alexa Integration of Gene Coexpression Network, GO Enrichment Analysis for Identification Gene Expression Signature of Invasive Bladder Carcinoma | Open Access Journals
ISSN: 2329-8936
Transcriptomics: Open Access
Make the best use of Scientific Research and information from our 700+ peer reviewed, Open Access Journals that operates with the help of 50,000+ Editorial Board Members and esteemed reviewers and 1000+ Scientific associations in Medical, Clinical, Pharmaceutical, Engineering, Technology and Management Fields.
Meet Inspiring Speakers and Experts at our 3000+ Global Conferenceseries Events with over 600+ Conferences, 1200+ Symposiums and 1200+ Workshops on
Medical, Pharma, Engineering, Science, Technology and Business

Integration of Gene Coexpression Network, GO Enrichment Analysis for Identification Gene Expression Signature of Invasive Bladder Carcinoma

Hanaa Hibishy Gaballah*
Department of Medical Biochemistry and Molecular Biology, Faculty of Medicine, Tanta University, Egypt.
Visiting researcher at Bioinformatic Centre, Chemical Research Lab, Kyoto University, Japan.
Corresponding Author : Hanaa Hibishy Gaballah
Visiting Researcher at Bioinformatic Centre
Chemical Research Lab, Kyoto University, Japan
E-mail: [email protected]
Received November 06, 2015; Accepted January 07, 2016; Published January 11, 2016
Citation: Gaballah HH (2016) Integration of Gene Coexpression Network, GO Enrichment Analysis for Identification Gene Expression Signature of Invasive Bladder Carcinoma. Transcriptomics 4:126. doi:10.4172/2329-8936.1000126
Copyright: © 2016 Gaballah HH. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Related article at Pubmed, Scholar Google

Visit for more related articles at Transcriptomics: Open Access


Bladder carcinoma is the most common malignancy of the urinary tract. Identification of genetic biomarkers for tumor invasiveness will help in earlier diagnosis and proper treatment. The present study aimed to integrate coexpression network and GO enrichment analysis for identification of prognostic markers and key genes that contribute to bladder cancer initiation and progression using a DNA microarray dataset (GSE 37317), invasive and noninvasive bladder cancer genes were compared by applying weighted gene co-expression network, gene ontology and pathway analysis. This study identified candidate genes (PURA, SRPK2, TRAK1, BRD2, and UPF3) that might have significant role in progression and invasiveness of bladder carcinoma. These markers might aid in early diagnosis of muscle invasiveness of bladder cancer. In conclusion; these finding may provide better understanding of the molecular mechanism of bladder cancer progression and invasiveness.

