| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Regular Articles |
1 Departments of General Surgery,
2 Biochemistry,
3 Pathology, and
4 Asan Institute for Life Science, University of Ulsan College of Medicine, Seoul;
5 Department of Biochemistry, Hanyang University, Seoul; and
6 Genome Research Center, Korea Research Institute of Bioscience and Biotechnology, Daejon, Korea
 |
ABSTRACT |
|---|
 Top ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The gene expression profiling of gastric cancer has been a challenging target for DNA microarray analysis (4, 5, 6) . Gastric cancers usually contain numerous inflammatory and stromal cells, which would interfere the profiling of cancer cells. To eliminate such interference, gastric cancer cell lines have been studied, and certain genes were suggested to be associated with particular behavioral changes (7, 8, 9, 10) . However, cultured cells may not necessarily reproduce the biological behaviors of tumor cells in the complex microenvironment in vivo.
Gastric cancers have a variety of pathological patterns and biological behaviors. In 1965, Laurén (11) proposed to classify gastric cancers into two groups, i.e., intestinal and diffuse, according to the histopathological pattern of growth. The classification is still valid in predicting the biological behaviors of stomach cancers; diffuse type cancers, which consist of single infiltrating cells, have worse prognosis than intestinal-type cancers of same stage (12) . Recently, Hasegawa et al. (13) have reported the expression profiling of intestinal type cancers using microdissection and amplification. They have demonstrated that microdissection and amplification would be an important approach for the analysis of complex tumors in vivo. The expression profiling of diffuse type cancers, however, still poses a considerable technical difficulty because it would not be easy to recognize and microdissect single-infiltrating cells enough to avoid error-prone multistep amplifications. Furthermore, gastric cancers may not always be classified easily based on such a dichotomy; there may be functionally distinct subgroups that would correlate with particular pathological patterns and/or biological behaviors. Molecular markers of those groups would lead to the treatment-oriented classifications and the development of more efficient therapeutics.
In the expression profiling experiments of complex tumors, several major questions remain to be answered before DNA microarray analysis may be applied to clinical research and/or medical practice. Can the global gene expression profiling detect subgroups of tumors with distinct pathological patterns and/or biological behaviors? What are appropriate experimental designs for efficient clustering and exploring relevant molecular markers? Also, do the experimental designs affect the content of information obtainable using the same samples and microarrays? How much reproducible are the results of expression profiling studies by other experiments using different tissue samples, DNA microarrays, and experimental designs?
The importance of appropriate experimental design of expression profiling has been reviewed recently (14 , 15) . The pairwise analysis of tumor and nontumor mucosa couples has been favored to reduce the factors of individual variation and/or pathological changes in the normal mucosa (14) . The mucosa of gastrectomy specimen frequently has various pathological changes associated with chronic H. pylori gastritis. The pathological steps of gastric carcinogenesis have not been characterized well thus far (16 , 17) ; mucosal epithelial cells in the chronic gastritis may have considerable changes in their genome or the expression profiles. For instance, the epithelial cells in H. pylori gastritis frequently show malgun cell change, which reflects extensive genomic damage and repair (18) .
In an effort to answer aforementioned questions, we have done the expression profiling of 22 gastric cancer/nontumor mucosa couples by pairwise analysis using 14K cDNA microarrays specifically designed for gastric cancer. Tumors were clustered based on the global expression profiles. For comparison, pooled samples of tumor and nontumor mucosa were similarly analyzed.
|
MATERIALS AND METHODS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Immediately after the pathological examination, representative tumor samples and adjacent nontumor mucosa samples were snap-frozen in liquid nitrogen. Tumor samples of 1 Cm3 were taken from the center avoiding necrosis. In case of early gastric cancers, serial sections were taken alternatively for sampling or histopathological examination. Nontumor mucosa samples, 2 x 2 Cm, were taken from the same topographical area as the tumors located 3 Cm apart from the margin. Sections were taken for histopathological examination from both tumor and mucosa samples. The tumor samples did not contain normal mucosal tissue, except for occasional entrapped gastric glands. The mucosa samples contained mucosa and a part of adherent submucosa: neither tumor nor dysplasia was included. They showed minimal to mild degree of chronic inflammation and/or atrophy according to the Updated Sydney system (19) . Most mucosa samples displayed minimal degree of the malgun cell change, which was characterized by the clear, enlarged nuclei and cytoplasm (18) .
