Home About us Editorial board Ahead of print Current issue Search Archives Submit article Instructions Subscribe Contacts Login 


 
 Table of Contents  
ORIGINAL ARTICLE
Year : 2018  |  Volume : 14  |  Issue : 1  |  Page : 201-207

Combining differential expression and differential coexpression analysis identifies optimal gene and gene set in cervical cancer


1 The Second Clinical College, School of Medicine, Wuhan University, Wuhan 430071; Department of Gynecologic Oncology, Hubei Cancer Hospital, Wuhan 430079, P.R. China
2 Department of Radio-chemotherapy Oncology, Zhongnan Hospital of Wuhan University, Wuhan 430071, P.R. China
3 Department of Gynecologic Oncology, Hubei Cancer Hospital, Wuhan 430079, P.R. China
4 Department of Gynecologic Oncology, Zhongnan Hospital of Wuhan University, Wuhan 430071, P.R. China

Date of Web Publication8-Mar-2018

Correspondence Address:
Prof. Hong-Bing Cai
Department of Gynecologic Oncology, Zhongnan Hospital of Wuhan University, No. 169 Donghu Road, Wuchang, Wuhan 430071
P.R. China
Login to access the Email id

Source of Support: None, Conflict of Interest: None


DOI: 10.4103/0973-1482.199787

Rights and Permissions
 > Abstract 

Objective: The objective of this study is to investigate the optimal gene and functional-related gene set in cervical cancer through combing the differential expression (DE) and differential coexpression (DC) analysis.
Materials and Methods: To achieve this, we first measured expression data of cervical cancer by incorporating DE and DC effects utilizing absolute t-value in t-statistic and Z-test, respectively. Then, we selected the optimal threshold pair to determine both high DE and high DC (HDE_HDC) partition on the basis of Chi-square maximization, and the best threshold pair divided all genes into four parts, including HDE_HDC, high DE and low DC (HDE_LDC), low DE and high DC (LDE_HDC), and low DE and low DC (LDE_LDC). Using the known functional gene sets, functional relevance of partition genes was explored to determine the best-associated gene set based on the functional information (FI) conception.
Results: Under the optimal threshold pair of 3.629 and 1.108 for DE and DC, respectively genes were divided into four partitions: HDE_HDC (311 genes), HDE_LDC (2072 genes), LDE_HDC (seventy genes), and LDE_LDC (1623 genes). Meanwhile, the gene set epidermis development was the best-associated gene set with the largest △G* = 10.496. Among the genes of epidermis development, zinc finger protein 135 (ZNF135) attained highest minimum FI gain of 41.226.
Conclusion: The combination of DE and DC analysis showed higher mean FI relative to individual DE and DC analyses. We successfully exhibited the optimal gene set epidermis development and gene ZNF135, which might be crucial for the prevention and treatment of cervical cancer.

Keywords: Cervical cancer, Chi-square maximization, differential coexpression, differential expression


How to cite this article:
Fang SQ, Gao M, Xiong SL, Chen HY, Hu SS, Cai HB. Combining differential expression and differential coexpression analysis identifies optimal gene and gene set in cervical cancer. J Can Res Ther 2018;14:201-7

How to cite this URL:
Fang SQ, Gao M, Xiong SL, Chen HY, Hu SS, Cai HB. Combining differential expression and differential coexpression analysis identifies optimal gene and gene set in cervical cancer. J Can Res Ther [serial online] 2018 [cited 2019 Nov 11];14:201-7. Available from: http://www.cancerjournal.net/text.asp?2018/14/1/201/199787


 > Introduction Top


Cervical cancer, arising from the cervix, is the second most common incident cancer and the third leading cause of cancer death in women worldwide,[1] resulting in more than 250,000 cervical cancer-related mortality annually.[1] Although the advances in disease treatment arise, patients with cervical cancer still present a poor prognosis due to late-stage detection and resistance to radiotherapy and chemotherapy.[2],[3] It is indicated that specific types of human papillomavirus infection are a high-risk factor to result in cervical cancer but not sufficient to cause malignant initiation,[4] and genetic alterations are essential for progression from precancerous disorder to invasive cancer.[5] It is predicted that molecular therapeutics might be crucial for the prevention and treatment of cervical cancer.

