Search tips
Search criteria 


Logo of neoplasiaGuide for AuthorsAbout this journalExplore this journalNeoplasia (New York, N.Y.)
Neoplasia. 2010 June; 12(6): 499–505.
PMCID: PMC2887090

A Gprc5a Tumor Suppressor Loss of Expression Signature Is Conserved, Prevalent, and Associated with Survival in Human Lung Adenocarcinomas1,2


Increasing the understanding of the impact of changes in oncogenes and tumor suppressor genes is essential for improving the management of lung cancer. Recently, we identified a new mouse lung-specific tumor suppressor—the G protein-coupled receptor 5A (Gprc5a). Microarray analysis of the transcriptomes of lung epithelial cells cultured from normal tracheas of Gprc5a knockout and wild-type mice defined a loss-of-Gprc5a gene signature, which revealed many aberrations in cancer-associated pathways. To assess the relevance of this mouse tumor suppressor to human lung cancer, the loss-of-Gprc5a signature was cross species compared with and integrated with publicly available gene expression data of human normal lung tissue and non-small cell lung cancers. The loss-of-Gprc5a signature was prevalent in human lung adenocarcinomas compared with squamous cell carcinomas or normal lung. Furthermore, it identified subsets of lung adenocarcinomas with poor outcome. These results demonstrate that gene expression patterns of Gprc5a loss in nontumorigenic mouse lung epithelial cells are evolutionarily conserved and important in human lung adenocarcinomas.


The identification of new effective biomarkers will undoubtedly improve clinical management of lung cancer and is tightly linked to a better understanding of the molecular events associated with the development and progression of the disease [1,2]. Both genetic and epigenetic aberrations in oncogenes and tumor suppressor genes have been implicated in lung cancer etiology. Such changes include mutations in KRAS [3], amplification of the epidermal growth factor receptor (EGFR) [4], and its mutation in adenocarcinomas [5,6], mutations in the tumor suppressor p53 [5], and epigenetic silencing of retinoic acid receptor beta (RARβ) [7]. We have previously identified a novel retinoic acid-regulated gene, mGprc5a/hGPRC5A, which is preferentially expressed in mouse and human normal lungs [8]. Earlier attempts at specific targeting of Gprc5a expression in mouse lung have shown that there are no significant developmental defects and that epithelial cell differentiation is normal, and lung structure is intact after lung-specific knockout of the gene and follow-up from early embryonal stages up to 3 months [9]. More recently, our group showed that Gprc5a knockout mice (Gprc5a-/-), when followed up to 2 years, develop spontaneous lung adenocarcinomas at age 12 months and onward, indicating that this gene is a lung-specific tumor suppressor [10]. Moreover, in the same study by our group, the expression of human GPRC5A messenger RNA (mRNA) was analyzed in a publicly available microarray data set and was found to be significantly lower in human lung adenocarcinomas and squamous cell carcinomas (SCCs) relative to normal lung [10].

Genome-wide expression profiling approach has been proven to be a useful method for the discovery of novel cancer subclasses [11–13]. Moreover, comparative genomics by directly comparing expression profiles of experimental mouse models and corresponding human diseases has highlighted conserved expression signatures and networks important for the phenotype under study [14–16]. Therefore, we surmised that information obtained from studying gene expression in the Gprc5a knockout mouse model could help us to begin to understand the molecular consequences of Gprc5a loss that may subsequently provide new insights into human lung cancer expression patterns.