The clinicopathological data of patients are summarized in Table 1
. Twenty advanced and 2 early cancers were included. Topographically, 11, 10, and 1 tumors were in the body, antrum, and cardia, respectively. In Korea, most tumors develop in the body and antrum, whereas cardiac tumors are relatively rare. Fifteen patients had regional lymph node metastasis, whereas 7 did not. Tumor-node-metastasis staging was according to the International Union Against Cancer system (20)
. Tumors were classified according to the Laurén’s classification (11)
. Some intestinal type tumors were additionally subclassified as papillary or solid type according to the predominant pathological components. The degree of lymphatic permeation, stromal and muscular infiltration, and desmoplastic stromal reaction were graded from 0 to 2. The acute and chronic inflammatory infiltration in the tumor was graded from 0 to 3 adopting the Updated Sydney system for gastritis (19)
.
|
For direct labeling of probes with oligodeoxythymidylic acid primer, the protocol of the Brown Lab at Stanford was adopted.8 For the hybridization, 50 µg of sample and reference RNA were labeled with Cy5 and Cy3, respectively. Reference RNA was prepared by mixing RNA from normal full-term placentas. For the pooling experiment, 50 µg of pooled RNA’s from tumor and nontumor mucosa were hybridized directly in quadruplication. Hybridized slides were scanned using a GenePix 4000B Scanner (Axon Instruments, Foster City, CA).
Data Analysis.
The log ratio values from all hybridizations were merged into a tabular file. For the unsupervised clustering analyses, the pairwise ratio of gene expression was calculated for every tumor/nontumor mucosa couple by subtracting the log ratio value of a nontumor mucosa from that of the matching tumor sample. Array elements on <80% of samples were excluded. Data of remaining 6075 genes were centered by subtracting the median log values. Hierarchical clustering was applied to both genes and samples using web-available software Cluster and TreeView written by Michael B. Eisen.9
The pairwise log ratios were analyzed using the SAM10 method, which provides the serial lists of both up-regulated and down-regulated genes at a given false significance rate of choice (23) . The pooled tumor/nontumor samples were also analyzed using SAM at the similar threshold level as the pairwise analysis to compare the efficiency of detection of differentially expressed genes. The hierarchical clustering was repeated using the genes selected by the individual pairwise analysis and pooled sample study, respectively.
In Situ Hybridization.
The gene expression was verified using in situ hybridization and/or immunohistochemistry. For In situ hybridization, 5-µM thick frozen sections were prepared using the same tissues analyzed for cDNA microarray. Frozen sections were fixed in 4% paraformaldehyde. To make appropriate probes, cDNAs were subcloned into pBluesript II SK (Stratagene, La Jolla, CA) vector to make linear DNA. Probes were made using T7 RNA polymerase and 35S-UTP and purified using Quick Spin column (Roche Molecular Biochemicals, Mannheim, Germany). Using the probes, the sections were hybridized at 58°C for 16 h. After washing, slides were coated with Kodak liquid emulsion NTB2 and exposed in dark for 1–2 weeks. Slides were developed in Kodak Developer D-19 and Fixer. Then, H&E staining was done.
Semiquantitative RT-PCR.
