Gene expression analysis of glioblastomas identifies the major molecular basis for the prognostic benefit of younger age
© Lee et al; licensee BioMed Central Ltd. 2008
Received: 07 March 2008
Accepted: 21 October 2008
Published: 21 October 2008
Glioblastomas are the most common primary brain tumour in adults. While the prognosis for patients is poor, gene expression profiling has detected signatures that can sub-classify GBMs relative to histopathology and clinical variables. One category of GBM defined by a gene expression signature is termed ProNeural (PN), and has substantially longer patient survival relative to other gene expression-based subtypes of GBMs. Age of onset is a major predictor of the length of patient survival where younger patients survive longer than older patients. The reason for this survival advantage has not been clear.
We collected 267 GBM CEL files and normalized them relative to other microarrays of the same Affymetrix platform. 377 probesets on U133A and U133 Plus 2.0 arrays were used in a gene voting strategy with 177 probesets of matching genes on older U95Av2 arrays. Kaplan-Meier curves and Cox proportional hazard analyses were applied in distinguishing survival differences between expression subtypes and age.
This meta-analysis of published data in addition to new data confirms the existence of four distinct GBM expression-signatures. Further, patients with PN subtype GBMs had longer survival, as expected. However, the age of the patient at diagnosis is not predictive of survival time when controlled for the PN subtype.
The survival benefit of younger age is nullified when patients are stratified by gene expression group. Thus, the main cause of the age effect in GBMs is the more frequent occurrence of PN GBMs in younger patients relative to older patients.
Glioblastoma Multiforme (GBM) persists as one of the most lethal forms of human cancer with a median survival of 12 to 14.6 months for patients who receive the latest surgical, radiation, and chemotherapy treatment . While increased efficacy of some chemotherapy drugs is evident, the prognosis remains dismal . Despite the number of patients diagnosed with primary malignant brain and central nervous system tumours is small (relative to other cancer types i.e. 1.35% of all primary malignant cancers, U.S.), the morbidity and mortality of these patients remains severe .
Gliomas are primarily identified by histological features established by the grading scale of the World Health Organization (WHO). Prominent features include necrosis, nuclear mitotic activity, vascular proliferation, and cellular atypia . In clinical oncology, the primary method for identifying survival factors in high-grade gliomas has been through Cox proportional hazards models. Covariates such as histological classification, performance status, and patient age at the time of surgery are particularly useful to predict patient outcomes [5, 6]. Numerous studies have indicated that lower tumour WHO grade and younger patient age are the most prominent indicators of longer survival. Of patients diagnosed with a malignant glioma, younger patients have a disproportionately higher likelihood of being diagnosed with a lower grade glioma relative to older patients. However, a frequently observed phenomenon is that even among patients with the same high grade of glioma, younger patients tend to have a longer survival time. The reason for this survival advantage has not been clear .
Our group in 2004 used a heterogeneous group of high-grade gliomas that were categorized as WHO grade III and grade IV to identify molecularly classifiable groups irrespective of histologic typing and correlated with survival . Despite this histological heterogeneity, these gliomas could be successfully stratified into at least three molecularly-defined glioma subgroups whose classification predicted patient survival. In particular, the HC1A tumours were identified as a good survival group, which were categorized by the high expression of genes typically expressed during neuronal development. The remaining groups HC2A and HC2B were characterized by mitotic and extra-cellular matrix related genes, respectively, and both demonstrated poor survival prognoses. Recently, the increased generation of whole genome microarray expression data has supported the gene expression signatures of the three previously identified GBM subgroups and the correlation with patient survival [9, 10]. Phillips and colleagues suggested the following terms to be more descriptive of the cellular characteristics: 'ProNeural' (PN) for HC1A, 'Proliferative' (Pro) for HC2A, and 'Mesenchymal' (Mes) for HC2B, which we adopt here .
