targets. Molecular prognostic markers might also have potential value in the early detection of OV (25). Prognostic evaluation models based on molecular biomarkers could guide individualized treatment and improve the therapeutic effectiveness of different treatments, and this was the impetus during the current bioinformatics analysis involving scrutinizing and integrating a large amount of genetic and clinical data on OV from the TCGA database.

In the present study, the stromal score, immune infiltration score and tumor purity scores of the TCGA OV dataset were calculated using the ESTIMATE algorithm. Immune cells have been shown to be significantly reduced in high purity tumor sample tissues (14). This negative association between immune score and tumor purity was also validated in the present study, and analysis of the TCGA OV data showed that low tumor purity, a high immune score, is associated with better survival.

A total of 379 DEGs were identified in relation to the immune score of the TCGA OV datasets, and functional enrichment analysis showed that these DEGs were associated with the occurrence, proliferation and metastasis of OV. From the GO cellular components result, the most common function of these DEGs is as components of membrane, such as plasma membrane and postsynaptic membrane. As biological activities and information transferring rely on changes on the cell or organelle membranes (26), these results indicate that these genes might be associated with intercellular information trafficking or to various biological activities. Exosomes, which are important for information transfer inter or inner cell, contribute to tumor progression and metastasis by mediating epithelial-to-mesenchymal transition, migration, invasion, angiogenesis and immune modulation as well as metabolic, epigenetic and stromal reprogramming into a cancer-associated phenotype (27). Exosomes are also associated with the development of OV (28). In addition, from both the GO biological processes and GO molecular functions results, the DEGs identified here are involved in biological processes related to cell proliferation and differentiation, GPCR signaling pathways and neurogenesis. These findings are consistent with previous research about the role of endocrine GPCRs in OV (29–31). GPCRs have been shown to be involved in many aspects of tumorigenesis in OV, including the promotion of aberrant growth, increased cell viability, angiogenesis and metastasis (32). In OV development the role of GPCRs is primarily through the regulation of metastasis and proliferation (21). According to the KEGG analysis result, the most significant biological processes for the identified DEGs is neuroactive ligand-receptor interactions. A previous study has shown that significant changes in the expression levels of estrogen and progesterone receptors might play one of the most important roles during the development of OV (33).

By carrying out both univariate/multivariate Cox's regression analysis and lasso-penalized Cox's regression, six genes were identified to be related to the prognosis of patients with OV, namely APOA5, CDC20B, PNPLA5, SATL1, UGT1A6 and ZSCAN4. The expression levels of these six genes in OV tissue were compared with normal ovarian tissue using the TCGA OV dataset, which showed that the expression of CDC20B and PNPLA5 in OV tissues was significantly greater than that of normal tissues. Moreover, KM curves, risk score and calibration curves showed the significant predictive ability of these two genes. Thus, a two-gene prognostic marker (CDC20B and PNPLA5) was established to predict the overall survival rate of OV, and this was verified using the external GSE32062 dataset.

These two gene biomarkers were used to calculate the survival probability of individual patients with OV and the risk factors were validated in the whole patient set that could be useful for predicting the prognosis of patients with OV. The validation results showed that the low-risk group had a better survival rate. The AUCs of the ROC curves of the whole OV cohort were 0.678 and 0.620 for 3- and 5-year overall survival, respectively. Taken together, these results showed that these two genes had a strong ability to predict the prognosis of ovarian cancer.

The survival analysis of CDC20B and PNPLA5 were performed individually, and the results of these analysis suggested that high levels of expression of both genes on their own were significantly associated with poor prognosis in patients with OV. The effects of CDC20B and PNPLA5 expression in the prognosis of ovarian cancer has not yet been described in the literature. CDC20B belong to the cell division cycle 20 (CDC20) family of regulatory proteins that interact with several other proteins at multiple points in the cell cycle and CDC20B is required for nuclear movement prior to anaphase and for chromosome separation (34). It has been reported that aberrant expression of CDC20 is associated with malignant progression and poor prognosis in various types of cancer and it has been shown that CDC20 knockdown inhibits the migration of the chemoresistant PANC-1 pancreatic cancer cells and the metastatic MDA-MB-231 breast cancer cell line (35). Thus, it was suggested that the development of specific CDC20 inhibitors might be a novel strategy for the treatment of cancer with elevated expression of CDC20 (35).

