Characterization of aberrant pathways across human cancers
© Ylipää et al; licensee BioMed Central Ltd. 2013
Published: 12 August 2013
Cancer is a broad group of genetic diseases which account for millions of deaths worldwide each year. Cancers are classified by various clinical, pathological and molecular methods, but even within a well-characterized disease, there is a significant inter-patient variability in survival, response to treatment, and other parameters. Especially in molecular level, tumours of the same category can appear significantly dissimilar due to complex combinations of genetic aberrations leading to a similar malignancy. We extended the current classification methods by studying tumour heterogeneity at pathway level.
We computed the rate of alterations in 1994 pathways and 2210 tumours consisting of eight different cancers. Using gene set enrichment analysis, each sample was computed a pathway aberration profile that reflected its molecular state. The profiles were analysed together to infer the characteristic aberration rates for each pathway within each cancer. Subgroups of tumours defined by similar pathway aberrations were identified using clustering analyses. The pathway aberration and gene expression profiles of the subgroups were consecutively compared across all eight cancer types to search for similar tumours crossing the standard classification.
We identified pathways and processes that were common to all cancers as well as traits that are unique to a cancer type or closely related cancers. Studying the gene expression patterns within the pathway context suggested potential alteration mechanisms. Clustering analysis revealed five clinically relevant subgroups of tumours in four cancers that exhibited significant differences in survival compared to others. The cross-cancer analysis of the subgroups resulted in the identification of tumours that shared potentially significant alterations.
This study represents the first effort to extend the molecular characterizations towards pathway level descriptions across the family of cancers. In addition to providing a proof-of-concept for single sample pathway aberration analysis in this context, we present a comprehensive pathway aberration dataset that can be used to study pathway aberration patterns within or across cancers. Significant similarities between subgroups of different cancers on pathway and gene expression levels provide interesting hypotheses for understanding variable drug response, or transferring treatments across diseases by identifying common druggable pathways or genes, for example.
The development of cancer is an evolutionary process that is driven by the acquisition of somatic genetic mutations which give cells a selective advantage against non-mutated cells . In order to become malignant cancer cells, normal cells need to acquire a set of mutations which confer "hallmark" traits, such as increased proliferation, immortality, and invasiveness . Usually, a single mutation is not enough to result in malignant growth, but there are plenty of different combinations of mutations which can alter the expression biochemical pathways leading to the same phenotypic effect . Acquiring these traits can be better described and understood as alterations in the balance of interaction networks of genes, proteins and other molecules, or pathways. Cancers can also be divided into clinically meaningful subtypes based on their gene expression patterns that may be indicative of response to a treatment, like Her2 positive breast cancers  or KIT positive gastro-intestinal tumours , for example. Many cancers have characteristic sets of somatic mutations that can be used to identify and classify the tumours ; few studies have even compared these across cancer types . However, cancers are not commonly classified or studied based on the acquired traits or alterations to the pathways because of the added complexity. Instead, pathway level changes are often concluded for the tumour subgroups that have first been identified by other means. Currently, most established cancer types and cancer grading systems (such as Gleason score for prostate cancers  and WHO grades for tumours of the central nervous system ) are not even based on genetic markers but instead on clinical parameters and phenotypic observations.
Multi-institutional projects, such as The Cancer Genome Project (TCGA) and International Cancer Genome Consortium (ICGC), are already improving the cancer classification by systematically gathering and analysing unprecedented amounts of microarray and deep sequencing data from multiple cancers. These data have been used to identify clinically meaningful subtypes based on genomic and transcriptomic [9–16] or epigenomic profiles [17, 18]. The standard approach of clustering samples based on only one type of data has recently been extended towards integration of multiple data types [19, 20]. Common practice is to follow up genomic and transcriptomic analyses by probing the frequently altered pathways in each identified subtype to infer the unique systems level characteristics of each subtype. Altered pathways can be identified by statistically combining knowledge on pathways' constituent genes, and their genomic (mutations, copy numbers) and/or transcriptomic (expression) state in the tumours . Recent increase in availability of microarray and deep sequencing data has made it also possible to identify the extent and the frequency at which pathways are aberrant in different cancers and cancer subtypes. There is currently great interest in extending characterizations and subtyping into systems level. One of the main goals is to improve poor drug response rates by matching drugs with the specific pathway alterations of the patient's cancer subtype .