Materials and Methods

WT-NLE and NULL-NLE Normal Epithelial Cells

The Gprc5a knockout mouse was generated previously in our laboratory [10]. Normal lung epithelial cells, WT-NLE and NULL-NLE, were derived from tracheas of mice (C57Bl/6 x 129sv) F1 with wildtype Gprc5a and mice lacking Gprc5a (knockout), respectively. Briefly, tracheas were dissected from 3-week-old Gprc5a WT and Gprc5a knockout mice and were cut into small pieces, which were incubated in a tissue-dissociating solution ACCUMAX from Innovative Cell Technologies (San Diego, CA). The dissociated cells and tissue fragments were then transferred to PRIMARIA tissue culture dishes (BD Biosciences, San Jose, CA) and incubated in AmnioMAX-C100 basal medium (GIBCO, Invitrogen, Grand Island, NY). The epithelial cells were detached by trypsinization, subcultured, and grown in keratinocyte serum-free medium (GIBCO, Invitrogen). The cell lines were karyotyped by G banding in the MD Anderson Institutional Molecular Cytogenetics Facility and were found to be of mouse origin.

RNA Extraction

Total RNA was isolated and purified using RNeasy columns (Qiagen, Valencia, CA). The cells were washed twice with ice-cold PBS, lysed and incubated with DNase I for RNA isolation according to the manufacturer's instructions. RNA quality based on the 28S/18S ribosomal RNA ratio (>1.5) was assessed using the RNA 6000 Nano Lab-Onchip and Agilent 2100 Bioanalyzer device (Agilent Technologies, Palo Alto, CA).

Microarray Sample Preparation, Hybridization, and Scanning

Synthesis of double-stranded complementary DNA was performed using the Superscript Choice system (Invitrogen) using 5 µg of total RNA for each strand. Biotin-labeled complementary RNA were synthesized by in vitro transcription reaction using the ENZO BioArray High-Yield RNA transcript labeling kit (Affymetrix, Santa Clara, CA). Fragmented complementary RNA were then hybridized to GeneChip Mouse Genome 430 2.0 arrays (Affymetrix) according to the manufacturer's instructions. The arrays were scanned with a GeneChip Scanner 3000 from Affymetrix, and raw image files were converted to probe set data (*.CEL files), using the Affymetrix GeneChip Operating Software. Expression microarray data have been submitted to the National Center for Biotechnology Information's Gene Expression Omnibus repository and are MIAME-compliant.

Derivation of a Loss-of-Gprc5a Signature

Raw microarray data files (*.CEL) were imported and analyzed using the BRB-ArrayTools v.3.7.0 developed by Dr. Richard Simon and BRB-ArrayTools Development Team [17]. Robust multiarray analysis [18] was used for normalization of gene expression data in R language environment. Differentially expressed gene features in the loss-of-Gprc5a signature were selected based on the criteria of P < .001 of a random variance t test with permutation and estimation of the false discovery rate and a two-fold difference in expression. The loss-of-Gprc5a signature was also analyzed using ingenuity pathways analysis (IPA; for functional pathways analysis.

Cross-species Analysis and Survival Analysis

To assess the expression of the loss-of-Gprc5a signature in human non-small cell lung cancer (NSCLC) compared with normal lung, we integrated unique orthologous members of the mouse gene signature with expression data from the data sets by Su et al. (27 lung adenocarcinomas and paired adjacent normal lung) [19] and Stearman et al. (20 lung adenocarcinomas and 19 adjacent normal lung; from 10 patients in replicate) [20]. To evaluate the clinical relevance of the mouse loss-of-Gprc5a signature in human lung cancer, we used available human NSCLC microarray data sets from the studies by Shedden et al. (National Cancer Institute Director's Challenge, 442 lung adenocarcinomas; [21], Bild et al. (Duke cohort, 58 adenocarcinomas and 53 SCCs; [22], and Bhattacharjee et al. (Harvard cohort, 125 adenocarcinomas) [12]. Raw microarray data from all data sets were analyzed and normalized using the BRB-ArrayTools. Before integration of the mouse loss-of-Gprc5a signature with the human lung cancer data sets, unique orthologous genes present in both murine (Affymetrix Gene-Chip Mouse Genome 430 2.0) and human (Affymetrix HG-U95A, HG-U95Av2, HG-U133A and HG-U133 plus 2.0) platforms were identified using NetAffx from Affymetrix ( Before integration of the mouse and human orthologous gene expression data into mixed mouse-human data sets, gene expression data were median-centered independently [15]. Hierarchical cluster analysis by average linkage was performed with Cluster 2.11, and results were visualized with TreeView programs (Michael Eisen Laboratory, Kaplan-Meier and log-rank test survival analyses while censoring for patients and based on Gprc5a-WT and-NULL clusters after hierarchical clustering analysis were performed using the R language environment v2.8.0 (