Single-stranded cDNA was synthesized with oligodeoxythymidylic acid primer from 3 µg of total RNA using Superscript II reverse transcriptase (Invitrogen Co.). Each single-stranded cDNA was diluted for subsequent PCR amplification by monitoring ß-actin as a quantitative control. Each PCR was carried out in a 20-µl volume of 1x PCR buffer for 10 min at 94°C for initial denaturing, followed by 25 cycles of 94°C for 30 s, 56°C for 30 s, and 72°C for 30 s, in the GeneAmp PCR system 9700 (Perkin-Elmer Applied Biosystems, Foster City, CA). The primer sequences are summarized in Table 2
.
|
|
RESULTS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
1.614 (Fig. 1)
. The age, differentiation, location, and lymph node metastasis did not correlate with the tumor clustering. The group A consisted of intestinal types, except for a case having mixed intestinal and diffuse type (Table 1)
. Compared with the other group, they tended to have intense chronic and/or acute inflammatory infiltration (Fig. 2A)
. Two early gastric cancers and 3 papillary carcinomas belonged to this group.
|
|
; B2, diffuse type (Fig. 2C)
; and B3, intestinal type tumors with extensive infiltration (Fig. 2D)
. In the subgroup B1, >30% of the tumor volume was in the dilated lymphatic space. In the subgroup B3, tumor cell aggregates infiltrated extensively into the lamina propria, muscularis and perineural space without significant stromal desmoplastic reaction. Neither B1 nor B3 pattern was present in group A tumors. In the diffuse type subgroup, B2, three tumors consisted of single-infiltrating cells entirely, whereas two cases had intestinal components as well. There was also a case of mixed type tumor in the group A; it had more inflammatory infiltration than the group B counterparts.
Tumor-Specific Gene Expression.
Upon the pairwise analysis of the 22 couples, 477 genes were detected to express differentially using SAM at the false significance rate 0.91% and 1.106 (Fig. 3)
: the up-regulated and down-regulated genes were 79 and 398, respectively (Table 3)
. Up-regulated genes included mitochondrial isocitrate dehydrogenase 2, IGFBP7, acid cluster protein 33, cathepsin B, midkine, tissue inhibitor of metalloproteinase 1, BAP1, NME1, ANXA2, protein phosphatase 4, TGFB1, and other genes and expressed sequence tags.
|
|
: they included LIPA, LIPF, APOB, lactotransferrin, PGC, SNC73, PSCA, translin-associated factor X, cathepsin E, TFF1, TFF2, mesoderm-specific transcript homologue, SEPP1, KLF4, LTBP4, calpain 2, basic leucine-zipper protein BZAP45, IFN
-inducible protein 27, immunoglobulin genes, and other genes and expressed sequence tags.
|
1.194, respectively. The overlapping genes with those of all tumor pairwise analysis are summarized in the right columns of Tables 3
and 4
; the numbers represent the serial ranking of differential expression in the given subtypes. More down-regulated genes than up-regulated genes overlapped among various subtypes and all tumor pairwise analysis. For instance, APOBwas down-regulated in the group A tumors predominantly. CA11was down-regulated in the group B tumors, especially in the subtype B2, diffuse-type cancers.
Pooled Sample Analysis.
To compare the efficiency of selecting differentially expressed genes in tumors, pooled samples of tumors and nontumor mucosa were hybridized directly using four chips. Upon the SAM analysis at the false significance rate 0.93%, which was comparable with that of the pairwise analysis above, 1389 genes were reported to express differentially in comparison with 477 genes detected by the pairwise analysis (Fig. 4)
. Alternatively, when the thresholds were set to detect comparable number of genes as that of the pairwise analysis, the false significance rate was 0.36%, which was much lower than that of the pairwise analysis (Table 5)
. Our data suggested that the pooling analysis might be more efficient for the detection of differential expression common to gastric carcinomas than the pairwise analysis.
|
|
. Among 477 genes reported by the former, 28.7 and 45.7% were included in the list of the latter at the false significance rate 0.36 and 0.93% ( 1.064 and 0.750), respectively. The overlapping genes are summarized in the right columns of Tables 3
and 4
. No clustering was achieved upon the hierarchical clustering of 22 tumors based on the genes detected by the pooling analysis (data not shown). The low concordance rate and failure of clustering suggested that the gene populations of the analyses did not necessarily concur.
Verification of Selected Genes.