In order to investigate the robustness and generality of the gene expression-based groupings of GBMs, we have aggregated the publicly available genome scale expression data of histologically defined GBMs. Only samples performed on the Affymetrix platforms were included for joint analysis and analytical simplicity. In total, 181 GBMs were identified from the published literature for which Affymetrix microarrays were performed and CEL files were available. For many of these samples, survival and patient age were available [8–13]. In order to have sufficient numbers of samples to explore the correlation of expression signatures with age in GBMs, we have also performed gene expression analysis of an additional 86 GBMs in addition to this publicly available dataset. From this combined analysis, we demonstrate that about 86% of all tumour biopsies classify strongly into one of the three molecular subgroups as defined by the available gene list from Freije et al., and for about 11%, there is strong evidence of a novel group of GBMs that share expression features of Pro and Mes subtypes from the bulk biopsy analysis. Further, we investigated if there was a correlation between the molecular subtypes in relation to age of the patient at time of diagnosis and survival. We find that PN tumours are substantially more commonly diagnosed in younger patients, and that there is no survival advantage of age independent of gene expression classification. In other words, the beneficial effect of younger age in patients diagnosed with GBM is entirely due to the observation that for yet undefined reasons, younger adult patients develop PN type GBMs more commonly than older patients.
Agglomeration of publicly available high-grade glioma microarray data confirms the presence of four molecularly distinct prognostic groups
Combined Data Sources
No. GBM IV
HC Class Distribution
Freije et al. 
Phillips et al. 
Rich et al. 
Mischel et al. 
Shai et al. 
Nutt et al. 
All of the microarray platforms can be used to identify the PN, Pro, Mes and ProMes subgroups of GBMs. Thus, we were able to merge data from the three most recent Affymetrix genome-scale expression arrays for humans. In terms of the gene voting compatibility between different microarray platforms, we calculate that the percentages of probesets which vote samples into their HC categories are quite similar. For the PN samples an average of 52%, 54%, and 47% of the PN probesets vote these tumours as PN across the U95Av2, U133A, and U133 Plus.2.0 arrays, respectively. The Pro samples also vote with an average of 72%, 69%, and 92% of the Pro probesets contributing to the positive votes among the three arrays. The ProMes category was identifiable from data from each of the three platforms with an average of 58% and 43% (Pro and Mes probesets, respectively), 64% and 45%, and 85% (only n = 3 U133 Plus 2.0 "Pro" samples) and 55% of their probesets. Finally, Mes samples vote with an average of 57%, 62%, and 51% of the Mes probesets on all three array types. Each sample voted over 47% of its category's probesets when voting into its respective HC class. Thus, even though differing sets of probes were used to identify the subgroups from the array platforms, the groups themselves were reliably identified.
The patient survival durations of each gene expression group confirms the prognostic utility of the gene signature-based predictor (Fig. 1B). PN GBM patients (n = 77) have a mean survival of 2.16 years (median = 1.4 years) while Pro (n = 61), Mes (n = 100), and ProMes (n = 29) GBM patients have a mean survival of 1.05 years (median = 283 days), 1.20 years (median = 322 days), and 1.44 years (median = 385 days), respectively. Moreover, only 35% of PN patients survive less than 1 year from the date of surgical resection, while about 60% of Pro, 49% of ProMes, and 58% of Mes patients succumb to their malignancies within the first year after resection. Part of the reason for the low survival is likely due to the lack of homogenous treatment across institutions over a period of several years. However, we note that among patients who received current standard of care (e.g. combined high dose radiotherapy, aggressive surgery, and Temozolomide (N = 36)), only 44% of these patients were alive at 12 months post diagnosis and only 39% (n = 14/36) survive up to or beyond 14.6 months. Thus, differences in therapeutic efficacy from the various studies included here are unlikely to alter the fundamental conclusions and that therapy is a minor contributor to the effects measured in terms of patient survival. The findings from this expanded dataset firmly corroborate the importance of gene signature based classification in categorizing glioblastomas into prognostically meaningful molecular groups. While there is little difference in survival between Pro, Mes, and ProMes tumour carriers, all have dramatically shorter patient survival times than the PN GBM patients. For the subset of the patients where treatment data were available, there were no differences in treatments attempted in the classification groups to account for the difference in survival length of the patients.
Molecular Classification & Age Analysis
For adult patients who are diagnosed with GBM, younger age at diagnosis is a strong predictor of longer patient survival. The mechanism for this observation has not been clear, and in the study by Freije et al., age and PN tumour type were clearly correlated . With the larger group of patients and GBM tumour biopsies available from the published work of several groups and new data reported here, we explore the relationship between the age of the patient, survival, and gene expression signature-based subtype of GBM. We sought to determine the relative importance of the PN subtype as compared to age of the patient. For these analyses, 28 GBM samples were removed from the analysis as no available age data were provided from the published data.
