Combining clinical and molecular data in regression prediction models: insights from a simulation study.

Data integration, i.e. the use of different sources of information for data analysis, is becoming one of the most important topics in modern statistics. Especially in, but not limited to, biomedical applications, a relevant issue is the combination of low-dimensional (e.g. clinical data) and high-dimensional (e.g. molecular data such as gene expressions) data sources in a prediction model. Not only the different characteristics of the data, but also the complex correlation structure within and between the two data sources, pose challenging issues. In this paper, we investigate these issues via simulations, providing some useful insight into strategies to combine low- and high-dimensional data in a regression prediction model. In particular, we focus on the effect of the correlation structure on the results, while accounting for the influence of our specific choices in the design of the simulation study.

Latent structure and factor reliability of the National Health Service Community Mental Health Service User Questionnaire.

BACKGROUND: National Health Service use the Community Mental Health Service User Questionnaire (NHS-CMH) to assess care quality. However, its reliability and internal validity is uncertain. AIMS: To test the NHS-CMH structure, reliability and item-level characteristics. METHODS: We used data from 11,373 participants who answered the 2017 NHS-CMH survey. First, we estimated the NHS-CMH structure using Exploratory Factor Analysis (EFA) in half of the dataset. Second, we tested the best EFA-derived model with Confirmatory Factor Analysis (CFA). We tested the internal validity, construct reliability (omega - ω), explained common variance of each factor (ECV), and item thresholds. RESULTS: EFA suggested a 4-factor solution. The structure derived from the EFA was confirmed, demonstrating good reliability for the four correlated dimensions: "Relationship with Staff" (ω = 0.952, ECV = 40.1%), "Organizing Care" (ω = 0.855, ECV = 21.4%), "Medication and Treatments" (ω = 0.837, ECV = 13.3%), and "Support and Well-being" (ω = 0.928, ECV = 25.3%). A second-order model with a high-order domain of "Quality of Care" is also supported. CONCLUSIONS: The NHS-CMH can be used to reliably assess four user-informed dimensions of mental health care quality. This model offers an alternative for its current use (item-level and untested sum scores analysis).

AL amyloidosis patients with low amyloidogenic free light chain levels at first diagnosis have an excellent prognosis.

The difference between involved minus uninvolved serum free light chains (dFLC) has been established as an invaluable hematologic parameter in systemic amyloid light chain (AL) amyloidosis. However, patients with an initial dFLC level <50 mg/L are currently deemed not evaluable for response to therapy. Therefore, we aimed to characterize this subgroup of patients and to define novel hematologic response parameters. We retrospectively analyzed 783 AL patients newly diagnosed at our center between 2002 and 2016. Patients with a dFLC level <50 mg/L showed smaller bone marrow plasmacytosis compared to patients with a dFLC level ≥50 mg/L (7% vs 10%, P < .001), but no significant differences in all analyzed chromosomal aberrations. Cardiac involvement was less frequent (45% vs 80%, P < .001) and less severe (Mayo 2004 stage III: 18% vs 51%, P < .001), whereas kidney involvement was more prevalent (83% vs 53%, P < .001) and proteinuria was higher (7.3 g/L vs 5.0 g/L, P < .001). In multivariate analyses, a dFLC level <50 mg/L appeared to be an independent prognostic factor with respect to overall survival (hazard ratio [HR] = 0.50, P = .003) and renal survival (HR = 0.56, P = .020). Patients with a dFLC level <50 mg/L showed a higher proportion of complete hematologic response after first-line therapy compared to patients with a dFLC level ≥50 mg/L (39% vs 9%, P < .001). Receiver-operating characteristics analysis identified a low-dFLC partial response (dFLC <10 mg/L for patients with a dFLC between 20 and 50 mg/L), which predicted overall and renal survival already at 3 months after the start of therapy. Importantly, a parallel Italian study validated this new hematologic remission parameter. The outcome of prospective clinical trials might be adversely influenced by exclusion of the favorable clinical subgroup with an initial dFLC <50 mg/L. We propose the appreciation of dFLC in hematologic response assessment for all patients with a baseline dFLC >20 mg/L.

NFATC1 activation by DNA hypomethylation in chronic lymphocytic leukemia correlates with clinical staging and can be inhibited by ibrutinib.

