Despite the numerous research and clinical studies, ovarian cancer (OC) still has a high mortality rate among gynecological cancers since lack of effective biomarkers for early detection and prognosis [1-3]. Also, the treatment efficiency of OC is low depending on multiple challenges such as late diagnosis, lack of reliable markers, the development of resistance to current therapeutics, and phenotype of heterogeneity [1,4]. For these challenges of OC, new studies need to reveal underlying molecular mechanisms, and to discover molecular biomarkers for early diagnosis, prevention, and targeted therapy [4].

MicroRNAs (miRNAs) are small non-coding RNAs that are about ~22 nucleotides in length. Their function is transcriptional and post-transcriptional regulation of gene expression by targeting mRNAs [5,6]. According to the effect on cancer, there are two types of miRNAs which are tumor suppressor miRNAs and oncomiRs. Depending on the type of cancer, oncomiRs or tumor suppressor miRNAs are inhibited or stimulated, respectively [7]. Especially, dysregulations of specific miRNAs affect cancer cell proliferation, differentiation, metastasis, and recurrence formation [8,9]. Various miRNAs have been shown to play different roles in OC. The mechanisms of miRNAs to impact OC and signaling pathways are still unknown. [2,10].

Many experimental studies have verified the detected interactions with bioinformatics analysis and proved the accuracy and predictivity with in-silico tools or databases. Although significant progress has been made about in-silico analysis in evaluating miRNAs, the need for new tools/databases is increasing day by day. It is also because of the limitations in existing tools/databases, that has increased with the development of high-throughput miRNAs technologies to analyze miRNA [11,12].

In this study, pathfindR was used for enrichment analysis of target genes. PathfindR uses active subnetworks, where an active subnetwork can be defined as a subnetwork of interconnected genes in a protein-protein interaction network (PIN), predominantly consisting of significantly altered genes. The tool initially maps the input genes with significance values onto the PIN and identifies active subnetworks, then it performs enrichment analysis on the identified subnetwork gene sets. In general, enrichment approaches overlook the relational information captured in the PIN and the genes neighboring the significant genes are not considered. By identifying active subnetworks, pathfindR exploits interaction information to enhance enrichment analysis. Active subnetworks allow the inclusion of possibly relevant genes that are not significant but connect significant genes in the PIN, and, in turn, the identification of phenotype-associated connected significant subnetworks [13]. This aids pathfindR uncover relevant mechanisms underlying the studied disease/phenotype.

In this study, we developed an in-silico approach to evaluate target genes of OC-related circulating and previously validated miRNAs which may be targeted for therapeutic approaches and utilized for OC management and diagnosis (Fig. 1).

Recently, a considerable number of research have been made to determine the effect of miRNAs regulate cancer hallmarks and to develop the early diagnosis and prognosis of cancer [18]. Nevertheless, several challenges such as various sampling methods, sample size, detection techniques, gender, and ethnicity or genetic background, affect the reliance on utilizing miRNAs as biomarkers [1,3,10,19]. One of the fundamental research areas for easily detectable, non-invasive, sensitive, and specific miRNA-based biomarker discoveries are of a great value for accurate and effective early diagnosis, risk prediction, prognosis, recurrence, and effective management of OC [8,10,12].

Identification of miRNAs' target genes has been the focus of computational biology for the last few years. Since detecting all possible miRNA targets with high-throughput technologies is laborious and costly, a wide variety of computational resources has been developed. The methodologies used by databases range from evolutionary conservation evaluations of putative miRNA binding sites to machine learning and classification algorithms. Continuous improvement is needed to develop new tools/databases to accurate predictions of miRNA targets [20,21]. One of these databases is miRDB online database and determines miRNA-target prediction and functional annotations. Estimating that 3.5 million target genes were regulated by 7,000 miRNAs in human, mouse, rat, dog, and chicken, in the latest version of miRDB (v6.0) major updated in 2019. In miRDB, miRNA binding and target down-regulation features used to predict miRNA targets with machine learning methods, generates a prediction scores are in the range of 0–100, and candidate genes with scores ≥ 50 are presented as predicted miRNA targets [16].