Multivariate and Univariate Cox proportional hazards analyses support that the molecular signature status of GBMs demonstrates greater hazard prediction than age
To measure the robustness of the PN and non-PN molecular classifier amongst several additional covariates, multivariate and univariate Cox proportional hazards analyses were performed across the GBM patients where age data were available (N = 239: PN (n = 67), non-PN (n = 172) Tables 2 and 3). The model tested the following covariates: age, glioma gene signature, and expression status of MGMT, VEGF, and EGFR. MGMT, VEGF, and EGFR were added in as individual expression covariates due to prior reports of their possible individual contributions in relation to survival prediction [14, 15]. Our analysis indicated that only increased VEGF expression significantly correlated with increasing age (data not shown), while MGMT and EGFR expression levels did not show significant correlations with age. While univariate analysis demonstrated as expected that the 'Age 40 and above' covariate represents a near significant marker for hazard (HR = 1.44, P-value = 0.067), age was not a significant survival predictor in the context of multivariate analysis (HR = 1.20, P-value = 0.39). Within the univariate model, the non-PN classification demonstrates a high hazard ratio (Table 3: HR = 1.94, P-value = 3.15e-5). Within the multivariate model, the most significant hazard covariate for patient outcome was in fact the non-PN molecular subtype (Table 2: HR = 1.93, P-value = 1.4e-4).
Multivariate Cox Proportional Hazards Ratio: Five Covariates
0.80 – 1.80
1.38 – 2.70
1.11 – 1.92
0.87 – 1.53
0.64 – 1.15
Univariate Cox Proportional Hazards Ratio: Five Covariates
0.97 – 2.14
1.42 – 2.66
0.97 – 1.66
1.03 – 1.76
0.68 – 1.21
This study uses available genome-scale gene expression based analyses across glioblastomas to further investigate observed age effects in patient survival and creates a unified dataset for further exploration of gene expression correlates in glioblastomas. We add to the literature 86 additional GBM biopsy gene expression profiles performed on the U133A and U133 Plus 2.0 platforms. We confirm that within histologically defined GBMs, there are robust and repeatedly observed gene expression signature based groups of GBMs, which produces a classification scheme within GBMs. Within this aggregate dataset of 267 GBMs, the three previously defined molecular glioma subgroups were robustly detected as defined by over-expression of a series of related genes: Neurogenesis/HC1A (a.k.a. ProNeural), Mitotic/HC2A (a.k.a. Proliferative), and Extra-Cellular Matrix related/HC2B (a.k.a. Mesenchymal). From the larger dataset, we observe evidence of a new subtype that highly co-expresses genes in both the Pro and Mes groups, which we term "ProMes". Of these four molecularly distinct groups identified, only PN portends a favourable prognosis relative to the other expression based groups.
In this study, we determine that the reason that younger patients diagnosed with GBMs have longer survival durations than older GBM patients is due to the observation that younger patients tend to develop the favourable PN GBM type more commonly relative to older patients. The reason for this observed age effect is not clear at this time, but may have to do with the precursor cell that develops into GBMs changing over time. One could hypothesize that the precursor cell that gives rise to the PN type diminishes in abundance in the CNS with advancing age. However, given that GBM incidence increases greatly with age, the absolute numbers of PN GBMs is actually numerically higher in older patient groups. Thus, we favour a model that the precursor cell type that gives rise to the Pro and Mes types of GBMs are increasingly likely to become neoplastic over time while this effect is not as pronounced within the PN type precursors.
PN type GBMs represent a unique tumour aetiology whose idiopathic molecular mechanisms manifest in survival periods spanning from two to ten years in contrast to ten to fifteen months for the Pro, Mes and ProMes types of GBMs. Historically, the percentage of GBM patients who demonstrate long term survival of 3 years or more has been reported to be approximately 5% . This 5% incidence rate matches well with the observations reported here in which 8% (18/267) of our PN GBM patients are observed to survive 3 years or longer. The identification of the PN subgroup is important for patient management and stratification into small phase II clinical trials for experimental therapeutics as uneven representation of the PN GBM diagnoses would greatly alter observed survival times irrespective of potentially active agents . The identification of the gene expression subtype is clearly more important for patient stratification within clinical trials than age.