To predict class, we adopted a previously developed model using six algorithms, namely, compound covariate predictor, linear discriminator analysis, nearest neighbors 1 and 3, nearest centroid, and support vector machine [15]. Lung adenocarcinomas from the study of Shedden et al. were used as a training set (n = 442), and adenocarcinomas from the Duke and Harvard cohorts were pooled as a test set (n = 183; DH cohort). An optimized classifier list was generated using a leave-one-out cross-validation approach. The six classification algorithms were then applied to the test set, and survival analysis was performed to assess the clinical significance of predicted Gprc5a-WT and Gprc5a-NULL groups by Kaplan-Meier survival analyses while censoring for patients and log-rank tests in R language environment.


Biological Characteristics of a Mouse Loss-of-Gprc5a Gene Signature in Lung Epithelial Cells Derived from Gprc5a Knockout Mice

We performed global gene expression analysis on the transcriptome of NLE cells isolated from Gprc5a wild-type and knockout mice, WT-NLE and NULL-NLE cells, respectively (Figure 1A) to understand potential molecular consequences of the loss of Gprc5a. A gene expression signature reflecting the loss of Gprc5a was found to be composed of 1586 gene features differentially expressed between these cells based on selection criteria of P < .001 of the univariate t test and a difference in expression by at least two-fold (Figure 1B). Functional analysis using IPA revealed the significant modulation of canonical cancer-related gene sets and pathways, such as cell death, cell growth and proliferation, cell cycle, growth factor and receptor signaling, survival, inflammation, tumor suppressor p53 signaling, and Wnt/β-catenin signaling pathways (all P < .001; Figure 1, C and D, and Table W1). Moreover, Gprc5a knockout cells displayed increased activation (marked by the red arrow) of oncogenic biological processes and pathways such as cell growth and proliferation, cell cycle, inflammation, and Wnt/β-catenin signaling pathway compared with wild-type cells (Figure 1, C and D). In addition, deactivation or inhibition (marked by blue arrow) of tumor-suppressive biological functions and pathways such as cell death, p53 signaling, and the G2/M cell cycle-controlling checkpoint was evident in the Gprc5a knockout cells (Figure 1, C and D). These results indicate that loss of Gprc5a in lung epithelial cells perturbs expression of genes strongly associated with cancer and important in mouse lung carcinogenesis.

Figure 1
Derivation of a mouse loss-of-Gprc5a signature. (A) Normal epithelial cells cultured fromtracheas of Gprc5a wild-type and knockout mice (WT-NLE and NULL-NLE, respectively) were used for global gene expression analysis to understand molecular consequences ...

Mouse Loss-of-Gprc5a Gene Signature Discriminates Human Lung Adenocarcinomas from Adjacent Normal Tissues

The association of Gprc5a loss with canonical cancer-related pathways prompted us to address the biological relevance of the loss-of-Gprc5a signature to human NSCLC. By using only orthologous genes present in both mouse and human microarray platforms, we first integrated the loss-of-Gprc5a expression signature with gene expression data of human lung adenocarcinomas and adjacent normal counterparts available from previous studies [19,20]. Hierarchical clustering and principal component analyses of the integrated data revealed that the mouse loss-of-Gprc5a signature was tightly clustered and associated with human lung adenocarcinomas relative to normal lung (Figures 2, A and B, and W1, A and B). In addition, expression of GPRC5A was significantly lower in lung adenocarcinomas than in adjacent normal tissues in both human data sets (Figure 2C).