Bladder carcinoma; WGCNAL; Coexpression; Invasiveness; GO analysis
Alteration of the balance of cell proliferation and cell death is the hallmark of cancer, which cause reprogramming of cellular metabolism to support neoplastic proliferation and to avoid immune destruction Ghafouri- Fard et al., [1]. Bladder cancer is the commonest malignancy of the urinary tract, with the incidence being four times higher in men than in women Ploeg et al., [2]. Approximately 25% of the urothelial tumors are presented by muscle invasion (stage ≥T2), the rest are confined to the bladder mucosa (Ta) or lamina propria (T1) Cheung et al., [3]. Transurethral resection (TUR) of the bladder tumor is the standard method used to determine the local stage of the neoplasm which also provides treatment for patients with non-muscle-invasive disease and staging information for those with muscle-invasive disease Biagioli et al., [4]. Prior efforts have identified urinary protein biomarkers of bladder cancer Wang et al., [5,6]. However, markers that can predict tumor stage have not been identified. Such markers would provide pre-transurethral resection information that could be used for patient counseling and determination of the extent of the TUR required in cases where muscle-invasive disease is predicted Schiffer et al., [7]. The vast amount of available high-throughput gene expression data has provided excellent opportunities for studying gene functions on a global scale. Comparing the gene-expression profiles of cancerous and healthy tissues can help in the identification and characterization of susceptible genes associated with cancer Chin et al., [8]. Weighted coexpression network analysis is a system biology method for describing the correlation patterns among genes across microarray samples. It can be used for finding clusters (modules) of highly correlated genes, for summarizing such clusters, for relating modules to one another and to external sample traits, and for calculating module membership measures Horvath et al., [9,10]. Network-based gene screening methods can be used to identify candidate biomarkers or therapeutic targets and to identify key genes that contribute to the disease phenotypes Kristensen et al., [11]. By focusing on correlated gene modules rather than on individual genes, the network approach may provide a means to bridge the gap from individual genes to complex traits Stranger et al., [12]. Thus, we aimed to apply WGCNA to identify more prognosis markers and key genes that contribute to bladder cancer initiation and progression and to dissect its related biological pathways and networks by integration of coexpression network analysis and GO enrichment analysis.
Data description, processing and differential expression analysis
The data set with the series accession number (GSE 37317) accessible at National Center for Biotechnology Information (NCBI), Gene Expression Omnibus data repository (http://www.ncbi.nlm. was used in this study. The 19 tissue samples for this microarray study consisted of 8 non-muscle invasive and 11 muscle invasive bladder cancers that were profiled for gene expression using the Affymetrix HG-U133A platform. The gene expression profile of one normal bladder tissue sample (GSM44682) was also used for the analysis of differential gene expression between normal and cancerous tissues. Raw gene expression data files were read into R statistical package using the Affy package and normalized using the Gene chip robust multichip average (GC-RMA) normalization method as implemented in Bioconductor Gautier et al., [13]. Then, the processed expression levels of each gene between normal samples and tumor samples were compared by linear modeling to identify a set of genes that are differentially expressed between normal and tumor samples. A false discovery rate (FDR) threshold of 5% was set to correct for multiple hypothesis testing. These analyses are performed in R statistical software with the Limma package Smyth [14].
Network construction and Module detection procedure
The expression profiles of these differentially expressed genes were inputted to construct weighted gene co-expression modules using the WGCNA R package Langfelder et al., [15]. Positive and negative correlations in gene expression may indicate different biological interactions (synergistic or antagonistic); hence, using the absolute value of the correlation may misinterpret biologically relevant information Song et al., [16]. Therefore, in order to identify direct gene-gene coexpression relationships among the set of differentially expressed genes, a signed weighted gene coexpression network was constructed by calculating the spearman’s correlations between gene expression levels for all pairs of genes, then, a signed similarity (Sij) parameter was derived: Sij=(1+cor (xi,xj))/2, where xi and xj consist of the expression of genes i and j across microarray samples Mason et al., [17]. The signed similarity (Sij) was then raised to power β to represent the connection strength (aij); where aij=Sijβ. This step aims to emphasize strong correlations and reduce the emphasis of weak correlations on an exponential scale. A power of β=2 was chosen so that the resulting networks exhibited approximate scale-free topology Zhang et al., [18]. The topological overlap was used to organize the genes into clusters or modules. To calculate the topological overlap for a pair of genes, we compare them in terms of their connection strengths with all other genes in the network. Hence, the topological overlap-based metrics consider the correlation changes to all other genes to determine the similarity between two genes, this proximity measure can be used as input of clustering method followed by module extraction Yip et al., [19]. Hereby, my approach of identification of co-expressed gene modules was based on a modification of WGCNA method by applying the spectral clustering method, instead of the suggested hierarchical clustering, considering topological overlap matrix (TOM) as a similarity measure. Despite being applied in earlier studies Zhu et al., [5] and proved some utility, interpretation of hierarchical clustering is complex; besides expression patterns of individual gene sequences become less relevant as the clustering process progresses Li et al., [20]. Spectral clustering methods correspond to a family of unsupervised learning algorithms. The clustering information can be obtained from analyzing the eigenvalues and eigenvectors of a matrix derived from pairwise similarities of the data Li et al., [20]. These eigenvectors become a new representation of the data, where the clusters form a localized structure Huang et al., [21]. Following