B cell receptor (BCR) signaling is a key for survival of chronic lymphocytic leukemia (CLL) cells, and BCR signaling inhibitors are clinically active. However, relapse and resistance to treatment require novel treatment options. To detect novel candidate therapeutic targets, we performed a genome-wide DNA methylation screen with custom arrays and identified aberrant promoter DNA methylation in 2,192 genes. The transcription factor NFATC1 that is a downstream effector of BCR signaling was among the top hypomethylated genes and was concomitantly transcriptionally upregulated in CLL. Intriguingly, NFATC1 promoter DNA hypomethylation levels were significantly variant in clinical trial cohorts from different disease progression stages and furthermore correlated with Binet disease staging and thymidine kinase levels, strongly suggesting a central role of NFATC1 in CLL development. Functionally, DNA hypomethylation at NFATC1 promoter inversely correlated with RNA levels of NFATC1 and dysregulation correlated with expression of target genes BCL-2, CCND1 and CCR7. The inhibition of the NFAT regulator calcineurin with tacrolimus and cyclosporin A and the BCR signaling inhibitor ibrutinib significantly reduced NFAT activity in leukemic cell lines, and NFAT inhibition resulted in increased apoptosis of primary CLL cells. In summary, our results indicate that the aberrant activation of NFATC1 by DNA hypomethylation and BCR signaling plays a major role in the pathomechanism of CLL.

DMBT1 expression in biliary carcinogenesis with correlation of clinicopathological data.

AIMS: Deleted in malignant brain tumours 1 (DMBT1) exerts functions in the regulation of epithelial differentiation and inflammation and has been proposed as a tumour suppressor. Because chronic inflammation is a hallmark of cholangiocarcinogenesis, the aim of this study was to investigate the expression of DMBT1 in biliary tract cancer (BTC) and to correlate this expression with clinicopathological data. METHODS AND RESULTS: The expression of DMBT1 protein was examined immunohistochemically in 157 BTC patients [41 intrahepatic (ICC), 60 extrahepatic cholangiocarcinomas (ECC) and 56 adenocarcinomas of the gallbladder (GBAC)]. Additionally, 56 samples of high-grade biliary intraepithelial neoplasia (BilIN 3) and 92 corresponding samples of histological non-neoplastic biliary tract tissues were included. DMBT1 expression was increased significantly in BilIN 3 compared to normal tissue (P < 0.0001) and BTC (P < 0.0001). BTC showed no significant difference in DMBT1 expression compared to non-neoplastic biliary tissue (P = 0.315). Absent DMBT1 expression in non-neoplastic biliary tissue of BTC patients was associated with poorer survival (P = 0.027). DMBT1 expression was correlated significantly with patients' age (P < 0.001). CONCLUSION: DMBT1 is expressed differently in cholangiocarcinogenesis and poorer patients' survival rates are associated with absent DMBT1 expression in non-neoplastic biliary tissue, suggesting a tumour-suppressive role of DMBT1 in early cholangiocarcinogenesis.

Lenalidomide/melphalan/dexamethasone in newly diagnosed patients with immunoglobulin light chain amyloidosis: results of a prospective phase 2 study with long-term follow up.

Chemotherapy in light chain amyloidosis aims to normalize the involved free light chain in serum, which leads to an improvement, or at least stabilization of organ function in most responding patients. We performed a prospective single center phase 2 trial with 50 untreated patients not eligible for high-dose treatment. The treatment schedule comprised 6 cycles of oral lenalidomide, melphalan and dexamethasone every 4 weeks. After 6 months, complete remission was achieved in 9 patients (18%), very good partial remission in 16 (32%) and partial response in 9 (18%). Overall, organ response was observed in 24 patients (48%). Hematologic and cardiac toxicities were predominant adverse events. Mortality at 3 months was low at 4% (n=2) despite the inclusion of 36% of patients (n=18) with cardiac stage Mayo 3. After a median follow-up of 50 months, median overall and event-free survival were 67.5 months and 25.1 months, respectively. We conclude that the treatment of lenalidomide, melphalan and dexamethasone is very effective in achieving a hematologic remission, organ response and, consecutively, a long survival in transplant ineligible patients with light chain amyloidosis. However, as toxicity and tolerability are the major problems of a 3-drug regimen, a strict surveillance program is necessary and sufficient to avoid severe toxicities. clinicaltrials.gov Identifier: 00883623 (Eudract2008-001405-41).