Figure 2
The mouse loss-of-Gprc5a signature differentiates human lung adenocarcinomas from normal lung or SCCs. Dendrograms of hierarchical cluster analysis of mixed mouse-human data sets including lung adenocarcinomas and adjacent normal lung samples from the ...

Mouse Loss-of-Gprc5a Signature Is Significantly Associated with Human Lung Adenocarcinomas than with SCCs

The mouse loss-of-Gprc5a expression signature was also integrated with human gene expression data containing 58 lung adenocarcinomas and 53 SCCs available from the study by Bild et al. (Duke cohort) [22]. In the hierarchical cluster analysis of the composite data, 51 of 58 adenocarcinomas coclustered with the loss-of-Gprc5a signature, whereas 43 of 53 SCCs samples were in the Gprc5a-WT cluster (Figure 2D and Table W2; P = 2.8 x 10-13 of the χ2 test). The closer association of the mouse loss-of-Gprc5a signature with lung adenocarcinomas rather than SCCs was also evident in the multidimensional space (Figure 2E).

We next explored the possibility that the loss-of-Gprc5a signature exhibits prognostic properties in human lung adenocarcinoma. The 58 lung adenocarcinomas patients were dichotomized on the basis of their similarity of gene expression to the loss-of-Gprc5a signature after hierarchical clustering alone and independent of the SCC samples (clustering heat maps not shown). Lung adenocarcinoma patients who expressed the loss-of-Gprc5a signature exhibited a significantly worse overall survival (OS) when compared with those who did not express the signature (P = .02 of the log-rank test; Figure 2F, left). Conversely, when the mouse loss-of-Gprc5a signature was integrated with human lung SCCs only, no significant differences in OS were found (P = .3 of the log-rank test; Figure 2F, right) between the cluster of SCC patients expressing the signature (NULL cluster) and the cluster that did not (WT cluster) after hierarchical cluster analysis (heat map not shown).

The expression of the mouse loss-of-Gprc5a gene signature was then assessed in a larger cohort of human lung adenocarcinomas, the NCI Director's Challenge study [21]. After hierarchical cluster analysis of the integrated data (Figure 3A), patients expressing the loss-of-Gprc5a gene signature displayed worse OS (P = .00003) than those lacking the signature (Figure 3B). To validate the association of loss-of-Gprc5a signature with poor survival, the NCI Director's Challenge data sets were used as a training set (n = 442), and gene expression data of adenocarcinomas from the Duke [22] and Harvard [12] cohorts (n = 183) were pooled as a validation set (DH cohort). The number of genes was optimized to minimize misclassification during a leave-one-out cross-validation approach (Figure 3C and Table 1). Various prediction algorithms showed that the OS of human lung adenocarcinoma patients in the DH validation cohort predicted to harbor the mouse loss-of-Gprc5a signature was significantly poorer than that of patients predicted to lack the signature (Figure 3D).

Figure 3
The mouse loss-of-Gprc5a gene signature is associated with poor survival in human lung adenocarcinoma. (A) The mouse loss-of-Gprc5a gene signature was integrated with and analyzed in the NCI Director's Challenge data sets (n = 748) by hierarchical cluster ...
Table 1
Performance of the Mouse Loss-of-Gprc5a Signature by Different Prediction Algorithms.


Pathway signatures representing aberrant activation of oncogenes or inactivation of tumor suppressor genes have provided important insights into lung cancer development [23]. Having recently identified a new mouse lung tumor suppressor gene (Gprc5a) [10], we surmised that characterization of the differential gene expression patterns of epithelial cells isolated from lungs of Gprc5a wild-type and knockout mice will improve the understanding of the molecular characteristics of loss of Gprc5a during the development of lung cancer. The results of this comparison have led to the derivation of a mouse loss-of-Gprc5a gene signature. This signature suggested that loss of Gprc5a is well associated with cancer-related molecular processes and pathways, including growth factor and receptor signaling, cell death, cell cycle, survival, cell growth and proliferation, and inflammation. The large variety of genes and pathways affected by the loss of Gprc5a expression attests to the pivotal role that this lung-specific tumor suppressor plays in mouse lung carcinogenesis.