It is likely that the use of genome-wide expression based molecular classification will result in less variation in tumour diagnoses and provide more specific guidance to clinicians. The agglomeration of gene expression datasets permits meta-analyses that were insufficiently powered in multiple individual publications. Resources for the sharing of genome-scale expression datasets have been set up at Array Express, Gene Expression Omnibus, and Celsius [20, 21]. Critical to the sharing of microarray data is providing raw microarray data as opposed to processed data. In order to facilitate this sharing, the NIH Neuroscience Microarray Consortium has established Celsius, which is a community resource of CEL (image) files performed on the Affymetrix platform for public distribution using programmatic tools. At the writing of this manuscript, the Celsius database contains CEL files from human experiments performed on U95Av2 arrays (n = 5 006), U133A arrays (n = 13 818), and U133 Plus 2.0 arrays (n = 10 376). To fully capture and leverage the value of microarray expression data, a greater commitment must be made to capture and share clinical covariates and raw expression data. For instance, in this study of 267 glioblastomas, a total of 433 GBM CEL files were initially identified across the U95Av2, U133A, and U133 Plus 2.0 platforms. Of these, thirty-six percent (160/433) were immediately removed from this study due to a lack of any clinical data. Additionally, not all microarray CEL files or their matched clinical data points are systematically retrievable.
Amongst the samples gathered at UCLA, we simultaneously gathered additional covariates such as extent of surgical resection, Karnofsky Performance Scores (KPS), lesion locations, and MRI scans. We have begun to make all of these data available through a web interface in order to promote data sharing and exploration of gene expression differences in gliomas. All of the data added here are deposited in Gene Expression Omnibus (GEO), and can be explored at our real-time survival-synchronized search engine "Probeset Analyzer" .
Clinical Decision Impact and Improved Public Disclosure
In large academic hospitals, tumours come from a wide variety of patients from across different cities, states, or countries. In contrast, local hospitals treat their regional constituencies. The potential for demographically-biased patient populations and biased tumour subsets is a possibility. These trends can reinforce particular treatment strategies at local institutions over time. For example, if patients from a community highly populated by retirees (e.g. southern Florida) presented with a GBM, clinicians would be apt to predict that these older patients would likely succumb to their malignancies within one year. Current treatment for patients diagnosed with high grade gliomas consists of surgical resection followed by toxic and expensive therapy schedules that are minimally effective. But if these elderly patients were suffering from a PN tumour, they would have a high likelihood for surviving at least two to three years or longer. These patients could then be distinguished from patients who otherwise present identically under the microscope or according to their patient biographical sketch. This would permit time to enrol in potentially beneficial clinical trials. Thus, if grade and age alone were considered for prognosis, these factors would lead clinicians to prescribe unnecessary treatments due to trends reinforced by regional sampling biases.
We provide an explanation for why younger patients diagnosed with GBM have longer life expectancies than older patients using accumulated whole-genome microarray expression data and clinical variables. We have discovered the reason that younger patients tend to survive longer is because they are more likely to present with PN gene signature tumours relative to the more common and aggressive Pro, Mes, and ProMes GBM types. The PN molecular classification predicts an enhanced survival performance by at least two to ten years irrespective of age when tested against age-matched, Pro, Mes, or ProMes molecularly-classified tumours. The application of Cox proportional hazards studies have also confirmed that having a non-PN tumour was the most statistically significant factor in predicting precipitous short survival over age by two orders of magnitude. These data lend more evidence to the clinical reality that high-grade gliomas can be molecularly distinct tumours. Therefore, different tumours with distinct aetiologies should be differentially segregated in terms of their treatment regimens and especially their clinical trial assignments. The benefits for clinicians would be a reduction in the heterogeneous admixture of genetic background for treatment cohorts and a potential reduction in the percentage of non-responders for molecularly-targeted investigational new drugs. Patients accurately classified and characterized for the biology of their tumour may potentially benefit from a variety of molecularly-targeted treatments as the inclusion criteria for clinical trials become simultaneously based on molecular signatures.
Microarray and Clinical Data Collection
Clinical data including histopathology, age, sex, and survival time from diagnosis were retrieved from 181 glioblastomas which have been reported within previous studies between 2003 and 2006 (Table 1) and for which CEL files (Affymetrix, Santa Clara, CA) were available from the authors. In addition, we collected 86 new patient-unique tumour biopsies from the UCLA Neuro-oncology Program (n = 55) and the Barrow Neurological Institute (n = 31) for a grand total of 267 glioblastomas. Newly acquired tumours were collected through institutional review board approved protocols and assigned WHO grades at UCLA Neuropathology or Barrow Neuropathology by PSM. Time of survival (days), sex, and, age were collected where available [see Additional file 1]. Patient age at the time of diagnosis was available for 239 patients and ranged from 18 to 86 years. Sex of the individual was available for those 239 patients (151 males and 88 females).