The expression of candidate genes was verified by in situ hybridization and/or semiquantitiative RT-PCR analysis. By in situ hybridization, ANXA2was strongly up-regulated in tumor (Fig. 5A)
, whereas nontumor mucosa was largely negative (Fig. 5B)
. In contrast, TFF2was expressed in normal glands (Fig. 5C)
, whereas the tumor was negative (Fig. 5D)
. No convincing expression of PSCA, KLF4, IFN -inducible protein 27, and SEPP1was detected in tumors, either. In addition, we have studied the expression of 2 genes, which were detected by the pooling analysis but not by the pairwise analysis; suppressor of Ty (Saccharomyces cerevisiae) 6 homologue and FXYD domain-containing ion transport regulator 3 were also down-regulated in tumors in comparison with nontumor mucosa (data not shown).
|
. Compared with the ß-actin control, the expression of ANXA2, IGFBP7, BAP1, and galectin 4 was verified to be up-regulated, whereas LIPF, TFF2, NAD(P)H dehydrogenase, and quinone 1 were down-regulated. We have also studied 2 genes that were detected to be differentially expressed by the pooling analysis but not pairwise analysis; CALU and paraspeckle protein 1 were confirmed to be up-regulated in tumors (Fig. 6)
. Taking the data of in situ hybridization and RT-PCR analysis together, it was concluded that both pairwise and pooling analyses yielded verifiable results.
|
|
DISCUSSION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Tumors in groups A and B had high and low levels of inflammatory infiltration, respectively, suggesting that the expression of inflammatory cells contributed to the profiling and clustering. However, the clustering of group B as three pathologically distinct subgroups of high-grade carcinomas suggested that it was largely based on the expression profiles of tumor cells. The fact that all early and papillary type cancers were clustered in the group A supported the notion. Also, all diffuse type carcinomas, except for one, were clustered in the group B.
Intriguingly, our lists of differentially expressed genes did not correlate well with those of previous studies (4, 5, 6 , 13) . Among the up-regulated genes, midkine, SIAH binding protein 1, and NME1 were in the previous studies as well; they have been implicated in the carcinogenesis or metastasis (24, 25, 26) . Among the down-regulated genes, CA11, TFF1, TFF2, APOB, and PGCwere included in the previous studies (13) . The down-regulation CA11(27) and TFFs1 and 2 (28 , 29) have been reported in gastric cancers. In our data, CA11 was down-regulated predominantly in the group B, particularly the diffuse type tumors, suggesting that the down-regulation of CA11 might be related to high-grade tumors. The down-regulation of APOB, PGC, LIPF, LIPA, and lactotransferrin appeared to reflect the loss of normal mucosal function in cancer: the predominant down-regulation in diffuse type cancers was compatible with the notion that they were relatively less differentiated and more aggressive than the intestinal type counterparts.
The low reproducibility of DNA microarray data among research groups has been a problem to be solved. Several factors may have been related to the apparent discrepancy among the published data. First, different tumor samples were applied. For instance, Hasegawa et al. (13) have analyzed only intestinal type cancers, whereas ours included six diffuse type or mixed type cancers. Given the extreme diversity of gastric cancers, the sampling may pose a significant difference in the results. Thus, it is suggested that much more tumors should be analyzed to get a comprehensive data of expression profiling of gastric cancers eventually.
Second, the selected gene lists may vary according to the thresholds applied in the prefiltering and subsequent analysis. Particularly, genes with low expression levels have posed a problem in DNA microarray analysis. They tend to be excluded from the analysis to reduce technical errors; however, biologically important genes such as most transcriptional regulators would be expressed in low levels. Then, how many of the low expressers should be excluded and how many differentially expressed genes should be selected for the convincing global gene expression profiling? There is no consentaneous answer to the question currently. We may need to consider as many genes as the experimental false significance rate would permit and to validate the data independently.
Third, the difference of DNA microarrays used in various studies may also be related to the discrepancy of results. Regardless of the general quality of microarrays, different coverage of genes would result in apparently discrepant ranking of differentially expressed genes. Obviously, it would be advantageous to use organ- or disease-specific microarrays that would cover most related genes expressed in the tissue and/or disease of interest.
