Background
microRNAs (miRNAs) are small noncoding RNAs, consisting of around 20-22 nucleotides, that regulate gene expression by binding complementary gene transcripts, thus causing the translational suppression of mRNA (Bartel 2009; Guo et al., 2010), (Volinia et al., 2006). Because of miRNAs’ negative regulation of gene expression, over 30% of microRNAs play critical roles in fundamental processes – such as differentiation, development, cell proliferation and apoptosis in almost all living organisms (Bartel, 2004), (Esquela-Kerscher and Slack, 2006), (Calin and Croce, 2006), (Lagos-Quintana, 2001). Because miRNAs commonly weaken and destroy their target mRNAs, reverse expression relationships with sequence partner of mRNAs is obviously expected (Baek et al., 2008), (Selbach et al., 2008).

Normal tissues present different miRNA expression profiles from cancer tissues (Lu et al., 2005), (Volinia et al., 2006). The dysregulation of miRNAs is able to facilitate tumor formation and development (Croce, 2009), (Lujambio and Lowe, 2012). Comparison between differentially expressed miRNA in cancer relative to the corresponding control has been already done in previous studies of ovarian cancer. The expression level is categorized as abnormally high and low, or no expression of miRNAs (Iorio et al., 2007), (Zhang et al., 2008), (Wyman et al., 2009). For example, microRNAs overexpressed in ovarian cancer are mir-27a, mir-27b mir-23b, miR-503, miR-346 and miR-424, which are correlated with the magnitude of metastasis (Wang, Kim, and Kim, 2014), (Park et al., 2013). It has been shown that mir-199a can repress the expression of CD44 gene, resulting in the suppression of the tumorigenicity and multidrug resistance of ovarian cancer-initiating cells (Cheng et al., 2012). Similarly, hsa-miR-140-3p targets RAD51AP1 gene, which is responsible for a common DNA damage response pathway, showed significantly decreased expression in ovarian cancer (Miles et al., 2012).

The Cancer Genome Atlas (TCGA) is a publicly available cancer genomic database that supplies the genomic data related with individual human cancer types (“The Cancer Genome Atlas - Data Portal” 2015). In the last decade, The Cancer Genome Atlas (TCGA) project has been a large-scale collaborative effort and a powerful database portal, which let us search and compare a comprehensive directory of molecular abnormalities in various cancers. The data led us to find up-regulated and low regulated miRNAs in ovarian cancer. We used an unbiased approach to select the most differentially expressed miRNAs in cancer and normal controls. We used a novel strategy to categorize the expression data of microRNAs in ovarian cancer because miRNAs and their target mRNAs have a potential to change molecular and biological processes in the cancer cells leading to the discovery of new therapeutic options.

Materials and Methods
Patient samples
Ovarian Cancer miRNAs and control data, Level 3, are downloaded from TCGA (02/05/2014). The data analysis is illustrated as a flow chart (Figure 1). According to the expression level, 485 cancer patients’ and 22 controls’ data are sorted and extracted by using the R statistical program. The R original script has been written to detect aberrant miRNAs in the cancer.

Figure 1 Flow chart of ovarian cancer miRNAs data process. The figure presents each step of the data mining work such as; downloading from TCGA, R statistical process, selection of the microRNA, REC score and molecular and biological functions. The flow chart explicates the whole data mining process

preprocessing
The data first downloaded and then separated into two groups, ovarian cancer and controls. Then the same ID microRNAs expression are collected as patients and controls separately. The data preprocessing has been completed by finding their fold changes.

Expression analysis
The extracted miRNAs have been applied to Cancerminer (www.cancerminer.org) which is a web-based tool that calculates the possible interaction between miRNAs-mRNAs and produces results as a REC score (“CancerMiner” 2015). The high and low expressed miRNAs (Figure 3) are hierarchically clustered by a bioinformatics tool, HCE 3.5 software program (http://www.cs.umd.edu/hcil/hce/) (“HCE - Hierarchical Clustering Explorer” 2015). The software program has different parameters, but in this paper Euclidean Distance has been used to cluster and find their correlation. Negatively affected molecular and biological functions of the target genes are categorized using http://www.pantherdb.org/ (“PANTHER - Gene List Analysis” 2015).

Figure 3 The high and the low expressed miRNAs in ovarian cancer. The miRNA changes profile is presented high and low expressed in ovarian cancer. After the comparison 96 non- expressed and 134 minimum expressed miRNAs are detected high expressed miRNAs. However, 139 and 156 high expressed miRNAs in control is expressed 0 and minimum in ovarian cancer

Results
The Extracted Data
The R code separates the expression of 680 different miRNAs into 3 main groups: high, low and not expressed (Figure 1). According to an expression value of reads per million mapped, the data shows 230 highly expressed miRNAs and 295 low expressed miRNAs (Figure 2). In addition, 155 of them are almost passive and are expressed in neither cancerous cells nor normal cells (Figure 2).