Currently, microarray technology has revealed the guiding principles for the molecular initiation and progression of complex diseases, enabling investigators to explore potential molecular biomarkers for early detection of cervical cancer. Growing evidence has demonstrated that differential expression (DE) analysis and differential coexpression (DC) analysis are powerful tools to explore diagnostic gene signatures and biological processes of complex diseases,[6],[7],[8],[9] and contribute greatly to the understanding of gene regulation systems.[10],[11] It is well confirmed that the propensity of many diseases can be reflected in the difference of gene expression levels.[12] In DE analysis, genes that present different expression levels across different conditions are identified, which is conductive to identifying cancer-specific gene signatures for distinguishing cancer patients from normal controls, and screening underlying candidate genes that assist better diagnosis and treatment of diseases at molecular level.[13] Different from DE analysis, DC analysis aims to study the potential interactions among individual genes, and further to reveal altered regulatory mechanisms by analyzing the difference in gene coexpression patterns between disease and normal subjects.[9],[14],[15] In general, both DE and DC analyses are beneficial for gene expression analysis, while they present different performance characteristics, one for individual genes and one for intrinsic gene interactions. The integration of two types of strategies might improve the testing power and provide new insights on dissecting complex disease mechanism.

Fortunately, Lui et al. presented an analytic framework for studying different DE and DC characteristics of the genes and indicated that the integration strategy could improve the detection of functionally relevant genes and contribute to a more systematic understanding of complex disease mechanism.[16] Thus, in the present study, the integration strategy was implemented to study the molecular alterations in cervical cancer and further to dissect the complex mechanism of cervical cancer. Details were described in the following sections.


 > Materials and Methods Top


Microarray data

To improve the statistical level and power, three microarray datasets of cervical cancer (E-GEOD-39001, E-GEOD-67522, and E-GEOD-63678) were recruited from the ArrayExpress (http://www.ebi.ac.uk/arrayexpress/) database. In these datasets, there were a total of 87 cervical cancer samples and 44 normal controls included in the current study. The characteristics of these three studies were shown in [Table 1]. To eliminate undesired batch effects on the gene expression values, batch mean-centering (BMC) method [17] employed in inSilicoMerging package (Brussels, Belgium. Support at https://insilicodb.com/)[18] was implemented to actually merge different datasets before further analysis. After preprocessing, a total of 4076 genes were obtained from the microarray data.
Table 1: Characteristics of the studies related to cervical cancer in our study

Click here to view


Differential expression analysis by t-test

To quantify the degree of DE of each gene, we performed the t-statistics for cervical cancer group (D) and normal control group (C) on the basis of the absolute t -value. A positive t-value of a gene indicated an up-regulation in cervical cancer; whereas a negative t-value represented a down-regulation. A higher absolute t-value showed a larger DE level. The absolute t-value ǀtaǀ for a given gene a, where a∈{1,…, k} (here k = 4076), was defined as:



Where and represented mean expression levels in cervical cancer and normal control respectively, nD and nC were standard deviations of expression levels in cervical cancer and normal control, and nD and nC were sample sizes of cervical cancer and normal control.

Differential coexpression analysis using the Fisher Z- test

To model DC analysis in D group and C group, we utilized the Fisher Z- test [19] to quantify the correlation difference by the following steps. In the first, we calculated the Pearson correlation coefficient)[20] between pairs of genes a and b in cervical cancer and normal conditions, respectively.



Where s was the number of samples; g (a, i) or g (b, i) was the expression level of gene a or b in the sample i; or represented the mean expression level of gene a or b; and n (a) orn (b) represented the standard deviation of expression level of gene a or b.