construction of a gene co-expression network, spectral clustering was run using specc function of the kernlab package implemented in R statistical software Karatzoglou et al., [22]. Number of centers was set to 18 and number of iterations to 200. Transcripts were clustered into 18 distinct gene clusters or modules whose number of genes ranged from 38 (the sixth cluster) to 71(the seventeenth cluster). Each module represents a group of genes with similar expression profiles across the samples and the expression profile pattern is distinct from those of the other modules. Singular value decomposition was performed using the module Eigengenes function in WGCNA package implemented in R software to summarize the expression levels of all genes in each module Xiao et al., [23]. The module eigengene roughly corresponds to the average of the signed normalized gene-expression values for a given sample. The module eigengene-based connectivity (k) for each gene was defined by correlating the expression profile of a gene with the module eigengene of its resident module. This measures how connected a given gene is to biologically interesting modules. The larger ki, the more similar is gene i to the summary profile of the module Min et al., [24]. Furthermore, we assessed the trait-based gene significance by correlating the gene expression profile to the tumor stage, which was represented as a binary value (zero for the non-invasive and one for the invasive tumor). Gene significance specifies the biological significance of the gene, the higher this value, the more significant a gene is Puniya et al., [25]. Consequently, scatterplots, spearman correlation coefficients and the corresonding p-values as computed by the cor.test function in R statistical software were used to relate the gene significance to the intramodular connectivity. The physiological relevance of each module was assessed by measuring the module significance which was defined as the absolute value of the correlation between the tumor stage and the module eigengene. Functional annotation of the modules was performed on the basis of their gene composition using DAVID database. This software, which is available for download at (http://, calculates the p value for the extent of enrichment of a given biological pathway/set by performing Fisher’s exact test. ‘KEGG_PATHWAY’ and ‘PANTHER_PATHWAY ‘were selected for pathway enrichment analysis of coexpressed genes compared with the background list of all genes on the array. For characterization of modules, ‘GO_BP_FAT’ and ‘GO_MF_FAT’ were selected.
The module significance was determined by correlating the trait with the module eigengene (Table 1). The three highest module significance (MS) scores observed were for the ninth module (MS=0.696, p=9.2 × 10−4), the thirteenth module (MS=0.648, p=2.6 × 10−3) and the twelfth module (MS=0.45, p=0.05). We also used the non-parametric Jonckheere-Terpstra trend test to evaluate the correlation between the module eigengenes and tumor stage. For a given physiological trait, a measure of gene significance was defined as the absolute value of the correlation between trait and gene expression values. For example, the tumor stage can be used to define a gene significance of the ith gene expression GS (i)=|cor (x (i), stage)| where x (i) is the gene expression profile of the ith gene. Furthermore, eigengene based gene connectivity (k) of the three modules, which showed the highest module significance values, was plotted against gene significance for all genes in each module. These plots, shown in (Figures 1-3), revealed that there is a direct relationship between the connectivity of a gene and the extent to which it is related to the tumor stage. The three modules showed a significant positive correlation presented by (r=0.566, p=2. × 10−6) for the ninth, (r=0.607, p=1.54 × 10−5) for the thirteenth and (r=0.373, p=1.9 × 10−3) for the twelfth module. So, it can be concluded that the connectivity within a physiologically relevant module is directly related to the gene significance. Using DAVID database, we tested each module for both enriched biological process and molecular function GO terms. The genes of the ninth module were markedly enriched for organelle organization (p=1.2 × 10−5), negative regulation of programmed cell death (p=2.4 × 10−4) and DNA replication (p=3.3 × 10−3). Moreover, the genes of the twelfth module are highly enriched for cell cycle regulation terms such as regulation of mitotic cell cycle (p=3.4 × 10−4), mitotic cell cycle checkpoints (p=6.6 × 10−4), cell cycle phase (p=1.2 × 10−3). They were also enriched for DNA recombination (p=8.0 × 10−4) and nuclear mRNA splicing (p=3.2 × 10−3). Meanwhile, the genes of the thirteenth module were significantly enriched for cellular localization (p=2.4 × 10−2) and defense response to virus (p=3.9 × 10−2). Concomitantly, we conducted a pathway enrichment analysis for each module using DAVID database in order to obtain further insight into the functional significance of the gene modules. ‘KEGG_PATHWAY’ and ‘PANTHER_PATHWAY‘ were selected for pathway enrichment analysis of co-expressed genes compared with the background list of all genes on the array. The ninth module was found to be enriched for genes within Pancreatic cancer pathway (p=4.1 × 10−2), B cell receptor signaling pathway (p=4.4 × 10−2) and Ras Pathway (p=3.7 × 10−2). Results for the functional analysis, those were significant at p<0.05 level, are shown in (Tables 2-5). Screening for genes that may have a possible role in the progression and invasiveness of bladder cancer was performed by ranking the gene expressions based on their correlations with the clinical tumor stage. Furthermore, as the three modules eigengenes showed significant correlations with the tumor stage (with p value ≤0.05), genes were ranked by their membership to each module or by their eigengene based connectivity (K) values. Furthermore, genes potentially involved in bladder cancer progression were screened by two methods the first is ranking genes within each significant module by their eigengene based connectivity (K) values. The second is assessing the trait-based gene significance measure correlating the gene expression profile to the tumor stage. Specifically, we used the 80th percentile of each screening variable, which resulted in six genes inside the ninth module and five genes in each of the twelfth and the thirteenth modules (Tables 6-8).