Figure 2 Ovarian Cancer Up and Down regulated miRNAs numbers. The numbers in the figure label ovarian cancer microRNA activities. The microRNAs are separated mainly into 3 groups, up regulated, down regulated and not expressed. From the numbers, it can be easily detected how many of the microRNAs expression have been changed or not as a result of the cancer

Once we found the high and low expressed miRNAs, we compared them to their expression in control samples. We found that some miRNAs are not expressed at all in control samples. Thus, we wanted to categorize the candidate miRNAs into two groups: group 1 is the miRNAs which show zero expression in controls and group 2 are the miRNA which show some expression in control samples (Figure 3). We have done this analysis for both high and low expressed miRNAs (Figure 3) (Supplementary Table 1). This analysis found 96 different miRNAs which are highly expressed in cancer, but showed no expression in control samples. Also, 134 miRNAs which are highly expressed in ovarian cancer show minimal expression in control samples.

Table 1 The up regulated miRNAs and their target mRNAs selected based on the REC score. miRNAs are selected based on their high expression in ovarian cancer as compared to controls. The target mRNAs were found using Cancerminer database and selected based on REC score. Some miRNAs were not found in the Cancerminer database and those were excluded.%73 of the microRNAs are related to tumorigenesis

For the low expression miRNAs, we found 139 miRNAs that showed zero expression in cancer samples and high expression in controls. Furthermore, we found 156 miRNAs that had a minimal expression in the cancer cells and high expression in controls (Figure 3).

Since some candidate miRNAs have very high or very low expression, we decided to select only the candidates with >50 fold change in expression. This analysis found 31 aberrantly high and 89 low expressed miRNAs (the list of candidates is given in Supplementary Table 2). The non-expressed miRNAs (n=155) are given in Supplementary Table 2.

Table 2 The down regulated miRNAs and their target mRNAs selected based on the REC score. miRNAs are selected based on their low expression in ovarian cancer as compared to controls. The target mRNAs were found using Cancerminer database and selected based on REC score. Some miRNAs were not found in the Cancerminer database and those were excluded from the Table. 22% of the microRNAs are related to tumorigenesis

Hierarchical clustering of the specific miRNAs
The high and low expressed miRNAs which are selected by the R program in the cancer are clustered hierarchically to understand the possible correlation among them. The most correlated 17 up-down regulated cancer miRNAs are clustered hierarchically. The clustering heat map shows us the miRNAs activity and relation between each other (Figure 4). The families of miRNAs such as, mir-509-1,-2,-3, mir-129-1,-2, mir-663, -b and mir-200a, -b are highly expressed and are the most correlated. The low expressed cancer miRNAs are also highly correlated with each other and consist of different miRNAs except mir-519a-2 and mir-519a-1 (Figure 4).

Figure 4 Hierarchical clustering of high and low express microRNA. As a result of the statistical analyses, 17 of the highest and 17 of the lowest correlated microRNAs are shown as a heat map. The first figure shows up regulated microRNAs, and the second figure shows down regulated microRNAs

REC score of the miRNAs and target genes
To find the relation between miRNA-target interactions, the Cancerminer tool has been developed [22]. The tool gives the result as a REC score which is a rank-based statistical approach that has been developed to understand that miRNA-mRNA pairs with negative expression association has significantly better predicted miRNA-target interactions related with weakly or positively associated pairs [22]. The expression of genes and miRNAs come out antagonistically. The highest expressed 31 miRNAs have been applied to the Cancerminer tool but just 22 of them have a determined REC score (Table 1). In addition to the REC score, the association score in ovarian cancer has been determined in 22 miRNAs (Table 1). Interestingly, most of the target mRNAs were involved in tumorigenesis. It is clearly found that 73% microRNAs are directly related to tumorigenesis.

Similar analysis has been done for the low expressed miRNAs (Table 2). Out of 89 miRNAs which had >50 lower expression, 54 miRNAs had defined REC score and the corresponding target mRNA. The down regulated microRNAs relation to tumorigenesis is only 22%. It is obviously shown that up regulated microRNAs are more related to tumorigenesis.

Molecular Activities, Biological Function, Cellular Component and Protein Class affected by the miRNAs
The cancer’s cellular activities are obviously affected by miRNAs. According to the REC score, the mRNAs are categorized in order to understand how the extracted miRNAs are active and negatively affecting molecular activities, biological function, cellular component (Figure 5), protein class and pathway (Figure 6) in the cancer type. It is easily observable what kinds of mechanisms change or could potentially change as detected by a percentage in the cancer type.