The within-group correlation values were Fisher z-transformed:





Finally, the following test statistic is computed for the difference of the z-transformed correlation between nD and nC for genes a and b:



Threshold determination for differential expression and differential coexpression by Chi-square maximization

In this part, we explored the relationship between DE and DC for each gene. To achieve this, we performed the calculation of thresholds value for DE and DC utilizing the Chi-square maximization, which was used to evaluate the dependency between two variables. The details were as follows:

Given k genes in the expression data, for an individual gene a, we selected one pair optimal thresholds for DE and DC, ta and za, respectively, ta was used for defining high or low DE; za was used for defining high or low DC. In this paper, we determined these two thresholds for each gene by Chi-square maximization. The detailed algorithm based on Chi-square maximization was previously described in the study of Lui et al.[16] The pair of optimal threshold was selected from a set of threshold candidates, {(zab, tb)} where b = {1,…, k}. Consider each pair of threshold candidates, every gene j where j = {1,…, k} can be categorized into one of following four parts: (1) low DE and low DC (LDE_LDC); (2) high DE and low DC (HDE_LDC); (3) low DE and high DC (LDE_HDC); and (4) high DE and high DC (HDE_HDC).

The occurrence frequency of each part was counted as , the number of expected frequency was defined as . Where M = {LDE, HDE}, N = {LDC, HDC}.

According to the above data, the Chi-square for the gene k was calculated as follows:



The pair of threshold candidate (za, tb) was considered as the optimal threshold pair for gene k when χk2 obtained the maximum value.

Furthermore, we calculated the statistical significance for the relationship between DE and DC for each gene a. In brief, when computing the Chi-square value by above steps, we also obtained the corresponding P- value on the basis of Chi-square distribution. Moreover, we initially adjusted the P- value by Bonferroni [21] and further corrected the adjusted P- value through Benjamini and Hochberg's method.[22]

Calculation of the functional relevance

In order to understand the functional roles of the selected genes in each part, we utilized predefined known functional genes sets which included of Gene Ontology (GO) sets,[23] Reactome pathways [24] and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways [25] as background to weight the functional relevance. Then, the Fisher's test [26] was performed to extract the related functional gene set by computing the P- value. The best-related functional gene set was defined as the most significant gene set associated with the partition with lowest P- value.

To compare the P- value of different part gene sets, we proposed the functional information (FI) to quantify the significance relationship between SHDE_HDC part of a gene and a functional gene set G. The formula was defined as follows:



The smaller adjusted P- value, the greater FI. The gain of FI (△G′) between the SHDE_HDC part and the individual HDE was computed as:



The gain of FI (△G′) between SHDE_HDC part and the individual HDC was computed as:



Moreover, we defined the total FI gain (△G*) by the minimum of individual FI gain of △G′ and △G″. The formula was as followed:

△G* = min (△G′, △G″)

Here, only both △G′ and △G″ were increased, △G* was high. When any one of △G′ and △G″ was low, the △G* was low. Furthermore, when △G* < 0, it represented that FI in the SHDE_HDC criteria set was less than individual HDE or HDC criteria. Otherwise, △G* > 0, it meant FI in the SHDE_HDC criteria set was greater than individual one. Finally, the best gene set was selected with the largest △G*.


 > Results Top


Data collection

On the basis of the combination of three datasets of cervical cancer, we obtained a total of 4076 genes from the expression data. To evaluate the functional relevance of selected gene set, a total of 7114 known functional gene sets were collected, of which 5895 sets from GO, 999 sets from Reactome pathways and 220 pathways from KEGG database. By intersecting with gene expression data, we selected 6738 background gene sets with the intersected gene size >3.