Cyclophosphamide-based stem cell mobilization in relapsed multiple myeloma patients: A subgroup analysis from the phase III trial ReLApsE.

OBJECTIVE: Analysis of the efficiency and toxicity of cyclophosphamide-based stem cell mobilization in patients with relapsed multiple myeloma (RMM). METHODS: Peripheral blood stem cells (PBSCs) were mobilized with high dose cyclophosphamide (2 g/m2 daily on days 1 and 2) and G-CSF plus pre-emptive/rescue plerixafor in RMM patients (first to third relapse) treated within the ReLApsE trial of the German-Speaking Myeloma Multicenter Group (GMMG). RESULTS: Mobilization was initiated with high-dose cyclophosphamide (HD-CY) and G-CSF in 30 patients. Fifteen patients received additional pre-emptive/rescue administration of plerixafor. Stem cell collection was successful (≥2×106 CD34+ cells per kg bw) in 77% (23/30 patients). Patients with prior high-dose melphalan collected a significantly lower median total number of PBSCs than patients without prior high-dose melphalan (3.3×106 vs 17×106 CD34+ cells/kg bw). Toxicity of HD-CY was frequent with 12 serious adverse events (SAE) in 37% of patients (11/30 patients). Infections accounted for the majority of SAE reports. In two patients, SAEs were lethal (septic shock). CONCLUSIONS: These data proof feasibility of PBSC collection at relapse but emphasize the importance of collection and storage of additional PBSC transplants during first-line treatment when mobilization is more efficient and less toxic.

SNPs in transporter and metabolizing genes as predictive markers for oxaliplatin treatment in colorectal cancer patients.

Oxaliplatin is frequently used as part of a chemotherapeutic regimen with 5-fluorouracil in the treatment of colorectal cancer (CRC). The cellular availability of oxaliplatin is dependent on metabolic and transporter enzymes. Variants in genes encoding these enzymes may cause variation in response to oxaliplatin and could be potential predictive markers. Therefore, we used a two-step procedure to comprehensively investigate 1,444 single nucleotide polymorphisms (SNPs) from these pathways for their potential as predictive markers for oxaliplatin treatment, using 623 stage II-IV CRC patients (of whom 201 patients received oxaliplatin) from a German prospective patient cohort treated with adjuvant or palliative chemotherapy. First, all genes were screened using the global test that evaluated SNP*oxaliplatin interaction terms per gene. Second, one model was created by backward elimination on all SNP*oxaliplatin interactions of the selected genes. The statistical procedure was evaluated using bootstrap analyses. Nine genes differentially associated with overall survival according to oxaliplatin treatment (unadjusted p values < 0.05) were selected. Model selection resulted in the inclusion of 14 SNPs from eight genes (six transporter genes, ABCA9, ABCB11, ABCC10, ATP1A1, ATP1B2, ATP8B3, and two metabolism genes GSTM5, GRHPR), which significantly improved model fit. Using bootstrap analysis we show an improvement of the prediction error of 3.7% in patients treated with oxaliplatin. Several variants in genes involved in metabolism and transport could thus be potential predictive markers for oxaliplatin treatment in CRC patients. If confirmed, inclusion of these variants in a predictive test could identify patients who are more likely to benefit from treatment with oxaliplatin.

Altered regulation of DNA ligase IV activity by aberrant promoter DNA methylation and gene amplification in colorectal cancer.

