RESEARCH ARTICLE Data Mining for Identification of Molecular Targets in Ovarian Cancer

Ovarian cancer is possibly the sixth most common malignancy worldwide, in Mexico representing the fourth leading cause of gynecological cancer death more than 70% being diagnosed at an advanced stage and the survival being very poor. Ovarian tumors are classified according to histological characteristics, epithelial ovarian cancer as the most common (~80%). We here used high-density microarrays and a systems biology approach to identify tissue-associated deregulated genes. Non-malignant ovarian tumors showed a gene expression profile associated with immune mediated inflammatory responses (28 genes), whereas malignant tumors had a gene expression profile related to cell cycle regulation (1,329 genes) and ovarian cell lines to cell cycling and metabolism (1,664 genes). Our laboratory results, using high density microarrays, showed gene expression and alternative splicing profiles in non-malignant, malignant ovarian tumors and ovarian cell lines (Juarez-Mendez et al., 2013). However, it is not clear the molecular interaction of deregulates genes in OC. Systems biology approach provides extraordinary tools to examine high complexity interaction of large gene expression data. Additionally, experimental evidence of proteins and RNA expression provided exceptionally information to search for molecular involved in prognosis, diagnosis and treatment of cancer. In this study we performed data-mining using high-density microarray and System biology using MetaCoreTM, Thomson Reuters to identify the most significant deregulated signaling pathways in non-malignant, malignant and ovarian cell lines.


Introduction
Ovarian cancers (OC) is the sixth most common malignance in the worldwide, and represent the fourth leading cause of gynecological cancer death (Jemal et al., 2008), this is mainly more than 70% of patients are diagnosed in advanced stages, and the five year survival is less to 50% (Jemal et al., 2008). OC is classified according to the ovarian tissue of origin, the epithelial ovarian cancer (EOC) is the most common (Cannistra, 2004). EOC is further classified into serous, cell clear, mucinous and endometrioid types, with serous type being the most common. In Mexico the incidence is of 10.1 cases per 100,000 women (Globocan, 2012). Several factors are involved in prognosis of OC such as: early detection, age, tumor stage, and familiar history of ovarian/breast cancer, among others.
The identification of molecular signature has improved our understanding of the molecular mechanism associated with ovarian cancer pathogenesis has identified molecular markers useful for diagnosis, prognosis and even as target for treatment (Chen et al., 2015). Recent data indicates that certain deregulated genes are associated whit tumor progression (Liu et al., 2015).
Unregulated proliferation, migration, invasion, and treatment resistance characterize the ovarian cancer cell as well as point mutation in BRCA1/2, copy number amplification, over/under gene expression, genetics and epigenetic modification of DNA among others. The Omics studies have improved the approaches in cancer research; they provide large-scale genomics analyses RESEARCH ARTICLE

Data Mining for Identification of Molecular Targets in Ovarian Cancer
Vanessa Villegas-Ruiz 1,2 , Sergio Juarez-Mendez 1 * of imbalances, gene expression, and proteomics profile. Our laboratory results, using high density microarrays, showed gene expression and alternative splicing profiles in non-malignant, malignant ovarian tumors and ovarian cell lines (Juarez-Mendez et al., 2013). However, it is not clear the molecular interaction of deregulates genes in OC.
Systems biology approach provides extraordinary tools to examine high complexity interaction of large gene expression data. Additionally, experimental evidence of proteins and RNA expression provided exceptionally information to search for molecular involved in prognosis, diagnosis and treatment of cancer. In this study we performed data-mining using high-density microarray and System biology using MetaCoreTM, Thomson Reuters to identify the most significant deregulated signaling pathways in non-malignant, malignant and ovarian cell lines.

Microarray gene expression
In this study we used microarray that included nonmalignant ovarian tumors (NMOT, N=2), malignant ovarian tumors (MOT, N=4), ovarian cell lines (OCL, N=4) and healthy ovarian tissue (HOT, N=4) according to our previous report (Juarez-Mendez et al., 2013). Microarray data analyses were performed using Partek Genomics Suite v6.6 software (Partk Incorporated, Saint Louis, MO). In brief, microarray data was summarized using Median Polis, quantile normalization, the background noise correction was archived using RMA and finally the data was log2 transformed. The microarray were compared as follows: NMOT vs HOT, MOT vs HOT and OCL vs HOT. The differential expressed genes were selected using cutoff fold change > 2 and < -2 and False Discovery Ratio (FDR) > 0.05.

Systems biology
The significant deregulated genes obtained by means of microarray gene expression were loaded in the Metacore portal; the significant data were labeled using ID gene and fold change. The ontology were analyzed using Enrichment analysis workflow, p-values were calculated according to dataset activated (p< 0.05).