Bladder carcinoma is the most common malignancy of the urinary tract. It is associated with high recurrence and mortality rates. The basis for bladder cancer development and progression is complex and involves genetic abnormalities, therefore; development of accurate surveillance tests to evaluate disease aggressiveness is a major clinical need Frantzi et al., [26]. Network analysis has been emerged as an attractive approach to decipher the occurrence and progression of complex diseases as it provides means to bridge the gap from individual genes to systems biology Zeng et al., [27]. Weighted gene coexpression network measures correlations among all the gene pairs and extracts modules based on the density of gene connections Liu et al., [28]. Using this approach, demonstrated that the co-expression modules for Hepatitis B virus/Hepatitis C virus -induced hepatocellular carcinoma revealed distinct progression patterns from hepatitis infection to hepatocellular carcinoma He et al., [29]. Therefore, the current study aimed to apply WGCNA for exploring the significant coexpressed modules, genes and the functionally enriched pathways involved in progression of bladder carcinoma from the noninvasive to the invasive phenotype. Several gene co-expression modules in bladder cancer were explored in this study. These modules are clusters of highly correlated genes; this high correlation could be a result of transcriptional coactivation, or the co-regulation of mRNA stability, resulting in a complex genetic network of closely related genes coordinately acting to achieve a specific biological function Zhang et al., [16]. Among these co-expressed modules, 3 physiologically relevant or significant modules were detected, as their module eigengenes are found to be highly correlated with the tumor stage. Furthermore a gene screening strategy was applied for detecting the genetic drivers responsible for the invasiveness of bladder carcinoma, composed of two criteria: the first is the high association with the clinical tumor stage, i.e., high values of gene significance and the second is the value of k, i.e., membership in a tumor stage-related module. In this context, the current study demonstrated that most of the highly connected genes of a module have higher gene significance relationships to the malignant phenotype. Many of the identified genes by this screening method have been previously suggested as associated with tumorigenesis and cancer prognosis. For example, PURA gene encodes a single-stranded DNAbinding protein which binds preferentially to the purine-rich element at origins of replication and in gene flanking regions. Thus, it is implicated in the control of both DNA replication and transcription. Deletion of this gene has been associated with myelodysplastic syndrome and acute myelogenous leukemia Lezon-Geyda et al., [30]. Also, it was found to be upregulated in HCV-related hepatocarcinogenesis De Giorgi et al., [31]. Well in line, SRPK2 gene product belongs to a family of cell cycle regulated protein kinases which phosphorylate Serine/ Arginine (SR) domain-containing proteins in nuclear speckles and mediate the pre-mRNA splicing. Of note, reported that SRPK2 gene was over-expresssed in lung adenocarcinoma and was associated with extensive stages which may be attributed to disruption of the splicing machinery Gout et al., [32]. Likewise, TRAK1 gene, encodes trafficking kinesin-binding protein 1, has been recently identified as an emerging novel cancer biomarker. An et al., [33] found that its elevated expression is correlated with poor prognosis in colorectal cancer patients. Noteworthy, BRD2 gene, encodes a transcriptional regulator belonging to the BET (the bromodomain and extra-terminal domaincontaining proteins), is characterized by tandem bromodomains that interact with acetylated histones and influence gene expression, cellcycle regulation, and development Belkina et al., [34]. In addition, BET proteins are associated with chromatin throughout mitosis, and thus facilitate rapid transcriptional reactivation of critical genes after mitosis. Recent studies showed that BET inhibitors inhibit growth in a range of hematopoietic malignant cell lines as well as prostate cancer Wyce et al., [35]. Concomitantly, UPF3 gene products are trans-acting factors of the nonsense-mediated mRNA decay pathway which is a surveillance mechanism that degrades transcripts containing nonsense mutations, preventing the translation of possibly harmful truncated proteins Chan et al., [36]. It has been reported that this pathway may have a role in hereditary gastric cancer by down regulation of the tumour suppressor E-cadherin transcripts in gastric cells Karam et al., [37]. Moreover, GO enrichement analysis of the significant coexpressed modules revealed that these modules are enriched with genes belonging to distinct cellular compartments. This finding is in agreement with previous findings by Wang et al., [38] who reported that proteins within the same compartment tend to be more highly connected. Pathway enrichment analysis of the co-expressed modules revealed activation of Pancreatic cancer pathway, B cell receptor signaling pathway and Ras Pathway. Conceivably, pancreatic cancer pathway, which involves apoptosis regulating genes, is implicated in many types of cancer and Ghosh et al., [39] suggested its major contribution in bladder cancer. Well in line, Arum et al., [40] have reported aberrant activation of B cell receptor signaling pathway in rat bladder carcinoma. Moreover, activated B cell receptor signaling pathway has been also linked to other types of human cancer as chronic lymphocytic leukemia Oppezzo et al., [41]. Concurring with the findings of this study, Ras signaling has been found to be activated in human bladder cancer Boulalas et al., [42], where its mutation cooperates with β-catenin activation to drive bladder tumourigenesis Ahmad et al., [43]. Altered Ras signaling was also cited as a contributing factor to various tumoral and nontumoral pathologies Fernández-Medarde et al., [44]. In conclusion, a gene coexpression network was constructed from DNA microarray gene expression data for human bladder cancer and reported significant genetic drivers and altered signaling pathways that might have a role in cancer initiation and progression. These finding shed light into the molecular mechanisms of bladder cancer, which is of great benefit for defining the driving molecules and pathways appropriate for noveltargeted therapies. However, these results need further experimental validation and can be considered as a starting point for research studies exploring the magnitude of the involvement of these candidate genes in progression of this type of cancer.
The author gratefully thanks Prof. Hiroshi Mamitsuka, the director of Bioknowledge Engineering lab, bioinformatics centre, Kyoto University, Japan, for his guidance and support throughout this research project.