Optimal threshold pair

To measure the relationship between DE and DC, we calculated the DE and DC variables of 4076 genes using t-test and Z-test, respectively. Based on the Chi-square maximization, the dependency between DE and DC variables for candidate thresholds was calculated. When Chi-square was maximized, we obtained the optimal threshold (tb = 3.629, za= 1.108). With the optimal threshold pair of (3.629, 1.108), genes were divided into four partitions, including HDE_HDC (covering 311 genes), HDE_ LDC (covering 2072 genes), LDE_HDC (covering seventy genes), and LDE_LDC (covering 1623 genes). The result showed that 311 genes in cervical cancer data had a significant HDE_HDC association (P = 2.24 × 10−17).

Functional relevance of high differential expression and high differential coexpression partition

For 311 genes in HDE_HDC partitions, the best-associated gene set was investigated by analyzing the FI. Totally, 169 functional gene sets were associated with HDE and HDC partitions. [Figure 1] shows the top 10 best-associated gene sets with the highest mean minimum FI ( G *) gain for HDC_HDE partitions. Relative to individual HDE and HDC partitions, the combined HDE_HDC criteria showed higher mean FI in two gene sets, epidermis development ( G * = 10.496) and proteasome core complex (△G* = 4.378).
Figure 1: Top 10 best-associated gene sets with the highest mean minimum FI gain. The mean FI of HDE_HDC partitions was represented by the green yield. The FI gain of HDC partitions was symbolized as the red one, and the HDE partitions were marked by blue pillar. HDE_HDC=High differential expression and high differential coexpression, FI=Functional information

Click here to view


In-depth analysis of epidermis development

To illustrate the DE and DC detailed, the best association gene set (epidermis development) was selected for further exploration. Among these partitions, the gene zinc finger protein 135 (ZNF135) gained the highest minimum FI gain of 41.226. [Figure 2] illustrates the scatterplots of DE and DC for ZNF135. In [Figure 2], the optimal threshold pair (3.629, 1.108) based on Chi-square maximization was represented using red dash lines. The optimal point was placed in the region of the high Chi-square values. With the optimal threshold pair of (3.629, 1.108), genes were divided into four partitions. The distribution of genes in the best association gene set (epidermis development) was shown in [Figure 2]b. The scatterplot of correlation between genes and ZNF135 in cervical cancer and normal states was shown in [Figure 3]. It could be easily found that most genes in HDE-HDC partition, color in red, were negatively correlated with ZNF135 in the normal state [Figure 3]a, while positively correlated with ZNF135 in the cervical cancer state [Figure 3]b. A network relationship between ZNF135 and the genes of epidermis development was shown in [Figure 4].
Figure 2: The scatterplot of DE and DC for ZNF135. Every point stood for one gene. The abscissa represented the absolute value of DC between a gene and ZNF135, which was symbolized as |z|. The ordinate represented the absolute value of DE of a gene, which was defined as |t|. The red lines stood for the thresholds for DC and DE. (a) The heat map of the Chi-square values for the threshold candidates. (b) The best gene set epidermis development in the HDE_HDC partitions. The genes in epidermis development were symbolized by triangles. The genes with HDE_HDC value were expressed by red triangles; HDE_LDC (green); LDE_HDC (blue); LDE_LDC (pink). DC=Differential coexpression, DE=Differential expression, HDE_HDC=High DE and high DC, HDE_LDC=High DE and low DC, LDE_HDC=Low DE and high DC, ZNF135=Zinc finger protein 135

Click here to view
Figure 3: The scatterplot of DE and correlation between ZNF135 and every gene in (a) normal state and (b) cervical cancer state. The HDE_HDC (red); HDE_LDC (green); LDE_HDC (blue); LDE_LDC (pink). The abscissa represented the correlation coefficients under the in normal and disease condition, respectively. The ordinate represented the DE (t) of the gene. The red lines stood for the thresholds for DE. DC=Differential coexpression, DE=Differential expression, HDE_HDC=High DE and high DC, HDE_LDC=High DE and low DC, LDE_HDC=Low DE and high DC, LDE_LDC=Low DE and low DC, ZNF135=Zinc finger protein 135