Total RNA was purified from fresh frozen tumour biopsies and visually inspected for tumour content using Qiagen RNAeasy columns and standard manufacturer's protocols. Labelled one round cRNA was generated using kits (GeneChip One-Cycle Target Labelling and Control Reagent) from Affymetrix. cRNA was quantified and 15 micrograms were hybridized to U133A and U133 Plus 2.0 arrays at the UCLA DNA Microarray Facility using standard protocols recommended by the manufacturer [see Additional file 2]. All newly generated CEL files were deposited into the Celsius microarray database and this system was used to normalize relative to other microarrays of the same Affymetrix platform using RMA with default settings from the Bioconductor R library [20, 21, 23, 24].
Combination of microarray data
The Celsius microarray database, which houses over 20 000 human CEL files on various array iterations, was used to quantify and normalize the three microarray platform datasets with comparison groups of 50 random samples selected from each respective platform for RMA normalization and quantification using default parameters. Only probesets from the U133A portion of the Freije et al. paper that are retained or map to the same gene were analyzed from each tumour type. All microarray CEL files analyzed in this study in conjunction with their clinical covariates are accessible from the Gene Expression Omnibus (GEO) (Series Accession number: GSE13041).
Sample membership by HC Classification Gene Voting
The Hierarchical Clustering (HC) classification for each glioma was determined by the gene voting strategy as described previously . Briefly, the mean value of each probeset was evaluated from all samples within each of the three microarray platforms U95Av2, U133A, and U133 Plus 2.0 separately. Second, the probesets from each sample were assigned "yes" or "no" votes if that probeset's value was above or below the aforementioned probeset mean of its platform. Third, the "yes" or "no" votes of each probeset from the 377 probesets contained within the U133A portion of the 595 HC probeset classifier (which was based on U133A and U133B data) were tallied and used to categorize every glioma into one of three HC molecular groups. All of the 377 probesets from the 377 used for classification were on the U133A and the U133 Plus 2.0 array types, and 177 of the probesets were able to be matched to the same genes from the older U95Av2 arrays. For probesets that were unavailable on the U95Av2 arrays, the mean value of the probesets on the U133A and U133 Plus 2.0 arrays were calculated and intercalated for the missing row values of the U95Av2 samples strictly for visualization purposes and not gene voting. Lastly, each tumour was voted into one of three HC groups based on the highest vote tally: 1A vote = (tally of 1A probesets above the mean)/(total 1A probesets); 2A vote = (tally of 2A probesets above the mean)/(total 2A probesets); 2B vote = (tally of 2B probesets above the mean)/(total 2B probesets). A few gliomas appeared to vote almost equally well into both the 2A and 2B categories and are defined as "2A2B", which is defined where the highest vote category must be a 2A or 2B and the second highest vote must be 2B or 2A, respectively with at least a 33% vote. The lists of probesets used for voting each tumour across each of the Affymetrix platforms is available [see Additional file 3].
Kaplan-Meier and Cox Regression Analysis
Kaplan-Meier survival plots and Cox proportional hazard regression analyses were implemented in R version 2.5.0 using the "survival" library. Survival data and vital status for UCLA samples were based on time in days elapsed from surgical resection to the date of death up to May 1, 2006. The full R code is available and documented [see Additional file 4].
MGMT, VEGFA, EGFR Expression State Determination
Expression states for the MGMT, VEGFA and EGFR probesets were called "ON" or "OFF" based on whether their probesets' expression levels were above or below their respective mean within each platform. The probesets employed for the genes from each platform are listed [see Additional file 3].
World Health Organization
Hierarchical Cluster group 1A (Neurogenesis)/ProNeural
Hierarchical Cluster group 2A (Mitotic)/Proliferative
Hierarchical Cluster group 2A2B/Proliferative Mesenchymal
Hierarchical Cluster group 2B (Extra-Cellular Matrix)/Mesenchymal
vascular endothelial growth factor A (GeneID: 7422)
epidermal growth factor receptor (GeneID: 1956), Temozolomide (Brand Name: Temodar®).
