In this study, distinct mRNAs, lncRNAs, and miRNAs were identified to further understand the molecular events related to CESC prognosis. In addition, a prognostic ceRNA network provides new insights for the prognosis of CESC.
CESE tumorigenesis involves a combination of multiple genetic alteration processes rather than simply due to a single gene alteration. However, almost all studies focused on a single or a single cluster ‘driver’ gene of CESC so far. Daniel et al  reported that Keratin-17 is a prognostic biomarker of CESC. Li et al  reported that FAM83A is a potential biomarker regulated by miR-20, which promotes the development of CESC through the PI3K/AKT/mTOR signaling pathway. At present, no single pivotal driver gene or gene cluster was reported to be superior to evaluate the prognosis of CESC. Moreover, TNM stage, as a major prognostic indicator, is based on anatomical information and does not reflect the biological heterogeneity of CESC. Hence, it is of great significance to identify the effective biomarkers and construct a multi-mRNA-based model based on survival-associated biomarkers to predict the prognosis of CESE.
Other studies constructed a ceRNA network of CESC, but there are some limitations. Song et al  constructed a CESC-associated ceRNA network which consisted of 50 lncRNAs, 81 mRNAs and 18 miRNAs, and found that several RNAs were associated with the prognosis. However, that study did not construct a prognostic model based on the prognostic RNAs for CESC. Another study conducted by Chen et al  also constructed a CESC-associated ceRNA network, however, the study only assessed the relationship between OS and a single gene. Although some studies have constructed a CESC-associated ceRNA and simultaneously developed a prognostic model, they did not use ROC curve to evaluate the predictive ability of the model. Prognostic models based on multiple mRNAs could provide more accurate predictions than single gene. Our study constructed a ceRNA network and identified several novel potential prognostic RNAs for CESC. Moreover, a prognostic model was constructed based on the prognostic mRNAs including OPN3, DAAM2, HENMT1, and CAVIN3. The evaluation ability was 0.726 of this model with ROC.
Previous studies developed several prognostic models based on multiple genes for CESC. Meng et al  developed a CESC prognosis model based on DSG2, ITM2A, CENPM, RIBC2, and MEIS2. Liu et al  constructed a multi-mRNA prognosis model which were comprised of ITGA5, HHEX, and S1PR4. Similarly, our study also constructed a CESC-associated prognosis model based on OPN3, DAAM2, HENMT1, and CAVIN3, and the evaluation ability was 0.726 of this model with ROC. Moreover, in our study, the Cox proportional hazard model was used, and we corrected for the following confounding factors that can greatly affect prognosis: T stage, N stage, M stage, and pathological stage. The RS signature of this model was significantly associated with the clinical outcome of CESC patients. These results indicate that the model we constructed is an independent prognostic factor.
The KEGG pathways analysis in the ceRNA network indicated that the targeted RNAs are significantly enriched in the TGF-beta signaling pathway and Cell adhesion molecules pathway. TGF-beta is a critical cancer-associated signaling pathway, which is involved in proliferation, apoptosis, differentiation, migration, and epithelia-mesenchymal transition (EMT) of cancer [18, 19]. TGF-beta signaling pathway plays vital role in CESC as well. Deng et al  reported that CD36 and TGF-β interact to promote the EMT of CESC. Yang et al  found that downregulation of SEMA4C could inhibit EMT, invasion, and metastasis of CESC via inhibiting TGF-beta. Cell adhesion molecules pathway plays a critical role in the development of CESC. Carvalho et al  reported that L1 cell adhesion molecule expression is associated with a poor prognosis. The biological process analysis revealed that the main disturbed biological process by those survival-associated RNAs in the ceRNA network is the epithelial cell proliferation. The cytokinetic homeostasis is controlled by the balance of cell proliferation and apoptosis. Previous studies have reported that excessive cell proliferative activity is a precancerous lesion in some types of tumors. T Obara indicated that the epithelial cell proliferation activity is significantly stepwise increased from normal gallbladder mucosa to cancer . However, different pathways are usually perturbed by different molecules and thus need to be further investigated in the laboratory.
Our study still has some limitations. Firstly, although the effects of OPN3, DAAM2, HENTM1, and CAVIN3 in our prognosis model on tumorigenesis have been reported in other types of cancer [24–27], the exact effects on CESC have yet to be fully elucidated and need to be verified by experiments. Secondly, the research data came from a single online database, and another independent cohort is needed to verify above results in the future. Thirdly, the information of cervical cancer patients from the TCGA should be assessed with another experimental method.