We hypothesize that classifying tumours based on clinical, phenotypic, or genomic markers may not be as informative as using pathway alterations since different tumours may appear similar and the molecular mechanisms to malignancy are undoubtedly variable. Furthermore, it is extremely difficult to predict the effect of DNA level changes (e.g. mutations, copy number changes, methylation levels) to the phenotype, and therefore we analysed data in the context that is as close to the phenotype as possible, the pathways. In this paper, we analyse multiple TCGA expression datasets from systems perspective. Using pathway data from five databases (1994 pathways) and expression data from eight cancers (2210 samples), we first infer the aberrant pathways in individual tumours, thus defining their pathway aberration profiles. Based on these profiles, we define a comprehensive catalogue of pathway alterations and their frequencies in respective cancers. By clustering the pathway aberration profiles, we are able to uncover clinically meaningful subtypes of cancers that have not been reported from TCGA cancer types by earlier studies. Finally, we compare the subgroups of different cancers together to find unexpected similarities on both pathway and gene expression level.
Gene expression data analysis and quality control
Number of samples
Lung squamous cell carcinoma
Uterine corpus endometrial adenocarcinoma
Kidney clear cell renal carcinoma
Acquisition of the pathway information
Number of pathways
Kyoto Encyclopedia of Genes and Genomes
GeneOntology Biological Processes
Computing pathway aberration profiles
To arrive with pathway aberration scores corresponding to enrichment and depletion of each pathway, we computed gene set enrichment scores inspired by the GSEA method by Subramanian et al.  for each pathway in each sample individually. These scores reflect the degree to which a pathway's genes are up- or downregulated compared to a reference in a sample. First we rank ordered the list of all measured genes based on their normalized expression difference against a reference (as described above). Then, walking-down the list of genes, we label each gene as 1 if it belongs to the pathway or 0 if it does not belong to the pathway. Starting from the most upregulated gene results in a score for enrichment and starting from the most downregulated gene a score for depletion. Drawing analogy to analysis of Receiver Operating Characteristic (ROC) curves, we derive the fraction of 1's vs. the fraction of 0's at each position of the list and compute the Area Under Curve (AUC) statistic which describe how far up or down the list are the pathway's genes. To estimate the statistical significance of the AUC, we permute the list of ordered genes and recompute the statistic for the permuted data 1000 times to generate a null distribution for the AUC. This method for generating null distributions was chosen, because in the single sample analysis, more complex models are difficult to justify as there is no meaningful way to evaluate correlations with phenotypes. Empirical confidence scores of the observed AUC are then calculated relative to the null distribution. By iterating the algorithm for each sample and each pathway individually, we obtained two scores for each sample-pathway combination, corresponding enrichment and depletion. Smaller value of the score indicated more significant trend. In order to avoid being overly conservative for an exploratory method, we chose not to control for the amount of false positives due to multiple testing problem. Combining the enrichment and depletion scores for each pathway in each sample, we created pathway aberration profiles that describe all the pathway aberrations in each sample. This was done by taking the smaller of enrichment and depletion scores, and transforming it into log2 space if it was enrichment and -log2 space if it was depletion.
Clustering of pathway aberration profiles
To investigate the similarities and differences between cancers on pathway level we hierarchically clustered the means of pathway aberration profiles with Euclidean distance metric and Ward's linkage method. To identify homogeneous subtypes, a two way clustering of the aberration profiles across samples and pathways was done using hierarchical clustering with the same distance metric and linkage method. Distinct branches of 20-30 samples were identified from the dendrogram and further studied as subgroups. Clustering of the identified subgroups was done using the same methodology and features consisting of mean enrichment and depletion frequencies within each subgroup.