We then hypothesized that the mouse loss-of-Gprc5a signature could be valuable in improving the understanding of differential gene expression patterns related to human lung carcinogenesis. This hypothesis is based in part on the assertion that conserved gene expression signatures resembling similar phenotypes in different species are functionally important for the specific phenotype [24]. Using comparative functional genomic approaches and integrating the mouse loss-of-Gprc5a signature with published human lung (normal and malignant) gene expression data, we found that the signature is 1) highly conserved in human NSCLC and sufficient for discriminating human lung adenocarcinomas from adjacent normal lung tissues, 2) better associated with gene expression patterns of lung adenocarcinomas than of SCCs, and 3) associated with poor prognosis of human lung adenocarcinoma patients.

We previously demonstrated by in silico analysis in a published data set by Bhattacharjee et al. [12] that GPRC5A mRNA is significantly decreased in lung adenocarcinomas, SCCs, carcinomas, and bronchioalveolar carcinomas compared with normal lung, suggesting that GPRC5A loss of expression is not cell lineage-specific. Importantly, both the Gprc5a knockout (NULL) and wild-type cells (WT), from which the loss-of-Gprc5a signature was derived, had been isolated from trachea of 3-month-old Gprc5a knockout and wild-type mice, respectively, and because we used the exact same isolation and culture conditions, we assume that the cells are from the same lineage. Therefore, we thought it was justifiable to integrate and cross species analyze the loss-of-Gprc5a signature in the normal mouse cells with both adenocarcinomas and SCCs. Although the signature is based on differences between epithelial cells derived from normal tracheas of wild-type and Gprc5a knockout mice, it was prevalent in human lung adenocarcinomas that develop from small airways in the lung periphery rather than SCCs that are derived from bronchial upper airway epithelial cells. It is unclear why, despite decreased GPRC5A mRNA levels in both subtypes of NSCLC, gene expression patterns downstream of Gprc5a loss are more relevant to human lung adenocarcinomas rather than to SCCs. The reason may be that Gprc5a knockout mice develop adenomas and adenocarcinomas and not SCCs [10]. Interestingly, Hassan et al. [25] recently demonstrated that an embryonic stem cell-like signature identified poorly differentiated lung adenocarcinoma with dismal prognosis but was not relevant for SCCs, suggesting that biological pathway-specific gene signatures may be differentially expressed and relevant in different subtypes of NSCLC.

We have previously analyzed publicly available transcriptome data that have indicated statistically significant decreases in mRNA levels of GPRC5A in adenocarcinoma (n = 139) and SCC (n = 21) compared with normal lung (n = 17) [10]. We do not know the mechanism of the decrease in GPRC5A expression, but it cannot be accounted for by gene loss because the frequency of chromosome 12p12.3 loss is low in NSCLC tumors. The expression is likely silenced by epigenetic mechanisms because we found that GPRC5A can be reexpressed after treatment of lung cancer cell lines with histone deacetylase inhibitors (unpublished observations). Interestingly, no mutations had been identified in the GPRC5A gene in samples from 44 human cancers (22 breast tumors, 11 colorectal tumors, and 11 glioblastomas) [26,27] as well as in 20 human NSCLC cell lines we have analyzed (unpublished observations).

In conclusion, we demonstrated the molecular similarity of the Gprc5a knockout mouse model to its human counterpart, indicating that it recapitulates fundamental features of human lung adenocarcinoma. Our study strongly suggests that loss of Gprc5a in the lung might influence the clinical outcome of lung adenocarcinoma. Thus, gene expression patterns associated with loss of Gprc5a would be valuable for the identification of novel biomarkers and targets for intervention.