Click here to view
Figure 4: The network between ZNF135 and the genes in epidermis development. The longer edges reflected the higher DC between a gene and ZNF135. The deeper red color represented the genes with higher DE. Genes with different DC and DE values were circled with different colors: HDE_HDC (red); HDE_LDC (green); LDE_HDC (blue); LDE_LDC (pink). DE=Differential expression, DC=Differential coexpression, HDE_HDC=High DE and high DC, HDE_LDC=High DE and low DC, LDE_HDC=Low DE and high DC, LDE_LDC=Low DE and low DC, ZNF135=Zinc finger protein 135

Click here to view



 > Discussion Top


The high mortality rate of cervical cancer is ascribed to disease recurrence despite cervical resection as well as ineffective treatment options for advanced disease. Numerous current studies have reported that genetic factors are the main cause of cervical cancer, and there is a close relationship between genetic polymorphisms and cervical cancer susceptibility.[27],[28] Therefore, the search of critical potential therapeutic targets was very important for improving the treatment and prognosis of cervical cancer.

In the current study, the combination of DE and DC method provided an efficient way to select both HDE and HDC genes in cervical cancer. Using this strategy, we can investigate whether the FI of an identified gene partition using the combining DC and DE criteria was higher than that using individual DE or DC criteria alone. This work demonstrated that combination method can improve the detection of some related gene sets on the basis of HDE and HDC criteria. Further investigation on the identified gene partitions and the associated functional pathways provided a systematic understanding of the disease mechanism. In this study, we identified the best-associated functional gene set epidermis development and the gene ZNF135 with the highest minimum FI gain of 41.226.

Epidermis development plays essential roles in the development of normal tissue. Previous studies have indicated the associations of epidermis development with cervical cancer. A study of Carlson declared that epidermis development showed a positive correlation to cervical cancer tissue.[29] In the cervix from K14E6 transgenic mice, epidermis development was altered in the gene expression profile in cervix tissue.[30] From a genome-wide gene expression profiling of cervical cancer, several differentially expressed genes were identified in cervical cancer, four of them were related to epidermis development pathway.[31] As the most important gene in epidermis development, ZNF135 is a protein-coding gene. Previous analysis has indicated that the human genome contained many zinc finger genes.[32] Systematic isolation and mapping of ZNF genes is a straightforward approach for the identification of novel candidate disease genes, for example, ZNF80 and ZNF77.[33] Tommerup andVissing indicated that ZNF135 mapping to regions commonly deleted in solid tumors.[34] Furthermore, ZNF135 has previously been reported as hypermethylated in cervical cancer.[35] However, the precise roles of ZNF135 in the development and progression of cervical cancer have not been reported exhaustively. Here, ZNF135 was indicated to be a key gene with the highest minimum FI, suggesting its critical role in cervical cancer, and should be dissected by further investigation.


 > Conclusion Top


In a word, we demonstrated that the proposed method can disclose the relationship between a gene and its associated functional gene sets with HDC and HDE characteristics, and further investigate the potentially prevention and prognostic marker in cervical cancer.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.

 
 > References Top

1.
Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-Tieulent J, Jemal A. Global cancer statistics, 2012. CA Cancer J Clin 2015;65:87-108.  Back to cited text no. 1
    
2.
de Freitas AC, Gomes Leitão Mda C, Coimbra EC. Prospects of molecularly-targeted therapies for cervical cancer treatment. Curr Drug Targets 2015;16:77-91.  Back to cited text no. 2
    
3.
Liu S, Pan X, Yang Q, Wen L, Jiang Y, Zhao Y, et al. MicroRNA-18a enhances the radiosensitivity of cervical cancer cells by promoting radiation-induced apoptosis. Oncol Rep 2015;33:2853-62.  Back to cited text no. 3
    
4.
Tjalma WA, Van Waes TR, Van den Eeden LE, Bogers JJ. Role of human papillomavirus in the carcinogenesis of squamous cell carcinoma and adenocarcinoma of the cervix. Best Pract Res Clin Obstet Gynaecol 2005;19:469-83.  Back to cited text no. 4
    