Associations between patient survival and subtype were computed with Mantel-Cox test of difference of Kaplan-Meier survival estimators. Associations between the subgroups and previous molecular subtype characterizations were computed using Fisher's exact test. Differential gene expressions were computed using Wilcoxon rank sum test. A p -value of 0.05 was considered the threshold of statistical significance in all tests. All analyses were made with Matlab version R2010b (MathWorks, Natick, MA).
Results and discussion
Identification of common and disease-specific sets of pathway aberrations in eight cancers
The 2210 gene expression profiles consisted of eight different cancer types (see Table 1) provided by the TCGA. Three cancers are represented by over 500 samples each, two cancers by at least 150 samples, and three cancers are represented by at least 50 samples each. Principal component analysis indicated that the gene expression profiles of colon and rectal carcinomas are very much alike, similarly as ovarian and uterine tumours, whereas glioblastoma has the most distinct gene expression pattern (Figure 1a). A very small subset from all cancers, except endometrial, closely resembled each other in gene expression level (blow-up panel in Figure 1a). Closer inspection revealed that these were data from only two patients according to patient ID numbers. The data from these patients were removed as well as all duplicate samples so that each sample represented a unique case.
Using a collection of 1994 pathways from five different databases (see Table 2), we investigated the similarities and differences between cancers in pathway level. Instead of pooling the data from each cancer type first, we computed the enriched and depleted pathways for each sample individually (Additional File 1). We then combined the pathway enrichment and depletion scores into pathway aberration profiles for each cancer type (see Methods), and hierarchically clustered them (Figure 1b). On pathway level, colon and rectal carcinomas have very few differences (blue branch). Also, the gynaecological cancers (BRCA, UCEC, and OV) clustered closely together (red branch), as could be expected. By comparing the average aberration rates across all cancer types, we observed that many of the biological processes considered as cancer hallmarks  are frequently aberrant in all tumour types. For example, GO terms Inflammatory response (81%), Immune response (80%), Cell-to-cell signalling (77%), Cell-to-cell adhesion (76%), and Cellular homeostasis (67%) were among the most frequently enriched processes, whereas DNA replication (98%), Regulation of cell cycle (98%), were among the most frequently depleted (Additional File 2).
However, by ranking the pathways according to their variability in alteration frequencies between cancers, we identified several pathways that were altered very frequently within one or few cancer types and only rarely in other cancers. These pathways can give rise to the observed differences in physiological and phenotypic properties across cancers, or they may only reflect the differences between host tissues or cells of origin, especially since there were no tissue-specific normal references available. For example, GO term Nervous system development which is enriched in 99.5% of GBM's, but hardly ever in other tumour types, is likely a cell type specific pathway rather than malignant alteration (Figure 1c). However, we identified processes that we think are actually more related to cancer than normal cell physiology. For example, platelet activation is enriched in 87% of kidney carcinomas, but only in 6% of uterine adenocarcinomas, which may translate to differences in tumour haemostatic activity or formation of cancer metastases through emergence of platelet-tumour cell aggregates  (Figure 1d). Another potentially interesting observation is the major difference in enrichment of the drug metabolism by Cytochrome p450 pathway which is closely related to multiple drug resistance, and also represents a potential therapeutic target . It is highly enriched in cancers of the kidney, lung, colon and rectum (64%-89%), lowly enriched in gynaecologic cancers (ovarian, breast, and endometrial) (15%-26%) and never in glioblastoma. In contrast, the pathway is actually depleted in 56% of GBMs (Figure 1e). There is also a significant difference in the enrichment and depletion frequencies of Irinotecan pathway which describes the biotransformation of the chemotherapy prodrug irinotecan to form the active metabolite which inhibits DNA topoisomerase I . The drug is used in the treatment of many different cancers, but there is large interpatient variability in response to it. The pathway is very frequently depleted in ovarian and breast cancers (81% and 89%, respectively, but rarely in cancers of kidney, colon, and rectum (4%, 8%, and 7%, respectively) (Figure 1f).