Supplementary Material

Supplementary Figures and Tables:


1This study was supported in part by Department of Defense grant W81XWH-06-1-0303 (to R.L.), the Samuel Waxman Cancer Research Foundation (to R.L.), the Jimmy LaneHewlett Fund for Lung Cancer Research (to R.L.), the Cancer Center Support grant P30 CA16672 (Affymetrix Core), and the 21C Frontier Functional Human Genome Project grant FG-4-2 of MEST (to J.-S.L.). The authors declare no conflict of interest.

2This article refers to supplementary materials, which are designated by Tables W1 and W2 and Figure W1 and are available online at


1. Meyerson M, Carbone D. Genomic and proteomic profiling of lung cancers: lung cancer classification in the age of targeted therapy. J Clin Oncol. 2005;23:3219–3226. [PubMed]
2. Xie Y, Minna JD. Predicting the future for people with lung cancer. Nat Med. 2008;14:812–813. [PMC free article] [PubMed]
3. Westra WH, Baas IO, Hruban RH, Askin FB, Wilson K, Offerhaus GJ, Slebos RJ. K-ras oncogene activation in atypical alveolar hyperplasias of the human lung. Cancer Res. 1996;56:2224–2228. [PubMed]
4. Weir BA, Woo MS, Getz G, Perner S, Ding L, Beroukhim R, Lin WM, Province MA, Kraja A, Johnson LA, et al. Characterizing the cancer genome in lung adenocarcinoma. Nature. 2007;450:893–898. [PMC free article] [PubMed]
5. Ding L, Getz G, Wheeler DA, Mardis ER, McLellan MD, Cibulskis K, Sougnez C, Greulich H, Muzny DM, Morgan MB, et al. Somatic mutations affect key pathways in lung adenocarcinoma. Nature. 2008;455:1069–1075. [PMC free article] [PubMed]
6. Pao W, Miller V, Zakowski M, Doherty J, Politi K, Sarkaria I, Singh B, Heelan R, Rusch V, Fulton L, et al. EGF receptor gene mutations are common in lung cancers from “never smokers” and are associated with sensitivity of tumors to gefitinib and erlotinib. Proc Natl Acad Sci USA. 2004;101:13306–13311. [PubMed]
7. Xu XC, Lee JS, Lee JJ, Morice RC, Liu X, Lippman SM, Hong WK, Lotan R. Nuclear retinoid acid receptor beta in bronchial epithelium of smokers before and during chemoprevention. J Natl Cancer Inst. 1999;91:1317–1321. [PubMed]
8. Cheng Y, Lotan R. Molecular cloning and characterization of a novel retinoic acid-inducible gene that encodes a putative G protein-coupled receptor. J Biol Chem. 1998;273:35008–35015. [PubMed]
9. Xu J, Tian J, Shapiro SD. Normal lung development in RAIG1-deficient mice despite unique lung epithelium-specific expression. Am J Respir Cell Mol Biol. 2005;32:381–387. [PubMed]
10. Tao Q, Fujimoto J, Men T, Ye X, Deng J, Lacroix L, Clifford JL, Mao L, Van Pelt CS, Lee JJ, et al. Identification of the retinoic acid-inducible Gprc5a as a new lung tumor suppressor gene. J Natl Cancer Inst. 2007;99:1668–1682. [PubMed]
11. Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A, Boldrick JC, Sabet H, Tran T, Yu X, et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature. 2000;403:503–511. [PubMed]
12. Bhattacharjee A, Richards WG, Staunton J, Li C, Monti S, Vasa P, Ladd C, Beheshti J, Bueno R, Gillette M, et al. Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses. Proc Natl Acad Sci USA. 2001;98:13790–13795. [PubMed]
13. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP, Coller H, Loh ML, Downing JR, Caligiuri MA, et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science. 1999;286:531–537. [PubMed]