The authors wish to thank the UCLA and Barrow Institute subjects for their participation in this study. The work was supported by the NIH Neuroscience Microarray Consortium UCLA Site (U24NS052108), The UCLA Gene Expression Shared Resource and the Singleton Brain Tumour Program.
- Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ, Belanger K, Brandes AA, Marosi C, Bogdahn U, Curschmann J, Janzer RC, Ludwin SK, Gorlia T, Allgeier A, Lacombe D, Cairncross JG, Eisenhauer E, Mirimanoff RO, European Organisation for Research and Treatment of Cancer Brain Tumor and Radiotherapy Groups; National Cancer Institute of Canada Clinical Trials Group: Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med. 2005, 352 (10): 987-996. 10.1056/NEJMoa043330.View ArticlePubMedGoogle Scholar
- Omuro AM, Faivre S, Raymond E: Lessons learned in the development of targeted therapy for malignant gliomas. Mol Cancer Ther. 2007, 6 (7): 1909-1919. 10.1158/1535-7163.MCT-07-0047.View ArticlePubMedGoogle Scholar
- CBTRUS. 2005–2006 Statistical Report: Primary Brain Tumors in the United States Statistical Report, 1998–2002 (Years Data Collected). 2006
- Louis DN, Posner J, Jacobs T, Kaplan R: Report of the Brain Tumor Progress Review Group. 2000, National Institutes of Health: Bethesda, MDGoogle Scholar
- Dong S, Nutt CL, Betensky RA, Stemmer-Rachamimov AO, Denko NC, Ligon KL, Rowitch DH, Louis DN: Histology-based expression profiling yields novel prognostic markers in human glioblastoma. J Neuropathol Exp Neurol. 2005, 64 (11): 948-955. 10.1097/01.jnen.0000186940.14779.90.View ArticlePubMedPubMed CentralGoogle Scholar
- Krex D, Klink B, Hartmann C, von Deimling A, Pietsch T, Simon M, Sabel M, Steinbach JP, Heese O, Reifenberger G, Weller M, Schackert G: Long-term survival with glioblastoma multiforme. Brain. 2007, 2596-2606. 10.1093/brain/awm204. 2007 Sep 4Google Scholar
- Wrensch M, Fisher JL, Schwartzbaum JA, Bondy M, Berger M, Aldape KD: The molecular epidemiology of gliomas in adults. Neurosurg Focus. 2005, 19 (5): E5-10.3171/foc.2005.19.5.6. 1–11View ArticlePubMedGoogle Scholar
- Freije WA, Castro-Vargas FE, Fang Z, Horvath S, Cloughesy T, Liau LM, Mischel PS, Nelson SF: Gene expression profiling of gliomas strongly predicts survival. Cancer Res. 2004, 64 (18): 6503-6510. 10.1158/0008-5472.CAN-04-0452.View ArticlePubMedGoogle Scholar
- Phillips HS, Kharbanda S, Chen R, Forrest WF, Soriano RH, Wu TD, Misra A, Nigro JM, Colman H, Soroceanu L, Williams PM, Modrusan Z, Feuerstein BG, Aldape K: Molecular subclasses of high-grade glioma predict prognosis, delineate a pattern of disease progression, and resemble stages in neurogenesis. Cancer Cell. 2006, 9 (3): 157-173. 10.1016/j.ccr.2006.02.019.View ArticlePubMedGoogle Scholar
- Rich JN, Hans C, Jones B, Iversen ES, McLendon RE, Rasheed BK, Dobra A, Dressman HK, Bigner DD, Nevins JR, West M: Gene expression profiling and genetic markers in glioblastoma survival. Cancer Res. 2005, 65 (10): 4051-4058. 10.1158/0008-5472.CAN-04-3936.View ArticlePubMedGoogle Scholar
- Mischel PS, Shai R, Shi T, Horvath S, Lu KV, Choe G, Seligson D, Kremen TJ, Palotie A, Liau LM, Cloughesy TF, Nelson SF: Identification of molecular subtypes of glioblastoma by gene expression profiling. Oncogene. 2003, 22 (15): 2361-2373. 10.1038/sj.onc.1206344.View ArticlePubMedGoogle Scholar