Figure 5 Changes in biological process, molecular function, and cellular component in ovarian cancer by high expressed miRNAs. Aberrantly expressed microRNAs repress some vital processes in the cell. According to REC score, mRNAs are listed and run through the REC program to find changes in their own biological process, molecular function and cellular component. All of the changes are given as a percentage in the figure to clearly detect the changes.

Figure 6 Changes of Protein Class and Pathway

Conclusion
The changes in microRNA expression between cancer and normal ovarian tissue samples, either high and low, show that there is a dynamic expression of the miRNAs. Although the sequence of miRNAs is complementary to many genes, they show a preference for specific genes. Depending on the antagonistic relationship between miRNAs and mRNAs, some of the functions in cells are suppressed. The main target of this study was to understand how miRNAs are aberrantly expressed in ovarian cancer, destroying the cellular balance and progressing to cancer. We found that out of 680 different miRNAs, 230 show high and 295 show low expression. Furthermore, 31 show >50 fold higher expression in cancer, and 89 show >50 fold lower expression in cancer as compared to control samples of ovarian tissue. The work will help the molecular geneticist and clinicians to make a new drug which targets the different genes.

Abbreviations
TCGA, The Cancer Genome Atlas Data Portal REC Score, association recurrence (REC) score Cancerminer, microRNA finder tool.

Competing interests
The authors declare that they have no competing interests.

Authors’ Contributions
SDogan carried out the Bioinformatics and data mining studies, analyzing microRNA expression level, performed the statistical analysis, and drafted the manuscript. AKozaric carried out the molecular effects of microRNA in the cancer. The design of the study has been done by the two authors. All authors read and approved the final manuscript.

Acknowledgements
First of all we are thankful for TCGA, The Cancer Genome Atlas data portal, providing ovarian cancer mircoRNAs gene expression data via online. We thank Sead Banda for designing of figures kindly.

References
Baek, Daehyun, Judit Villén, Chanseok Shin, Fernando D. Camargo, Steven P. Gygi, and David P. Bartel. 2008. “The Impact of microRNAs on Protein Output.” Nature, 455 (7209): 64-71. doi:10.1038/nature07242
http://dx.doi.org/10.1038/nature07242

Bartel, David P. 2004. “MicroRNAs: Genomics, Biogenesis, Mechanism, and Function.” Cell, 116 (2): 281-97
http://dx.doi.org/10.1016/S0092-8674(04)00045-5

Bartel, David P. 2009. “MicroRNAs: Target Recognition and Regulatory Functions.” Cell, 136 (2): 215-33. doi:10.1016/j.cell.2009.01.002
http://dx.doi.org/10.1016/j.cell.2009.01.002

Calin, George A., and Carlo M. Croce. 2006. “MicroRNA Signatures in Human Cancers.” Nature Reviews Cancer , 6 (11): 857-66. doi:10.1038/nrc1997
http://dx.doi.org/10.1038/nrc1997

“CancerMiner.” 2015. Accessed April 14. http://cancerminer.org/.

Cheng, Weiwei, Te Liu, Xiaoping Wan, Yongtao Gao, and Hui Wang. 2012. “MicroRNA-199a Targets CD44 to Suppress the Tumorigenicity and Multidrug Resistance of Ovarian Cancer-Initiating Cells: MicroRNA-199a Inhibits Ovarian CIC Growth.” FEBS Journal, 279 (11): 2047-59. doi:10.1111/j.1742-4658.2012.08589.x
http://dx.doi.org/10.1111/j.1742-4658.2012.08589.x

Croce, Carlo M. 2009. “Causes and Consequences of microRNA Dysregulation in Cancer.” Nature Reviews Genetics, 10 (10): 704-14. doi:10.1038/nrg2634
http://dx.doi.org/10.1038/nrg2634

Esquela-Kerscher, Aurora, and Frank J. Slack. 2006. “Oncomirs- microRNAs with a Role in Cancer.” Nature Reviews Cancer, 6 (4): 259-69. doi:10.1038/nrc1840
http://dx.doi.org/10.1038/nrc1840

Guo, Huili, Nicholas T. Ingolia, Jonathan S. Weissman, and David P. Bartel. 2010. “Mammalian microRNAs Predominantly Act to Decrease Target mRNA Levels.” Nature, 466 (7308): 835-40. doi:10.1038/nature09267
http://dx.doi.org/10.1038/nature09267

“HCE - Hierarchical Clustering Explorer.” 2015. Accessed April 14. http://www.cs.umd.edu/hcil/hce/.