14. Bennett CN, Green JE. Unlocking the power of cross-species genomic analyses: identification of evolutionarily conserved breast cancer networks and validation of preclinical models. Breast Cancer Res. 2008;10:213. [PMC free article] [PubMed]
15. Lee JS, Chu IS, Mikaelyan A, Calvisi DF, Heo J, Reddy JK, Thorgeirsson SS. Application of comparative functional genomics to identify best-fit mouse models to study human cancer. Nat Genet. 2004;36:1306–1311. [PubMed]
16. Sweet-Cordero A, Mukherjee S, Subramanian A, You H, Roix JJ, Ladd-Acosta C, Mesirov J, Golub TR, Jacks T. An oncogenic KRAS2 expression signature identified by cross-species gene-expression analysis. Nat Genet. 2005;37:48–55. [PubMed]
17. Simon R, Lam A, Li MC, Ngan M, Menenzes S, Zhao Y. Analysis of gene expression data using BRB-Array Tools. Cancer Inform. 2007;3:11–17. [PMC free article] [PubMed]
18. Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, Speed TP. Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res. 2003;31:e15. [PMC free article] [PubMed]
19. Su LJ, Chang CW, Wu YC, Chen KC, Lin CJ, Liang SC, Lin CH, Whang-Peng J, Hsu SL, Chen CH, et al. Selection of DDX5 as a novel internal control for Q-RT-PCR from microarray data using a block bootstrap re-sampling scheme. BMC Genomics. 2007;8:140. [PMC free article] [PubMed]
20. Stearman RS, Dwyer-Nield L, Zerbe L, Blaine SA, Chan Z, Bunn PA, Jr, Johnson GL, Hirsch FR, Merrick DT, Franklin WA, et al. Analysis of orthologous gene expression between human pulmonary adenocarcinoma and a carcinogen-induced murine model. Am J Pathol. 2005;167:1763–1775. [PubMed]
21. Shedden K, Taylor JM, Enkemann SA, Tsao MS, Yeatman TJ, Gerald WL, Eschrich S, Jurisica I, Giordano TJ, Misek DE, et al. Gene expression- based survival prediction in lung adenocarcinoma: a multi-site, blinded validation study. Nat Med. 2008;14:822–827. [PMC free article] [PubMed]
22. Bild AH, Yao G, Chang JT, Wang Q, Potti A, Chasse D, Joshi MB, Harpole D, Lancaster JM, Berchuck A, et al. Oncogenic pathway signatures in human cancers as a guide to targeted therapies. Nature. 2006;439:353–357. [PubMed]
23. van't Veer LJ, Bernards R. Enabling personalized cancer medicine through analysis of gene-expression patterns. Nature. 2008;452:564–570. [PubMed]
24. Lee JS, Thorgeirsson SS. Comparative and integrative functional genomics of HCC. Oncogene. 2006;25:3801–3809. [PubMed]
25. Hassan KA, Chen G, Kalemkerian GP, Wicha MS, Beer DG. An embryonic stem cell-like signature identifies poorly differentiated lung adenocarcinoma but not squamous cell carcinoma. Clin Cancer Res. 2009;15:6386–6390. [PMC free article] [PubMed]
26. Parsons DW, Jones S, Zhang X, Lin JC, Leary RJ, Angenendt P, Mankoo P, Carter H, Siu IM, Gallia GL, et al. An integrated genomic analysis of human glioblastoma multiforme. Science. 2008;321:1807–1812. [PMC free article] [PubMed]
27. Sjoblom T, Jones S, Wood LD, Parsons DW, Lin J, Barber TD, Mandelker D, Leary RJ, Ptak J, Silliman N, et al. The consensus coding sequences of human breast and colorectal cancers. Science. 2006;314:268–274. [PubMed]

Articles from Neoplasia (New York, N.Y.) are provided here courtesy of Neoplasia Press