Colorectal cancer (CRC) presents as a very heterogeneous disease which cannot sufficiently be characterized with the currently known genetic and epigenetic markers. To identify new markers for CRC we scrutinized the methylation status of 231 DNA repair-related genes by methyl-CpG immunoprecipitation followed by global methylation profiling on a CpG island microarray, as altered expression of these genes could drive genomic and chromosomal instability observed in these tumors. We show for the first time hypermethylation of MMP9, DNMT3A and LIG4 in CRC which was confirmed in two CRC patient groups with different ethnicity. DNA ligase IV (LIG4) showed strong differential promoter methylation (up to 60%) which coincided with downregulation of mRNA in 51% of cases. This functional association of LIG4 methylation and gene expression was supported by LIG4 re-expression in 5-aza-2'-deoxycytidine-treated colon cancer cell lines, and reduced ligase IV amounts and end-joining activity in extracts of tumors with hypermethylation. Methylation of LIG4 was not associated with other genetic and epigenetic markers of CRC in our study. As LIG4 is located on chromosome 13 which is frequently amplified in CRC, two loci were tested for gene amplification in a subset of 47 cases. Comparison of amplification, methylation and expression data revealed that, in 30% of samples, the LIG4 gene was amplified and methylated, but expression was not changed. In conclusion, hypermethylation of the LIG4 promoter is a new mechanism to control ligase IV expression. It may represent a new epigenetic marker for CRC independent of known markers.

Dose-intensified bendamustine followed by autologous peripheral blood stem cell support in relapsed and refractory multiple myeloma with impaired bone marrow function.

Therapeutic options in heavily pretreated relapsed/refractory multiple myeloma patients are often very limited because of impaired bone marrow function. Bendamustine is effective in multiple myeloma and has a favourable toxicity profile. We hypothesized that dose-intensified bendamustine (180 mg/m2 , day 1 and 2) followed by autologous blood stem cell support (ASCS) would improve bone marrow function with low post-transplant toxicity in patients with severely impaired haematopoiesis. We analyzed 28 consecutive myeloma patients, with a median of three prior lines of therapy (range 2-7), who had relapsed from the last treatment with very limited bone marrow function and were therefore ineligible for conventional chemotherapy, novel agents or trial enrolment. Dose-intensified bendamustine with ASCS improved haematopoiesis as reflected by increased platelet counts (median 40/nl vs 94/nl, p = 0.0004) and white blood cell counts (3.0/nl vs 4.8/nl, p = 0.02) at day +100. The median time until engraftment of platelets (>50/nl) was 11 days (0-24 days) and of white cell counts (>1.0/nl) 0 days (0-24 days). At least, a minimal response was achieved in 36% of patients. The disease stabilization rate was 50% while the median progression-free survival rate was limited to 2.14 months. Most importantly, patients were once again eligible for alternative treatments including enrolment into clinical trials. We conclude that dose-intensified bendamustine followed by ASCS is safe and feasible for multiple myeloma patients with very limited bone marrow reserve. Copyright © 2015 John Wiley & Sons, Ltd.

Identification of HERC5 and its potential role in NSCLC progression.

For better lung cancer diagnosis and therapy, early detection markers of tumor dissemination are urgently needed, as most lung cancers do not show symptoms until extensive metastasis formation has already taken place. Our previous studies showed that in non-small cell lung cancer (NSCLC) early tumor dissemination is associated with a loss of chromosome 4q12-q32 and the presence of disseminated tumor cells (DTC) in the bone marrow. In order to identify the potential target gene in this region, a screen for methylation-dependent expression was performed. Lung cancer cell lines showing a loss of 4q as well as a normal bronchial epithelial cell line as control were treated with 5-aza-2'-deoxycytidine (5-aza-CdR) followed by expression profiling. Seven genes within the 4q target region, which have been associated with a positive DTC status before were found to be regulated by hypermethylation. QRT-PCR in an independent sample set identified HERC5 as a potential target gene. Quantitative methylation analysis of these lung tissue samples revealed that HERC5 promoter hypermethylation was significantly associated with positive DTC status (p = 0.020) and occurrence of brain metastases (p = 0.015). In addition, hypermethylation of the HERC5 promoter in NSCLC was identified as a predictor for poor survival for Stage I adenocarcinoma patients (p = 0.022) and also for poor overall survival in metastatic lung cancer patients (p = 0.028). In conclusion, HERC5 may function as a prognostic marker and is associated with tumor dissemination in lung cancer.

Global alterations of DNA methylation in cholangiocarcinoma target the Wnt signaling pathway.