Pathway aberration profiles identify clinically significant subtypes in glioblastoma, breast cancer, colon cancer, and ovarian cancer
In Figure S8ainAdditional File 4, we show the mean pathway aberration profiles for five pathways that are differentially aberrant in the three KIRC subgroups. The expression levels of the glycolysis-related genes indicate how exactly this pathway is depleted in KIRC subgroup 2 and enriched in subgroup 3. DPYS, UPB1 and PPAP2B are consistently upregulated in samples where glycolysis pathway is not enriched, whereas a group of eight downregulated genes (ACLY, PFKFB3, MLXIPL, SLC35D1, TPI1, RPIA, AMPD2, and PFKB4) are a characteristic of the glycolysis-enriched subgroup 3 in addition to a fair amount of upregulated genes, including subunits of NADH dehydrogenase complex, ATP synthase complex, cytochrome c oxidase, and ubiquinol-cytochrome reductase. In FigureS8b inAdditional File 4, tumours in the more lethal OV subgroup 12 were particularly enriched in metabolic, immune response, transcriptional and translational pathways. Metabolic and immune pathways were also enriched in the more lethal BRCA subgroup 14 (in FigureS8c inAdditional File 4). BRCA subgroups 3 and 4 that consisted of a significant portion of the Basal-like tumours were highly enriched in adaptive immune system processes (Adaptive immune response GO category enriched in 92% of the tumours in this subgroup compared to 20% in other tumours) such as Lymphocyte activation, and TNFa/NFkB signalling , which may have significant clinical implications. Other pathway-level features of the subgroup included less frequently enriched secretion, and more frequently depleted catabolic pathways. The more lethal COAD subgroup 6 was enriched of transcriptional and metabolic pathways (in FigureS8d inAdditional File 4). Collectively, some of the most variably aberrant pathways in these cancers included metabolic pathways such as oxidative phosphorylation, transcriptional and translational pathways, immune system related pathways, processes such as haemostasis, apoptosis and cell proliferation, and signalling pathways such as TNFa/NFkB signalling (Figure S8a-g inAdditional File 4). This may indicate that not all of these processes are necessary to the cancer cells, or that there exist alternative molecular mechanisms to acquiring the phenotypes that are described by these pathways.
A family tree of cancer: Pathway level comparison reveals functionally similar subgroups across cancer types
In the four mixed branches (lime, green, red, and blue), we compared the pathway aberration differences between the mixed tumours to the tumours of the same type in other branches. In the upper left corner of Figure 5, we show six pathways that are differently aberrant between the five OV subgroups and BRCA subgroup 10 in the lime branch comparing to the rest of OV and BRCA subgroups. Immune response and selenium metabolism were the most strikingly differently aberrant pathways. For the six pathways, we show the expression differences of few differently expressed genes (p <0.01) which may provide clues to understanding why the pathways are altered. For example in immune response pathway, CD2 and FASLG are not overexpressed in the lime branch subgroups compared to other BRCA and OV tumors. Similarly BTK is underexpressed in the lime branch subgroups.
In the green branch, COAD1 and BRCA11 share the depleted TNFa/NFkB signaling pathway, and enriched integrin adhesion pathway in comparison to other BRCA and COAD tumors. Significant differences in some of the genes in the selected pathways are observed, for example, in TNFa/NFkB pathway AKAP12 is underexpressed in BRCA and COAD tumors excluding COAD1/BRCA11 where GNAI1 and GNA11 are overexpressed. The red branch consisted of three types of cancers that share the enrichment in integrin adhesion, immune system, B cell receptor signaling, IL-4 signaling, and adherens junctions in contrast to the rest of the BRCA, OV, and LUSC tumors. Interestingly, TGFB1, TGFB2, TGFBR1, and TGFBR2 are all upregulated in these tumors. Subgroups of OV and BRCA in the blue branch are characterized by the frequently depleted apoptosis and beta-catenin phosphorylation pathways.