- Shai R, Shi T, Kremen TJ, Horvath S, Liau LM, Cloughesy TF, Mischel PS, Nelson SF: Gene expression profiling identifies molecular subtypes of gliomas. Oncogene. 2003, 22 (31): 4918-4923. 10.1038/sj.onc.1206753.View ArticlePubMedGoogle Scholar
- Nutt CL, Mani DR, Betensky RA, Tamayo P, Cairncross JG, Ladd C, Pohl U, Hartmann C, McLaughlin ME, Batchelor TT, Black PM, von Deimling A, Pomeroy SL, Golub TR, Louis DN: Gene expression-based classification of malignant gliomas correlates better with survival than histological classification. Cancer Res. 2003, 63 (7): 1602-1607.PubMedGoogle Scholar
- Herrlinger U, Rieger J, Koch D, Loeser S, Blaschke B, Kortmann RD, Steinbach JP, Hundsberger T, Wick W, Meyermann R, Tan TC, Sommer C, Bamberg M, Reifenberger G, Weller M: Phase II trial of lomustine plus temozolomide chemotherapy in addition to radiotherapy in newly diagnosed glioblastoma: UKT-03. J Clin Oncol. 2007, 24 (27): 4412-4417. 10.1200/JCO.2006.06.9104.View ArticleGoogle Scholar
- Zhou YH, Tan F, Hess KR, Yung WK: The expression of PAX6, PTEN, vascular endothelial growth factor, and epidermal growth factor receptor in gliomas: relationship to tumor grade and survival. Clin Cancer Res. 2003, 9 (9): 3369-3375.PubMedGoogle Scholar
- Martinez R, Schackert G, Yaya-Tur R, Rojas-Marcos I, Herman JG, Esteller M: Frequent hypermethylation of the DNA repair gene MGMT in long-term survivors of glioblastoma multiforme. J Neurooncol. 2007, 83 (1): 91-93. 10.1007/s11060-006-9292-0.View ArticlePubMedGoogle Scholar
- Ishii D, Natsume A, Wakabayashi T, Hatano H, Asano Y, Takeuchi H, Shimato S, Ito M, Fujii M, Yoshida J: Efficacy of temozolomide is correlated with 1p loss and methylation of the deoxyribonucleic acid repair gene MGMT in malignant gliomas. Neurol Med Chir (Tokyo). 2007, 47 (8): 341-349. 10.2176/nmc.47.341. discussion 350View ArticleGoogle Scholar
- Hegi ME, Diserens AC, Gorlia T, Hamou MF, de Tribolet N, Weller M, Kros JM, Hainfellner JA, Mason W, Mariani L, Bromberg JE, Hau P, Mirimanoff RO, Cairncross JG, Janzer RC, Stupp R: MGMT gene silencing and benefit from temozolomide in glioblastoma. N Engl J Med. 2005, 352 (10): 997-1003. 10.1056/NEJMoa043331.View ArticlePubMedGoogle Scholar
- Mischel PS, Cloughesy TF, Nelson SF: DNA-microarray analysis of brain cancer: molecular classification for therapy. Nat Rev Neurosci. 2004, 5 (10): 782-792. 10.1038/nrn1518.View ArticlePubMedGoogle Scholar
- Day A, Carlson MR, Dong J, O'Connor BD, Nelson SF: Celsius: a community resource for Affymetrix microarray data. Genome Biol. 2007, 8 (6): R112-10.1186/gb-2007-8-6-r112.View ArticlePubMedPubMed CentralGoogle Scholar
- The Celsius Project. [http://genomics.ctrl.ucla.edu/wiki/index.php/Celsius]
- Probeset Analyzer. [http://www.probesetanalyzer.com]
- Bolstad BM, Irizarry RA, Astrand M, Speed TP: A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics. 2003, 19 (2): 185-193. 10.1093/bioinformatics/19.2.185.View ArticlePubMedGoogle Scholar
- Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, Ellis B, Gautier L, Ge Y, Gentry J, Hornik K, Hothorn T, Huber W, Iacus S, Irizarry R, Leisch F, Li C, Maechler M, Rossini AJ, Sawitzki G, Smith C, Smyth G, Tierney L, Yang JY, Zhang J: Bioconductor: open software development for computational biology and bioinformatics. Genome Biol. 2004, 5 (10): R80-10.1186/gb-2004-5-10-r80.View ArticlePubMedPubMed CentralGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1755-8794/1/52/prepub
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.