UNLABELLED: The molecular mechanisms underlying the genesis of cholangiocarcinomas (CCs) are poorly understood. Epigenetic changes such as aberrant hypermethylation and subsequent atypical gene expression are characteristic features of most human cancers. In CC, data regarding global methylation changes are lacking so far. We performed a genome-wide analysis for aberrant promoter methylation in human CCs. We profiled 10 intrahepatic and 8 extrahepatic CCs in comparison to non-neoplastic biliary tissue specimens, using methyl-CpG immunoprecipitation (MCIp) combined with whole-genome CpG island arrays. DNA methylation was confirmed by quantitative mass spectrometric analysis and functional relevance of promoter hypermethylation was shown in demethylation experiments of two CC cell lines using 5-aza-2'deoxycytidine (DAC) treatment. Immunohistochemical staining of tissue microarrays (TMAs) from 223 biliary tract cancers (BTCs) was used to analyze candidate gene expression at the protein level. Differentially methylated, promoter-associated regions were nonrandomly distributed and enriched for genes involved in cancer-related pathways including Wnt, transforming growth factor beta (TGF-ß), and PI3K signaling pathways. In CC cell lines, silencing of genes involved in Wnt signaling, such as SOX17, WNT3A, DKK2, SFRP1, SFRP2, and SFRP4 was reversed after DAC administration. Candidate protein SFRP2 was substantially down-regulated in neoplastic tissues of all BTC subtypes as compared to normal tissues. A significant inverse correlation of SFRP2 protein expression and pT status was found in BTC patients. CONCLUSION: We provide a comprehensive analysis to define the genome-wide methylation landscape of human CC. Several candidate genes of cancer-relevant signaling pathways were identified, and closer analysis of selected Wnt pathway genes confirmed the relevance of this pathway in CC. The presented global methylation data are the basis for future studies on epigenetic changes in cholangiocarcinogenesis.

Adaptation of the Cognitive Screening Test (Triagem Cognitiva - TRIACOG) for computer-mediated assessments: TRIACOG-Online.

This study aims to present the adaptation, evidence of content validity and results of a pilot study of the Cognitive Screening Test - Online (TRIACOG-Online) in a clinical sample of patients after stroke. The process comprised four stages: 1) Adaptation of the instructions, stimulus and responses; 2) Seven experts analyzed the equivalence between the previous printed version and the online version; 3) A pilot study was carried out with seven adults who had experienced a stroke in order to check the comprehension and feasibility of the items; and 4) The development of the final version of TRIACOG-Online. Expert validity testing of the questionnaire yielded a content validity index (CVI) of 100% for correspondence and construct in 13 items, and a CVI of 87.71% in four items. In the pilot study, problems related to the internet led to the decision to use a single section form. No difficulties were observed in carrying out the tasks and understanding the instructions. Participants reported being able to adequately visualize the stimuli and remain motivated to complete the tasks presented. It was shown that TRIACOG-Online evaluated the same constructs as the pencil-and-paper version, can be used in remote neuropsychological assessments and face-to-face settings.

Protein phosphatase 1, regulatory subunit 15B is a survival factor for ERα-positive breast cancer.

Breast cancer is a heterogeneous disease at both the clinical and molecular levels. This heterogeneity may give rise to different therapy responses. Molecular profiling has facilitated identification of signatures for stratifying patients who would potentially benefit from given therapies. Previously, we reported on a subset of genes with the potential for predicting response of primary breast cancer to neoadjuvant chemotherapy. Herein, we report that patients with luminal (estrogen receptor α [ERα]-expressing) breast cancer were enriched for nonresponders. To identify novel factors that contribute to the survival of breast cancer cells, a loss-of-function screen was performed with a subset of genes overexpressed in patients with disease resistant to chemotherapy. This approach led us to identify protein phosphatase 1, regulatory subunit 15B (PPP1R15B) as a factor with a potentially essential role in the survival of ERα-positive breast cancer cells. Functional analyses showed that PPP1R15B depletion results in impaired proliferation due to unsuccessful transition of cells from G1 to S phase of the cell cycle, and apoptosis induction. Moreover, our data revealed a regulatory role for PPP1R15B in activating ERα. Furthermore, a high level of PPP1R15B mRNA expression was associated with poor outcome following tamoxifen-based therapy. Accordingly, knockdown of PPP1R15B expression sensitized tamoxifen-resistant MCF-7 breast cancer cells to tamoxifen while reducing ERα abundance in these cells. Our findings reveal a novel role for PPP1R15B in the survival and therapy response of ERα-positive breast cancer and may open new avenues for tumor subtype-specific therapeutic strategies in the era of personalized medicine.