Following huge efforts to measure the genomes and transcriptomes of different cancers, this study represents the first effort to extend the current molecular characterizations towards comparative and pathway level descriptions across the family of cancers. Studying large collections of tumour samples at pathway level enabled us to create a comprehensive catalogues of altered pathways from where we inferred the characteristic aberrations for each cancer. As such, this study is also the first proof-of-concept study for utilizing single sample pathway aberration analysis in this context. Importantly, our approach adds another layer of information on top of the classical markers retaining the option to study gene expression or other genomic features in the context of pathways as well. Based on the pathway aberration profiles alone, we identified clinically significant subtypes of glioblastoma, breast cancer, colon cancers, and ovarian cancer. In contrast to subtypes identified using genomic data, phenotypic characteristics of our subtypes can be easily hypothesized from their unique pathway aberrations. We also identified significant similarities between subgroups of different cancers on pathway and gene expression levels which provide interesting avenues for understanding variable drug response or transferring treatments across cancer types by identifying common druggable pathways or genes, for example. These results demonstrate the applicability of our approach, and the value of the aberration data as a resource for future investigations where integrating e.g. copy number, mutation, and epigenetic data to our results should provide plenty of intriguing insight.
The results published here are in whole or part based upon data generated by The Cancer Genome Atlas pilot project established by the NCI and NHGRI (dbGaP Study Accession: phs000178.v8.p7). Information about TCGA and the investigators and institutions who constitute the TCGA research network can be found at http://cancergenome.nih.gov/. The authors express their gratitude to Matti Annala for acquiring the data.
Publication of this article was supported by the following agencies and grants: Academy of Finland (project no. 132877, 251790), Emil Aaltonen Foundation, Finnish Funding Agency for Technology and Innovation Finland Distinguished Professor Programme, and National Cancer Institute grant no. U24CA143835.
This article has been published as part of BMC Systems Biology Volume 7 Supplement 1, 2013: Selected articles from the 10th International Workshop on Computational Systems Biology (WCSB) 2013: Systems Biology. The full contents of the supplement are available online at http://www.biomedcentral.com/bmcsystbiol/supplements/7/S1.
- Vogelstein B, Kinzler KW: Cancer genes and the pathways they control. Nat Med. 2004, 10 (8): 789-99. 10.1038/nm1087.View ArticlePubMedGoogle Scholar
- Hanahan D, Weinberg RA: Hallmarks of cancer: the next generation. Cell. 2011, 144 (5): 646-74. 10.1016/j.cell.2011.02.013.View ArticlePubMedGoogle Scholar
- Vogelstein B, Lane D, Levine AJ: Surfing the p53 network. Nature. 2000, 408 (6810): 307-10. 10.1038/35042675.View ArticlePubMedGoogle Scholar
- Bange J, Zwick E, Ullrich A: Molecular targets for breast cancer therapy and prevention. Nat Med. 2001, 7 (5): 548-52. 10.1038/87872.View ArticlePubMedGoogle Scholar