The pathfindR tool was used for the enrichment analysis of target genes within the scope of the study. While pathfindR allows better identification of disease-related pathways, it should be noted that the tool requires a significance level per each input gene. To overcome this limitation, we followed an approach where we used the scores for miRNA-target gene pairs to calculate a significance level. Additionally, some genes (usually a small proportion) that are not in the PIN (that do not have any curated interactions) have to be discarded, resulting in the loss of possibly relevant genes.

In our previous studies, three downregulated circulating miRNAs (hsa-miR-1909-5p, hsa-miR-885-5p, and hsa-let-7d-3p) in OC patients were consistently validated by comparison with healthy individuals. In these studies, circulating miRNAs, which were determined to be dysregulated by microarray, were then validated with the qPCR [14,15]. In this study, we predicted the validated circulating miRNAs' target genes with miRDB database and performed functional analysis by pathfindR tool to understand the pathogenesis of OC especially the molecular mechanisms of its development. Also, overall survival and pathological stage analyses were evaluated with differentially miRNAs' target genes which are commonly found in the TCGA and GTEx datasets.

Two hundred and twenty-nine, 90, and 23 of target genes of hsa-miR-1909-5p, hsa-miR-885-5p, and hsa-let-7d-3p were separately determined via miRDB (score ≥ 67), respectively (Fig. 2). Enriched KEGG pathways of target genes were performed with pathfindR tool. Totally, 339 genes included in active-subnetwork-oriented enrichment analysis and as a result, 24 enriched KEGG pathways were identified (Table 2). Top three pathways with the lowest p-values are ubiquitin-mediated proteolysis (p < 0.001), spliceosome (p < 0.001), and Notch signaling pathway (p < 0.001), respectively (Fig. 3). TRAF6, UBA7, UBE2K, UBE2N, and WWP1 genes regulate the ubiquitin-mediated proteolysis pathway. While SRSF2, SRSF6, and U2SURP are involved in the spliceosome, ATXN1L and KAT2B are adjusted to the Notch pathway. Although hsa-miR-885-5p targets ATXN1L, KAT2B, SRSF2, SRSF6, TRAF6, UBE2K, UBE2N, U2SURP, and WWP1 genes, hsa-miR-1909-5p just targets UBA7 gene. No target genes of hsa-let-7d-3p were found to be associated with these three pathways.

Our pathway analysis determined different pathways that can be related to the OC development. Of these pathways, only ubiquitin-mediated proteolysis, spliceosome, and Notch signaling pathway showed that their p-value is less than 0.001. Besides the effects of these pathways on OC, they also have roles on other cancer types. Firstly, ubiquitin-mediated proteolysis is an essential mechanism that is responsible for 80%–90% of intracellular protein degradation and is involved in many cellular processes, including tumorigenesis, tumor survival and apoptosis [22,23]. Ubiquitin-mediated pathway modulates BRCA1/2, p53, ERBB2 gene expressions, ERK pathway, cyclin-dependent cell cycle regulation which are related to OC [24]. Bazzaro et al. [25] claimed that upregulation of proteasome subunit levels occurs in OC and proteasome inhibitors may have utility in the treatment of OC. Secondly, splicing mechanism is a crucial process that regulates cellular proliferation, differentiation, and survival [26]. Dysregulation of splicing processes and splicing factor genes contribute to cancer, including OC [27,28]. Especially, splicing machinery mutations especially contribute to tumorigenesis. Additionally, it is critical to understand that the molecular mechanism of RNA splicing is causing the development of drug resistance in cancer treatment [26]. Regulation of the relationship between splicing and cancer can be led to splicing-based therapies for cancer treatment [29]. Thirdly, Notch signaling pathway regulates not only cell self renewal and differentiation but also a cell to cell communication [30]. Upregulation of Notch signaling pathway proteins have been identified in OC [31]. Several studies have revealed that it is related to poor overall and disease free survival time, and more advanced stages [30,32]. The Notch signaling pathway plays a specific role in deregulation of signaling cascade has been associated with OC. Targeted therapy against Notch pathway activation can offer clinical benefit to OC [30,31].