Tables and Figures at a glance


Table icon Table icon Table icon Table icon
Table 1 Table 2 Table 3 Table 4
Table icon Table icon Table icon Table icon
Table 5 Table 6 Table 7 Table 8


Figures at a glance

Figure Figure Figure
Figure 1 Figure 2 Figure 3
Select your language of interest to view the total content in your interested language
Post your comment

Share This Article

Relevant Topics

Recommended Conferences

  • 2nd International Conference on Nucleic Acids , Molecular Biology & Biologics
    August 31-September 01, 2017 Philadelphia, Pennsylvania, USA
  • 2nd World Congress on Human Genetics & Genetic Disorders
    November 02-03, 2017 Toronto, Canada
  • 3rd International Conference on Genetic and Protein Engineering
    Nov 08-Nov 09, 2017 Las Vegas, USA

Article Usage

  • Total views: 7957
  • [From(publication date):
    June-2016 - Jul 20, 2017]
  • Breakdown by view type
  • HTML page views : 7893
  • PDF downloads :64

Post your comment

captcha   Reload  Can't read the image? click here to refresh

Peer Reviewed Journals
Make the best use of Scientific Research and information from our 700 + peer reviewed, Open Access Journals
International Conferences 2017-18
Meet Inspiring Speakers and Experts at our 3000+ Global Annual Meetings

Contact Us

© 2008-2017 OMICS International - Open Access Publisher. Best viewed in Mozilla Firefox | Google Chrome | Above IE 7.0 version