- Heinrich MC, Blanke CD, Druker BJ, Corless CL: Inhibition of KIT tyrosine kinase activity: a novel molecular approach to the treatment of KIT-positive malignancies. J Clin Oncol. 2002, 20 (6): 1692-703. 10.1200/JCO.20.6.1692.View ArticlePubMedGoogle Scholar
- Bell DW: Our changing view of the genomic landscape of cancer. J Pathol. 2010, 220 (2): 231-43.PubMed CentralPubMedGoogle Scholar
- Beroukhim R, Mermel CH, Porter D, Wei G, Raychaudhuri S, Donovan J, Barretina J, Boehm JS, Dobson J, Urashima M, Mc Henry KT, Pinchback RM, Ligon AH, Cho YJ, Haery L, Greulich H, Reich M, Winckler W, Lawrence MS, Weir BA, Tanaka KE, Chiang DY, Bass AJ, Loo A, Hoffman C, Prensner J, Liefeld T, Gao Q, Yecies D, Signoretti S, Maher E, Kaye FJ, Sasaki H, Tepper JE, Fletcher JA, Tabernero J, Baselga J, Tsao MS, Demichelis F, Rubin MA, Janne PA, Daly MJ, Nucera C, Levine RL, Ebert BL, Gabriel S, Rustgi AK, Antonescu CR, Ladanyi M, Letai A, Garraway LA, Loda M, Beer DG, True LD, Okamoto A, Pomeroy SL, Singer S, Golub TR, Lander ES, Getz G, Sellers WR, Meyerson M: The landscape of somatic copy-number alteration across human cancers. Nature. 2010, 463 (7283): 899-905. 10.1038/nature08822.PubMed CentralView ArticlePubMedGoogle Scholar
- Gleason DF: The Veteran's Administration Cooperative Urologic Research Group: histologic grading and clinical staging of prostatic carcinoma. Urologic Pathology: The Prostate. Edited by: Tannenbaum M. 1977, Philadelphia: Lea and Febiger, 171-198.Google Scholar
- Louis DN, Ohgaki H, Wiestler OD, Cavenee WK, Burger PC, Jouvet A, Scheithauer BW, Kleihues P: The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol. 2007, 114 (2): 97-109. 10.1007/s00401-007-0243-4.PubMed CentralView ArticlePubMedGoogle Scholar
- Barretina J, Taylor BS, Banerji S, Ramos AH, Lagos-Quintana M, Decarolis PL, Shah K, Socci ND, Weir BA, Ho A, Chiang DY, Reva B, Mermel CH, Getz G, Antipin Y, Beroukhim R, Major JE, Hatton C, Nicoletti R, Hanna M, Sharpe T, Fennell TJ, Cibulskis K, Onofrio RC, Saito T, Shukla N, Lau C, Nelander S, Silver SJ, Sougnez C, et al: Subtype-specific genomic alterations define new targets for soft-tissue sarcoma therapy. Anal Nat Genet. 2010, 42 (8): 715-21. 10.1038/ng.619.View ArticleGoogle Scholar
- Verhaak RG, Hoadley KA, Purdom E, Wang V, Qi Y, Wilkerson MD, Miller CR, Ding L, Golub T, Mesirov JP, Alexe G, Lawrence M, O'Kelly M, Tamayo P, Weir BA, Gabriel S, Winckler W, Gupta S, Jakkula L, Feiler HS, Hodgson JG, James CD, Sarkaria JN, Brennan C, Kahn A, Spellman PT, Wilson RK, Speed TP, Gray JW, Meyerson M, Getz G, Perou CM, Hayes DN, Cancer Genome Atlas Research Network: Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell. 2010, 17 (1): 98-110. 10.1016/j.ccr.2009.12.020.PubMed CentralView ArticlePubMedGoogle Scholar
- Cancer Genome Atlas Research Network: Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature. 2008, 455 (7216): 1061-8. 10.1038/nature07385.View ArticleGoogle Scholar
- Cancer Genome Atlas Research Network: Integrated genomic analyses of ovarian carcinoma. Nature. 2011, 474 (7353): 609-15. 10.1038/nature10166.View ArticleGoogle Scholar
- Cancer Genome Atlas Network: Comprehensive molecular portraits of human breast tumours. Nature. 2012, 490 (7418): 61-70. 10.1038/nature11412.View ArticleGoogle Scholar
- Cancer Genome Atlas Network: Comprehensive molecular portraits of human colon and rectal cancer. Nature. 2012, 487 (7407): 330-7. 10.1038/nature11252.View ArticleGoogle Scholar