According to these studies, our findings also indicate that our dysregulated miRNAs are significant in OC via pathway related-target genes. Briefly, circulating downregulated miRNAs could not suppress their target genes in these pathways, hence, have activated these pathways, resulting in the development of OC. As we can see from the previous studies, these pathways were found to be related with OC. We have discovered target genes of three downregulated circulating miRNAs (hsa-miR-1909-5p, hsa-miR-885-5p, and hsa-let-7d-3p) and also related-pathways of these genes. We propose novel mechanisms between miRNAs, target genes and OC that have not been elucidated previously using pathfindR. Our findings can be used as a diagnostic tool in OC. Our perspective on ATXN1L, KAT2B, SRSF2, SRSF6, TRAF6, UBA7, UBE2N, UBE2K, U2SURP, WWP1 and in OC will promote more extensive research on the molecular mechanisms of hsa-miR-1909-5p, hsa-miR-885-5p, and hsa-let-7d-3p and provide a reference for improving the clinical outcome of OC.

The development of prognostic multigene classification protocols can benefit the understanding of tumor biology as well as the prediction of cancer progression and treatment strategies. One critical issue is determining the properly combining the genes [33]. However, studies on the overall survival-related profiles in OC patients have progressed, whereas there have been no large-scale studies based on multicenter validation of gene expression profiles for prediction of disease progression or recurrence in OC patients [34]. Furthermore, pathological staging, which can be determined after surgery and examination of the removed tumor tissue, is likely to be more accurate than clinical staging because it provides direct insight into the extent and nature of the disease [35]. Additionally, differentially expressed miRNAs' target genes, which were identified as multiple candidate genes in OC, were integrated our multigene analysis into the overall survival outcomes and pathological stage analysis using TCGA and GTEx datasets (Table 3). The hsa-miR-885-5p, which was the most gene-targeting score (n = 64) and the most associated miRNA with OC, followed by the hsa-miR-1909-5p and hsa-let-7d-3p miRNAs which were gene-targeting scores (n = 20) and (n = 9), respectively. After FDR analysis, no significant differentially expressed target genes were found to be associated with overall survival. However, we found 24 genes that were targeted by three miRNAs identified in the pathological stage of OC (Fig. 4). Notably, ZNF407 and UBN2 had the lowest p-values for their association with OC pathological stages (p = 0.017856) (Table 3). The molecular role and mechanism of ZNF407 and UBN2 in the development and progression of OC is not well understood. Missense mutations in ZNF407 affect tumor progression in gastrointestinal stromal tumors were reported [36]. Moreover, Tan et al. [37] showed that ZNF407 functiones as a promotive factor in colorectal cancer metastasis and accelerates cell proliferation by regulating phosphoinositide 3-kinase/AKT-mediate pathway. UBN2 is widely expressed in tumor tissues and encodes a nuclear protein that interacts with viral and cellular transcription factors [38]. Zhao et al. [39] suggested that high UBN2 protein expression is an independent prognostic marker to identify patients with poor clinical outcomes in colorectal cancer by affecting the Ras/MAPK pathway. Thus, these genes are likely to be facilitated in therapeutic approaches for OC.

In conclusion, we offer in-silico evidence that validated circulating miRNAs' target genes and enriched pathways are related to OC and have potential roles in theranostics applications. Especially, enrichment pathways and pathological stage-related genes can be combined with validated miRNAs, their multiple analysis can further enhance the molecular etiology of the OC and also can be employed in future research for biomarker and drug development related to OC. Further experimental investigations are required to validate our results which will ultimately provide a new perspective for translational applications in OC management. Our study will allow a greater understanding of broader clinical application prospects.