Assessment of cognitive functioning using the Mini Mental State Examination in men with Alcohol and Crack Cocaine Use Disorder.

INTRODUCTION: Consumption of substances has been associated with cognitive impairment. The Mini Mental State Examination (MMSE) is an easy-to-apply screening tool used to assess cognitive functions. OBJECTIVES: To evaluate the cognitive performance of individuals with alcohol (AUD) and/or crack cocaine use disorder (CUD) and polysubstance use using the MMSE and to investigate the impact of substance use profile and the moderation effect of educational level on MMSE performance. METHODS: Cross-sectional study with 508 adult male inpatients diagnosed with substance use disorders (245 with AUD, 85 with CUD, and 178 with polysubstance use). Cognitive performance was assessed using the MMSE scale (total and composite scores). RESULTS: Individuals with AUD had worse total MMSE scores and scored worse for all three MMSE components compared to individuals with polysubstance use (p < 0.001, oral/written language comprehension, p < 0.001, attention/memory, and p = 0.007, motor functions). MMSE scores were positively correlated with educational level (p < 0.017), but were not associated with age, recent drug use, or years of drug use. Educational level moderated the impact of substance use on MMSE performance, especially total score and composite language comprehension score. Individuals with a low educational level (≤ 8 years) had worse performance than those with a high educational level (≥ 9 years), mainly in individuals with AUD (p < 0.001). DISCUSSION: Individuals with a low educational level and alcohol use are more prone to present cognitive impairment than crack cocaine users, especially involving language aspects. Better-preserved cognitive function could impact treatment adherence and might guide the decision of therapeutic strategies.

No prognostic value of IDH1 mutations in a series of 100 WHO grade II astrocytomas.

Mutations in the gene encoding isocitrate dehydrogenase 1 (IDH1) have been identified in approximately 70-80 % of astrocytomas and oligodendrogliomas of WHO grades II and III, and in secondary glioblastomas. In addition, a low incidence of IDH2 mutations has been detected in these tumors, and the occurence of IDH1 and IDH2 mutations is mutually exclusive. For patients with anaplastic gliomas and glioblastomas with IDH1 mutations, overall survival was significantly longer than for patients with wild-type tumours. However, the prognostic value of IDH1 in low-grade gliomas remains ambiguous. IDH1 codon 132 and IDH2 codon 172 mutation status were determined by direct sequencing for a retrospective series of 100 patients with histologically diagnosed Astrocytomas WHO Grad II (A II), and investigated for association with patient outcome. For the patient cohort analysed, median progression-free survival (PFS) was 44.6 months (95 %-CI 1.0-267.0), time to progression (median time to malignant progression (TtMP) was 74.9 months (95 %-CI 1.6-236.2), and median overall survival (OS) was 81.4 months (95 %-CI 5.5-274.8). IDH1 mutations were identified in 79 % of the patients. IDH2 mutations were not observed. Univariate and multivariate analysis revealed no association between IDH1 mutation status and PFS, TtMP, and OS. Furthermore, there were no significant differences regarding PFS, TtMP, and OS between patients with and without IDH1 mutations who did not receive adjuvant treatment. The prognostic value of IDH1 mutations in low-grade astrocytomas is rather low compared with that in high-grade gliomas.

Elastic SCAD as a novel penalization method for SVM classification tasks in high-dimensional data.