5.
Martin CM, Astbury K, McEvoy L, O'Toole S, Sheils O, O'Leary JJ. Gene expression profiling in cervical cancer: Identification of novel markers for disease diagnosis and therapy. Methods Mol Biol 2009;511:333-59.  Back to cited text no. 5
    
6.
Doncheva NT, Kacprowski T, Albrecht M. Recent approaches to the prioritization of candidate disease genes. Wiley Interdiscip Rev Syst Biol Med 2012;4:429-42.  Back to cited text no. 6
    
7.
Liu SS, Cheung AN, Ngan HY. Differential gene expression in cervical cancer cell lines before and after ionizing radiation. Int J Oncol 2003;22:1091-9.  Back to cited text no. 7
    
8.
Deng SP, Zhu L, Huang DS. Predicting hub genes associated with cervical cancer through gene co-expression networks. IEEE/ACM Trans Comput Biol Bioinform 2016;13:27-35.  Back to cited text no. 8
    
9.
Choi JK, Yu U, Yoo OJ, Kim S. Differential coexpression analysis using microarray data and its application to human cancer. Bioinformatics 2005;21:4348-55.  Back to cited text no. 9
    
10.
Wyrick JJ, Young RA. Deciphering gene expression regulatory networks. Curr Opin Genet Dev 2002;12:130-6.  Back to cited text no. 10
    
11.
Tanay A, Sharan R, Kupiec M, Shamir R. Revealing modularity and organization in the yeast molecular network by integrated analysis of highly heterogeneous genomewide data. Proc Natl Acad Sci U S A 2004;101:2981-6.  Back to cited text no. 11
    
12.
Zhao J, Yang TH, Huang Y, Holme P. Ranking candidate disease genes from gene expression and protein interaction: A Katz-centrality based approach. PLoS One 2011;6:e24306.  Back to cited text no. 12
    
13.
Macgregor PF, Squire JA. Application of microarrays to the analysis of gene expression in cancer. Clin Chem 2002;48:1170-7.  Back to cited text no. 13
    
14.
Minguez P, Dopazo J. Assessing the biological significance of gene expression signatures and co-expression modules by studying their network properties. PLoS One 2011;6:e17474.  Back to cited text no. 14
    
15.
Ruan J, Dean AK, Zhang W. A general co-expression network-based approach to gene expression analysis: Comparison and applications. BMC Syst Biol 2010;4:8.  Back to cited text no. 15
    
16.
Lui TW, Tsui NB, Chan LW, Wong CS, Siu PM, Yung BY. DECODE: An integrated differential co-expression and differential expression analysis of gene expression data. BMC Bioinformatics 2015;16:182.  Back to cited text no. 16
    
17.
Sims AH, Smethurst GJ, Hey Y, Okoniewski MJ, Pepper SD, Howell A, et al. The removal of multiplicative, systematic bias allows integration of breast cancer gene expression datasets – Improving meta-analysis and prediction of prognosis. BMC Med Genomics 2008;1:42.  Back to cited text no. 17
    
18.
Taminau J, Meganck S, Lazar C, Steenhoff D, Coletta A, Molter C, et al. Unlocking the potential of publicly available microarray data using inSilicoDb and inSilicoMerging R/Bioconductor packages. BMC Bioinformatics 2012;13:335.  Back to cited text no. 18
    
19.
Cohen J, Cohen P, West SG, Aiken LS. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. London: Routledge; 2013.  Back to cited text no. 19
    
20.
Li M, Zhang H, Wang JX, Pan Y. A new essential protein discovery method based on the integration of protein-protein interaction and gene expression data. BMC Syst Biol 2012;6:15.  Back to cited text no. 20
    
21.
Bonferroni CE. Teoria statistica delle classi e calcolo delle probabilita. Florence: Libreria internazionale Seeber; 1936.  Back to cited text no. 21
    