Fourth, the experimental designs may also affect the data considerably. In our study, we applied pairwise analysis of tumor/nontumor mucosa from the same patients to reduce the unknown effects of individual variation. The considerable influence of experimental designs on the obtainable results was demonstrated by the comparison of pairwise analysis with pooling analysis of same samples. The pooling analysis was more efficient than the pairwise analysis for the detection of genes with differential expression in cancers. Because severe sample variations were to be considered as experimental errors in the statistical analysis, the pooling analysis was expected to yield more sensitive detection of differentially expressed genes than the individual case analysis did. Indeed, the reported genes of the former were more than those of the latter by about three times at the same false significance rate. However, less than half of genes in the selected lists of two experiments coincided, and no satisfactory clustering of tumors was achieved based on genes detected by the pooling analysis. Despite the discrepancy, the expression of genes in both pools was verifiable by in situ hybridization and/or semiquantitative RT-PCR. Taken together, the microarray analysis of tumors may yield variable results and efficiency according to the experimental designs applied. It is suggested that DNA microarray experiments should be designed depending on the purpose of experiment.
In our lists, there were genes implicated in gastric cancer and/or carcinogenesis. Cathepsin B has been reported to correlate with the depth of invasion in early gastric cancers (30) . Tissue inhibitor of metalloproteinase 1 has been reported to be particularly up-regulated in schirrhous-type carcinomas (31) , and the up-regulation in stomach cancers was reported to correlate with serosal invasion, lymphatic permeation, and poor prognosis (32) . ANXA2is one of a calcium-dependent phospholipid-binding protein family that plays a role in the regulation of cellular growth and in signal transduction pathways (33) .
There were genes the down-regulation of which has been implicated in cancer. KLF4is a zinc finger-containing transcription factor, which arrests the cellular proliferation at the G1-S boundary of the cell cycle (34 , 35) . Programmed cell death 4 inhibits tumor promoter-induced neoplastic transformation (36) . The down-regulation of LTBP4has been implicated in the prostate cancer (37) . It is noteworthy that TGF-ß1, which has been implicated in carcinogenesis (38) , is in our list of up-regulated genes in gastric cancer. The down-regulation of LTBP4along with up-regulation of TGF-ß suggests that the TGF-ß pathway may be important in the pathobiology of gastric cancer.
SEPP1has a protective role against peroxynitrite-mediated oxidation and reduces phospholipid hydroperoxide in vitro (39) . It has been reported that SEPP1 was down-regulated in prostate cancers and inversely correlated with the progression of prostate cancer (40) . Gastric cancers appear to develop in the background of extensive cellular genomic damage (18) , which, in turn, develops frequently in the milieu of longstanding oxidative stress by acute foveolitis of H. pylori gastritis (41) . The down-regulation of SEPP1may have gastric epithelial cells to be more vulnerable to the inflammation-related oxidative stress.
Our data suggest that the molecular classification of stomach cancer is feasible despite the considerable complexity and variety. It needs to be confirmed whether the pairwise analysis of tumor/nontumor couples is indeed advantageous for the clustering. In any case, the fact that different data may be obtained by the experiments using different samples, microarrays, and experimental designs suggest that multiplicated global microarray analyses are required to get a comprehensive global expression profiling and relevant clustering of complex tumors, preferably using standardized protocols that would facilitate the comparison and interpretation of the data among the research groups.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Inchul Lee, Department of Pathology, University of Ulsan College of Medicine, 388-1 Poongnap-Dong, Songpa-Gu, Seoul 138-736, Korea. Phone: 82-2-3010-4551; Fax: 82-2-472-7898; E-mail: iclee{at}amc.seoul.kr
7 Internet address: http://kugi.kribb.re.kr. 
8 Internet address: http://cmgm.stanford.edu/pbrown/mguide/index.html.
9 Internet address: http://www.microarrays.org/software
10 The abbreviations used are: SAM, significance analysis of microarray; RT-PCR, reverse transcription-PCR; IGFBP7, insulin-like growth factor binding protein 7; BAP1, BRCA1-associated protein 1; NME1, nonmetastatic cells 1; ANXA2, annexin A2; TGF, transforming growth factor; LIPA, lipase A; LIPF, gastric lipase; APOB, apolipoprotein B; PGC, progastricsin; PSCA, prostate stem cell antigen; SEPP1, selenoprotein P1; TFF, trefoil factor; KLF4, Kruppel-like factor 4; LTBP4, latent-transforming growth factor ß binding protein 4; CALU, calumenin.