Gene Expression
In order to identify deregulated genes associated to NMOT, MOT and OCL, we performed microarray analysis using a normal tissue (HOT) as a base line reference. The comparative microarray analysis showed differential expressed genes as follows: NMOT (N = 28), MOT (N = 1329) and OCL (N = 1664) Figure 1. Interestingly, in MOT and OCL we identified that ~60% genes were down regulated, unlike to NMOT in wich ~14% were down regulated. Our results showed an apparent progression of   On the other hand, the up and down regulated genes were mapped by chromosomal. The deregulated genes NMOT-associated were mapped to only 14 chromosomes:  NMOT, MOT and OCL. NMOT showed more up regulated genes distributed in 14 chromosomes, while, MOT and OCL more than 50% of differential expressed genes were down regulated. The most representative difference between MOT and OCL was Y chromosome. MOT showed up regulation in contrast to OCL wich showed down regulation.

Enrichment of deregulated genes
The cell has a high level of complexity in molecular interaction. In order to identify the gene ontology associated to NMOT, MOT and OCL, we performed a systems biology analysis using MetaCoreTM, Thomson Reuters. The deregulated genes were loaded in MetaCore portal, after that, we performed an enrichment analysis. The expressed genes were ontology-based classified the top five are ranked in Table 1.
The enrichment analysis in NMOT showed processes associated to immune response, inflammation, vessel morphogenesis and chemotaxis among others. On the other hand, we observed in MOT and OCL several processes associated to cell cycle such as: chromosome condensation, metaphase checkpoint, mitosis initiation, spindle assembly, G2-M and S Phase among others Table  1.
Several marks give the malignant phenotype in cancer cell such as: cell proliferation, angiogenesis, and self-survival. The transcriptome analysis in ovarian cancer and ovarian cell lines showed that cell cycle is the most significant cellular process deregulated. In order to integrate signaling pathways of deregulated genes in NMOT, MOT and OCL, we built a network based on gene expression profile.

Network reconstruction
The reconstructed network was performed using a curate data by means of MetaCoreTM, Thomson Reuters system biology (SB) approach. The SB analysis reveal 15 significant networks associated to NMOT. We selected the top five based on significant and number of seed Table 2. The seed were deregulated genes observed in microarray.
According to number of seed, we used the top network, including: FGF4, C1QTNF5, ITGA11, SP1 and FGFR2. Additionally, eight genes were significant and included in regulation of cell-cell process Figure 4.
On the other hand, 30 signaling pathways were   identified in MOT. The top five are shown in the table 3. We focused in the most significant related genes, including 25 deregulated genes such as: ATP2C2, UCK2, CRISPLD2, OLFML3 and KIAA0240 among others.
The target gene in this signaling pathway is the estrogen receptor protein ESR1; 16 genes were down regulated and nine up regulated. The GO processes were associated to cell cycle process, including: translation, elongation, gene expression and cell cycle checkpoint, among others.
After that, we build the network of the most significant and related genes deregulated Figure 5. Finally, we analyzed 1664 deregulated genes OCL-associated and 30 networks were identified; the top five networks are shown in the Table 4. The most significant network contains 25 seed including: MRPS28, MURC, FAM54A, GBGT1 among others. The most significant ontology was associated to mitochondrial process including: initiation, elongation and translation. The significant network is shown in Figure 6.

Discussion
A great challenge in cancer research is the understanding of such a complex trait as well as the identification of   Table 3.
Asian Pacific Journal of Cancer Prevention, Vol 17, 2016 1697 DOI:http://dx.doi.org/10.7314/APJCP.2016.17.4.1691 Data Mining for Identification of Molecular Targets in Ovarian Cancer molecular markers that could help to predict treatment response, better classification of tumors and the identification of druggable targets. The microarray gene expression is an extraordinary tool that provides a wealth of data about differentially expressed genes. In cancer, several cellular processes are involved such as: cell cycle, proliferation, apoptosis evasion, inflammation, migration and metastasis, among others (Hanahan and Weinberg, 2000;Hanahan and Weinberg, 2011).
On the other hand, alpha(q)-specific peptide GPCRs, alpha(q)-specific amine GPCRs and serotonin receptor were associated to schizophrenia. Our results could be suggesting that non-malignant ovarian tumor, share elements with malignant ovarian tumors. However, theses molecules are not integrated in cancer signaling pathways.
Several models are used to investigate the molecular basis of the phenomena in cancer research; we included cancer cell lines to investigate in vitro cancer. Our results showed a differential gene expression profile, as expected (Figure 1-3). Additionally, we identified 730 genes with correlation between MOT and OCL, 599 and 934 were exclusively for MOT and OCL, respectively. These data indicate large differences between the two models of cancer we used.
In conclusion, the great challenges in cancer are the early detection prognosis and treatment. Using microarray gene expression and systems biology approaches we could identify the most significant signaling pathways in non-malignant, malignant and ovarian cancer cell lines. The significant genes identified in non-malignant and malignant ovarian tumors could be useful as potential markers of disease.