PNPLA5 is a member of the patatin-like phospholipase family, and the protein it encodes has been shown to inhibit transacylation (36). In addition, PNPLA5 has been related to the initiation of autophagy (37). Autophagy is a cellular clearance system that removes unnecessary or dysfunctional components (38), and this process requires neutral lipids that are stored in lipidic droplets (37). During autophagy, neutral lipid storage is mobilized to support the formation of autophagy membranes (37). Autophagy is related to malignant transformation, and it affects tumor progression and therapeutic responses in malignant cells (39). Targeting autophagy has been seen as a promising anti-tumor therapy (39,40). The results of the present study confirm that the expression of PNPLA5, which is required for autophagy, is related to the prognosis of ovarian cancer, suggesting that PNPLA5 might be both, a predictive biomarker and a reliable target for ovarian cancer treatment.

Taken together, the current results suggest that CDC20B and PNPLA have considerable potential for predicting the prognosis of patients with OV. Therefore, collected data was utilized to construct a nomogram predictive model. Several nomograms have been constructed for predicting outcomes in patients with OV, which possess superior predictive ability comparing with the widely utilized FIGO staging system. In 2012, Barlin et al (41) identified a number of parameters, including age, stage, debulking, ASA and HBOC, for predicting disease-specific survival after surgery based on the outcomes of 478 patients with OV. In 2013, Lee et al (42) evaluated other parameters, including largest residual tumor size, number of organ sites of metastasis, status, CA125 and haemoglobin, for predicting long-term survival in patients who were initially responsive to a platinum-based regimen but subsequently suffered recurrence. A more recent study analyzed prognosis based on the log of the odds between the number of metastatic lymph nodes and the number of non-metastatic lymph nodes in OV (43,44). Compared to these previous nomograms, the present proposed model used simpler factors to predict the 3- and 5-year survival rate of individual patients since the expression level of these two genes could be checked just by blood tests, making this a more suitable model for clinical applications. In addition, the CDC20B and PNPLA5 genes used to establish the nomogram were identified based on immune scores, suggesting that these two genes might be potential biomarkers for targeted immunotherapy against OV.

Improving the accuracy of survival estimates is extremely important for clinical decisions regarding the treatment and follow-up of OV. The generated nomogram model presented two advantages: i) As the 3- and 5-year survival rate of individual patients with OV could be calculated through the nomogram, gynecologists might be able to devise more reasonable follow-up schedules for different patients; and ii) the nomogram is based on the immune score, thus the two genes used might not only be biomarkers for predicting prognosis, but might also be potential immunotherapy targets.

Although the nomogram had potentially strong prediction capabilities, there were still a number of limitations to the present study. Firstly, the study was based on retrospective data and thus, there was inevitably some inherent bias relative to the selection of patients. Secondly, well-known prognostic factors such as chemotherapy data and tumor markers, such as cancer antigen 125 and human epididymis protein 4, were not included in the nomogram, since the data for these paraments were incomplete. In summary, the present study demonstrated that CDC20B and PNPLA5 are independently associated with the prognosis of patients with OV. In addition, a nomogram model based on the expression levels of CDC20B and PNPLA5 to predict 3- and 5-year survival among patients with OV was established and validated, which can further contribute to individualized clinical decisions regarding the treatment of OV.

In conclusion, the present study combined TCGA data with the ESTIMATE algorithm to study the correlation between genes and the prognosis of patients with OV. By step-by-step statistical verification, two genes were identified to be associated with the prognosis of OV. In addition, these two genes were used in combination with the age of the patients to establish a prediction model to better evaluate the prognosis of patients with OV. However, the present study has clear limitations, such as the lack of prognostic factors and tumor markers, future research should focus on the design of trials able to further explore the prognostic value of CDC20B, PNPLA5 and age in combination with other prognostic factors and tumor markers.

Supplementary Material

Supporting Data

Acknowledgements

Not applicable.

Funding

No funding was received.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Authors' contributions

HL, JW, XW and HM made substantial contributions to conception and design of this study. JW curated the data. HL and XW performed the formal analyses. QW and SH carried out the investigation. ZM and LL developed the methodology. XL, YH, SL and HP performed the software analyses. HL drafted the manuscript. HL, JW and HM revised it critically for important intellectual content. HM gave final approval of the version to be published. All authors read and approved the final manuscript.

Ethics approval and consent to participate

Not applicable.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References