Iorio, Marilena V., Rosa Visone, Gianpiero Di Leva, Valentina Donati, Fabio Petrocca, Patrizia Casalini, Cristian Taccioli, et al. 2007. “MicroRNA Signatures in Human Ovarian Cancer.” Cancer Research, 67 (18): 8699-8707
http://dx.doi.org/10.1158/0008-5472.CAN-07-1936

Jacobsen, Anders, Joachim Silber, Girish Harinath, Jason T Huse, Nikolaus Schultz, and Chris Sander. 2013. “Analysis of microRNA-Target Interactions across Diverse Cancer Types.” Nature Structural & Molecular Biology, 20 (11): 1325–32. doi:10.1038/nsmb.2678
http://dx.doi.org/10.1038/nsmb.2678

Lujambio, Amaia, and Scott W. Lowe. 2012. “The Microcosmos of Cancer.” Nature, 482 (7385): 347-55. doi:10.1038/nature10888
http://dx.doi.org/10.1038/nature10888

Lu, Jun, Gad Getz, Eric A. Miska, Ezequiel Alvarez-Saavedra, Justin Lamb, David Peck, Alejandro Sweet-Cordero, et al. 2005. “MicroRNA Expression Profiles Classify Human Cancers.” Nature, 435 (7043): 834-38. doi:10.1038/nature03702
http://dx.doi.org/10.1038/nature03702

Miles, Gregory D., Michael Seiler, Lorna Rodriguez, Gunaretnam Rajagopal, and Gyan Bhanot. 2012. “Identifying microRNA/mRNA Dysregulations in Ovarian Cancer.” BMC Research Notes, 5 (1): 164
http://dx.doi.org/10.1186/1756-0500-5-164

“PANTHER - Gene List Analysis.” 2015. Accessed April 14. http://pantherdb.org/.

Park, Young Tae, J. Y. Jeong, M. J. Lee, K. I. Kim, Tae-Heon Kim, Y. D. Kwon, Chan Lee, Ok Jun Kim, and Hee-Jung An. 2013. “MicroRNAs Overexpressed in Ovarian ALDH1-Positive Cells Are Associated with Chemoresistance.” J Ovarian Res, 6 (1): 18
http://dx.doi.org/10.1186/1757-2215-6-18

Selbach, Matthias, Björn Schwanhäusser, Nadine Thierfelder, Zhuo Fang, Raya Khanin, and Nikolaus Rajewsky. 2008. “Widespread Changes in Protein Synthesis Induced by microRNAs.” Nature, 455 (7209): 58-63. doi:10.1038/nature07228
http://dx.doi.org/10.1038/nature07228

“The Cancer Genome Atlas - Data Portal.” 2015. Accessed April 14. https://tcga-data.nci.nih.gov/tcga/.

Volinia, Stefano, George A. Calin, Chang-Gong Liu, Stefan Ambs, Amelia Cimmino, Fabio Petrocca, Rosa Visone, et al. 2006. “A microRNA Expression Signature of Human Solid Tumors Defines Cancer Gene Targets.” Proceedings of the National Academy of Sciences of the United States of America, 103 (7): 2257-61
http://dx.doi.org/10.1073/pnas.0510565103

Wang, Yongchao, Sangmi Kim, and Il-man Kim. 2014. “Regulation of Metastasis by microRNAs in Ovarian Cancer.” Frontiers in Oncology 4 (June). doi:10.3389/fonc.2014.00143
http://dx.doi.org/10.3389/fonc.2014.00143

Wyman, Stacia K., Rachael K. Parkin, Patrick S. Mitchell, Brian R. Fritz, Kathy O’Briant, Andrew K. Godwin, Nicole Urban, Charles W. Drescher, Beatrice S. Knudsen, and Muneesh Tewari. 2009. “Repertoire of microRNAs in Epithelial Ovarian Cancer as Determined by Next Generation Sequencing of Small RNA cDNA Libraries.” Edited by Sudhansu Kumar Dey. PLoS ONE, 4 (4): e5311. doi:10.1371/journal.pone.0005311
http://dx.doi.org/10.1371/journal.pone.0005311

Zhang, Lin, Stefano Volinia, Tomas Bonome, George Adrian Calin, Joel Greshock, Nuo Yang, Chang-Gong Liu, et al. 2008. “Genomic and Epigenetic Alterations Deregulate microRNA Expression in Human Epithelial Ovarian Cancer.” Proceedings of the National Academy of Sciences, 105 (19): 7004-9
http://dx.doi.org/10.1073/pnas.0801615105