BACKGROUND: Classification and variable selection play an important role in knowledge discovery in high-dimensional data. Although Support Vector Machine (SVM) algorithms are among the most powerful classification and prediction methods with a wide range of scientific applications, the SVM does not include automatic feature selection and therefore a number of feature selection procedures have been developed. Regularisation approaches extend SVM to a feature selection method in a flexible way using penalty functions like LASSO, SCAD and Elastic Net.We propose a novel penalty function for SVM classification tasks, Elastic SCAD, a combination of SCAD and ridge penalties which overcomes the limitations of each penalty alone.Since SVM models are extremely sensitive to the choice of tuning parameters, we adopted an interval search algorithm, which in comparison to a fixed grid search finds rapidly and more precisely a global optimal solution. RESULTS: Feature selection methods with combined penalties (Elastic Net and Elastic SCAD SVMs) are more robust to a change of the model complexity than methods using single penalties. Our simulation study showed that Elastic SCAD SVM outperformed LASSO (L1) and SCAD SVMs. Moreover, Elastic SCAD SVM provided sparser classifiers in terms of median number of features selected than Elastic Net SVM and often better predicted than Elastic Net in terms of misclassification error.Finally, we applied the penalization methods described above on four publicly available breast cancer data sets. Elastic SCAD SVM was the only method providing robust classifiers in sparse and non-sparse situations. CONCLUSIONS: The proposed Elastic SCAD SVM algorithm provides the advantages of the SCAD penalty and at the same time avoids sparsity limitations for non-sparse data. We were first to demonstrate that the integration of the interval search algorithm and penalized SVM classification techniques provides fast solutions on the optimization of tuning parameters.The penalized SVM classification algorithms as well as fixed grid and interval search for finding appropriate tuning parameters were implemented in our freely available R package 'penalizedSVM'.We conclude that the Elastic SCAD SVM is a flexible and robust tool for classification and feature selection tasks for high-dimensional data such as microarray data sets.

Molecular signatures classify astrocytic gliomas by IDH1 mutation status.

To identify novel glioma-associated pathomechanisms and molecular markers, we performed an array-based comparative genomic hybridization analysis of 131 diffuse astrocytic gliomas, including 87 primary glioblastomas (pGBIV), 13 secondary glioblastomas (sGBIV), 19 anaplastic astrocytomas (AAIII) and 12 diffuse astrocytomas (AII). All tumors were additionally screened for IDH1 and IDH2 mutations. Expression profiling was performed for 74 tumors (42 pGBIV, 11 sGBIV, 13 AAIII, 8 AII). Unsupervised and supervised bioinformatic analyses revealed distinct genomic and expression profiles separating pGBIV from the other entities. Classifier expression signatures were strongly associated with the IDH1 gene mutation status. Within pGBIV, the rare subtype of IDH1 mutant tumors shared expression profiles with IDH1 mutant sGBIV and was associated with longer overall survival compared with IDH1 wild-type tumors. In patients with IDH1 wild-type pGBIV, PDGFRA gain or amplification as well as 19q gain were associated with patient outcome. Array-CGH analysis additionally revealed homozygous deletions of the FGFR2 gene at 10q26.13 in 2 pGBIV, with reduced FGFR2 mRNA levels being frequent in pGBIV and linked to poor outcome. In conclusion, we report that diffuse astrocytic gliomas can be separated into 2 major molecular groups with distinct genomic and mRNA profiles as well as IDH1 gene mutation status. In addition, our results suggest FGFR2 as a novel glioma-associated candidate tumor suppressor gene on the long arm of chromosome 10.

BRAF gene duplication constitutes a mechanism of MAPK pathway activation in low-grade astrocytomas.

The molecular pathogenesis of pediatric astrocytomas is still poorly understood. To further understand the genetic abnormalities associated with these tumors, we performed a genome-wide analysis of DNA copy number aberrations in pediatric low-grade astrocytomas by using array-based comparative genomic hybridization. Duplication of the BRAF protooncogene was the most frequent genomic aberration, and tumors with BRAF duplication showed significantly increased mRNA levels of BRAF and a downstream target, CCND1, as compared with tumors without duplication. Furthermore, denaturing HPLC showed that activating BRAF mutations were detected in some of the tumors without BRAF duplication. Similarly, a marked proportion of low-grade astrocytomas from adult patients also had BRAF duplication. Both the stable silencing of BRAF through shRNA lentiviral transduction and pharmacological inhibition of MEK1/2, the immediate downstream phosphorylation target of BRAF, blocked the proliferation and arrested the growth of cultured tumor cells derived from low-grade gliomas. Our findings implicate aberrant activation of the MAPK pathway due to gene duplication or mutation of BRAF as a molecular mechanism of pathogenesis in low-grade astrocytomas and suggest inhibition of the MAPK pathway as a potential treatment.