- Cancer Genome Atlas Research Network, Hammerman PS, Hayes DN, Wilkerson MD, Schultz N, Bose R, Chu A, Collisson EA, Cope L, Creighton CJ, Getz G, Herman JG, Johnson BE, Kucherlapati R, Ladanyi M, Maher CA, Robertson G, Sander C, Shen R, Sinha R, Sivachenko A, Thomas RK, Travis WD, Tsao MS, Weinstein JN, Wigle DA, Baylin SB, Govindan R, Meyerson M: Comprehensive genomic characterization of squamous cell lung cancers. Nature. 2012, 489 (7417): 519-25. 10.1038/nature11404.View ArticleGoogle Scholar
- Noushmehr H, Weisenberger DJ, Diefes K, Phillips HS, Pujara K, Berman BP, Pan F, Pelloski CE, Sulman EP, Bhat KP, Verhaak RG, Hoadley KA, Hayes DN, Perou CM, Schmidt HK, Ding L, Wilson RK, Van Den Berg D, Shen H, Bengtsson H, Neuvial P, Cope LM, Buckley J, Herman JG, Baylin SB, Laird PW, Aldape K, Cancer Genome Atlas Research Network: Identification of a CpG island methylator phenotype that defines a distinct subgroup of glioma. Cancer Cell. 2010, 17 (5): 510-22. 10.1016/j.ccr.2010.03.017.PubMed CentralView ArticlePubMedGoogle Scholar
- Sturm D, Witt H, Hovestadt V, Khuong-Quang DA, Jones DT, Konermann C, Pfaff E, Tönjes M, Sill M, Bender S, Kool M, Zapatka M, Becker N, Zucknick M, Hielscher T, Liu XY, Fontebasso AM, Ryzhova M, Albrecht S, Jacob K, Wolter M, Ebinger M, Schuhmann MU, van Meter T, Frühwald MC, Hauch H, Pekrun A, Radlwimmer B, Niehues T, von Komorowski G: Hotspot mutations in H3F3A and IDH1 define distinct epigenetic and biological subgroups of glioblastoma. Cancer Cell. 2012, 22 (4): 425-37. 10.1016/j.ccr.2012.08.024.View ArticlePubMedGoogle Scholar
- Shen R, Mo Q, Schultz N, Seshan VE, Olshen AB, Huse J, Ladanyi M, Sander C: Integrative subtype discovery in glioblastoma using iCluster. PLoS One. 2012, 7 (4): e35236-10.1371/journal.pone.0035236.PubMed CentralView ArticlePubMedGoogle Scholar
- Vaske CJ, Benz SC, Sanborn JZ, Earl D, Szeto C, Zhu J, Haussler D, Stuart JM: Inference of patient-specific pathway activities from multi-dimensional cancer genomics data using PARADIGM. Bioinformatics. 2010, 26 (12): i237-45. 10.1093/bioinformatics/btq182.PubMed CentralView ArticlePubMedGoogle Scholar
- Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, Mesirov JP: Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA. 2005, 102 (43): 15545-50. 10.1073/pnas.0506580102.PubMed CentralView ArticlePubMedGoogle Scholar
- Curigliano G: New drugs for breast cancer subtypes: targeting driver pathways to overcome resistance. Cancer Treat Rev. 2012, 38 (4): 303-10. 10.1016/j.ctrv.2011.06.006.View ArticlePubMedGoogle Scholar
- Jurasz P, Alonso-Escolano D, Radomski MW: Platelet-cancer interactions: mechanisms and pharmacology of tumour cell-induced platelet aggregation. Br J Pharmacol. 2004, 143 (7): 819-826. 10.1038/sj.bjp.0706013.PubMed CentralView ArticlePubMedGoogle Scholar
- Murray GI: The role of cytochrome P450 in tumour development and progression and its potential in therapy. J Pathol. 2000, 192 (4): 419-26. 10.1002/1096-9896(2000)9999:9999<::AID-PATH750>3.0.CO;2-0.View ArticlePubMedGoogle Scholar
- Mathijssen RH, Marsh S, Karlsson MO, Xie R, Baker SD, Verweij J, Sparreboom A, McLeod HL: Irinotecan pathway genotype analysis to predict pharmacokinetics. Clin Cancer Res. 2003, 9 (9): 3246-53.PubMedGoogle Scholar
- Dong J, Jimi E, Zeiss C, Hayden MS, Ghosh S: Constitutively active NF-kappaB triggers systemic TNFalpha-dependent inflammation and localized TNFalpha-independent inflammatory disease. Genes Dev. 2010, 24 (16): 1709-17. 10.1101/gad.1958410.PubMed CentralView ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.