Received 5/ 6/03. Revised 7/29/03. Accepted 9/18/03.
|
REFERENCES |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  N.-S. Oh, J.-S. Park, Y.-J. Jeon, J.-H. Oh, S.-Y. Jeong, J.-O. Yang, Y.-W. Park, H.-S. Yoo, and N.-S. Kim Generation of Expression Clone Set for Functional Proteomics of Human Gastric and Liver Cancers Experimental Biology and Medicine, October 1, 2009; 234(10): 1220 - 1229. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Kim, Y. Kim, H.-T. Kim, D. W. Kim, Y. Ha, J. Kim, C.-H. Kim, I. Lee, and K. Song TC1(C8orf4) Is a Novel Endothelial Inflammatory Regulator Enhancing NF-{kappa}B Activity J. Immunol., September 15, 2009; 183(6): 3996 - 4002. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I.-S. Song, N. S. Oh, H.-T. Kim, G.-H. Ha, S.-Y. Jeong, J.-M. Kim, D.-I. Kim, H.-S. Yoo, C.-H. Kim, and N.-S. Kim Human ZNF312b Promotes the Progression of Gastric Cancer by Transcriptional Activation of the K-ras Gene Cancer Res., April 1, 2009; 69(7): 3131 - 3139. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K.-S. Chung, Y.-J. Jang, N.-S. Kim, S.-Y. Park, S.-J. Choi, J.-Y. Kim, J.-H. Ahn, H.-J. Lee, J.-H. Lim, J.-H. Song, et al. Rapid Screen of Human Genes for Relevance to Cancer Using Fission Yeast J Biomol Screen, June 1, 2007; 12(4): 568 - 577. [Abstract] [PDF]
|
|
|
|
 |
|
 B. Kim, H. Koo, S. Yang, S. Bang, Y. Jung, Y. Kim, J. Kim, J. Park, R. T. Moon, K. Song, et al. TC1(C8orf4) Correlates with Wnt/{beta}-Catenin Target Genes and Aggressive Biological Behavior in Gastric Cancer. Clin. Cancer Res., June 1, 2006; 12(11): 3541 - 3548. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Jung, S. Bang, K. Choi, E. Kim, Y. Kim, J. Kim, J. Park, H. Koo, R. T. Moon, K. Song, et al. TC1 (C8orf4) Enhances the Wnt/{beta}-Catenin Pathway by Relieving Antagonistic Activity of Chibby Cancer Res., January 15, 2006; 66(2): 723 - 728. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Wei, M. Kanai, S. Huang, and K. Xie Emerging role of KLF4 in human gastrointestinal cancer Carcinogenesis, January 1, 2006; 27(1): 23 - 31. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. Rieder, A. J. Tessier, X. T. Qiao, B. Madison, D. L. Gumucio, and J. L. Merchant Helicobacter-induced Intestinal Metaplasia in the Stomach Correlates with Elk-1 and Serum Response Factor Induction of Villin J. Biol. Chem., February 11, 2005; 280(6): 4906 - 4912. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J.-M. Kim, H.-Y. Sohn, S. Y. Yoon, J.-H. Oh, J. O. Yang, J. H. Kim, K. S. Song, S.-M. Rho, H. S. Yoo, Y. S. Kim, et al. Identification of Gastric Cancer-Related Genes Using a cDNA Microarray Containing Novel Expressed Sequence Tags Expressed in Gastric Cancer Cells Clin. Cancer Res., January 15, 2005; 11(2): 473 - 482. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Meeting Abstracts Online |