22.
Benjamini Y, Hochberg Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. J R Stat Soc Ser B (Methodol) 1995;57:289-300.  Back to cited text no. 22
    
23.
Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. Gene ontology: Tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 2000;25:25-9.  Back to cited text no. 23
    
24.
Croft D, O'Kelly G, Wu G, Haw R, Gillespie M, Matthews L, et al. Reactome: A database of reactions, pathways and biological processes. Nucleic Acids Res 2011;39:D691-7.  Back to cited text no. 24
    
25.
Kanehisa M, Goto S, Sato Y, Furumichi M, Tanabe M. KEGG for integration and interpretation of large-scale molecular data sets 2012;40:D109-14.  Back to cited text no. 25
    
26.
Fisher RA. On the interpretation of χ2 from contingency tables, and the calculation of P. J R Stat Soc 1922;85:87-94.  Back to cited text no. 26
    
27.
Madeleine MM, Johnson LG, Smith AG, Hansen JA, Nisperos BB, Li S, et al. Comprehensive analysis of HLA-A, HLA-B, HLA-C, HLA-DRB1, and HLA-DQB1 loci and squamous cell cervical cancer risk. Cancer Res 2008;68:3532-9.  Back to cited text no. 27
    
28.
Shi TY, He J, Wang MY, Zhu ML, Yu KD, Shao ZM, et al. CASP7 variants modify susceptibility to cervical cancer in Chinese women. Sci Rep 2015;5:9225.  Back to cited text no. 28
    
29.
Carlson MW. Gene expression profiling and modeling of cervical cancer. The University of Texas at Austion, Dissertation; 2006. Available from: http://www.lib.utexas.edu/etd/d/2006/carlsonm15434/carlsonm15434.pdf. [Last accessed on 2016 May 12].  Back to cited text no. 29
    
30.
Mendoza-Villanueva D, Diaz-Chavez J, Uribe-Figueroa L, Rangel-Escareão C, Hidalgo-Miranda A, March-Mifsut S, et al. Gene expression profile of cervical and skin tissues from human papillomavirus type 16 E6 transgenic mice. BMC Cancer 2008;8:347.  Back to cited text no. 30
    
31.
Wong YF, Cheung TH, Tsao GS, Lo KW, Yim SF, Wang VW, et al. Genome-wide gene expression profiling of cervical cancer in Hong Kong women by oligonucleotide microarray. Int J Cancer 2006;118:2461-9.  Back to cited text no. 31
    
32.
Bray P, Lichter P, Thiesen HJ, Ward DC, Dawid IB. Characterization and mapping of human genes encoding zinc finger proteins. Proc Natl Acad Sci U S A 1991;88:9563-7.  Back to cited text no. 32
    
33.
Huebner K, Druck T, LaForgia S, Lasota J, Croce CM, Lanfrancone L, et al. Chromosomal localization of four human zinc finger cDNAs. Hum Genet 1993;91:217-22.  Back to cited text no. 33
    
34.
Tommerup N, Vissing H. Isolation and fine mapping of 16 novel human zinc finger-encoding cDNAs identify putative candidate genes for developmental and malignant disorders. Genomics 1995;27:259-64.  Back to cited text no. 34
    
35.
Wu JH, Liang XA, Wu YM, Li FS, Dai YM. Identification of DNA methylation of SOX9 in cervical cancer using methylated-CpG island recovery assay. Oncol Rep 2013;29:125-32.  Back to cited text no. 35
    


    Figures

  [Figure 1], [Figure 2], [Figure 3], [Figure 4]
 
 
    Tables

  [Table 1]



 

Top
 
 
  Search
 
Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

  >Abstract>Introduction>Materials and Me...>Results>Discussion>Conclusion>Article Figures>Article Tables
  In this article
>References

 Article Access Statistics
    Viewed1595    
    Printed57    
    Emailed0    
    PDF Downloaded65    
    Comments [Add]    

Recommend this journal


[TAG2]
[TAG3]
[TAG4]