Search

Supplementary materials and methods for ALDEFLUOR assay and CD44 staining and data showing that patient-derived GSCs are enriched for ALDH+ cells and express CD44.

Supplementary Tables 1 and 2, Figures 1-9. Supplemental Table 1. H-score analysis for SPRY2 staining in 10 EGFRvIII-negative and 10 EGFRvIII-positive tumors. Supplemental Table 2. Upregulated genes shared by human GBMs expressing EGFRvIII and 9L.EGFRvIII rat tumors compared to wild-type EGFR human GBMs or 9L.EV rat tumors. Supplemental Figure 1. SPRY2 is efficiently knocked down by shRNA expression in GBM cell lines. Supplemental Figure 2. SPRY2 knockdown by a second non-overlapping shRNA reduces cellular proliferation, and SPRY2 knockdown by transient siRNA transfection reduces colony formation in soft agar in EGFRvIII-expressing cells. Supplemental Figure 3. SPRY2 knockdown increases cellular sensitivity to EGFR and c-MET co-inhibition. Supplemental Figure 4. SPRY2 knockdown promotes response to EGFR and c-MET coinhibition in GSC cells. Supplemental Figure 5. p38 and JNK control anchorage-independent growth and response to EGFR and c-MET co-inhibition. Supplemental Figure 6. shRNA-mediated knockdown of MKP-1 or MKP-5 reduces MKP-1 or MKP-5 mRNA level. Supplemental Figure 7. SPRY2 protein expression in kidney and cerebellum sections by immunohistochemical analysis. Supplemental Figure 8. SPRY2 correlates well with ERK phosphorylation in a panel of GBM cell lines. Supplemental Figure 9. TCGA GBM dataset analysis reveals that SPRY2 expression is associated with reduced patient survival.

Comparing diverse single-cell RNA sequencing (scRNA-seq) datasets generated by different technologies and in different laboratories remains a major challenge. Here we address the need for guidance in choosing algorithms leading to accurate biological interpretations of varied data types acquired with different platforms. Using two well-characterized cellular reference samples (breast cancer cells and B cells), captured either separately or in mixtures, we compared different scRNA-seq platforms and several preprocessing, normalization and batch-effect correction methods at multiple centers. Although preprocessing and normalization contributed to variability in gene detection and cell classification, batch-effect correction was by far the most important factor in correctly classifying the cells. Moreover, scRNA-seq dataset characteristics (for example, sample and cellular heterogeneity and platform used) were critical in determining the optimal bioinformatic method. However, reproducibility across centers and platforms was high when appropriate bioinformatic methods were applied. Our findings offer practical guidance for optimizing platform and software selection when designing an scRNA-seq study. A comprehensive comparison of 20 single-cell RNA-seq datasets derived from the two cell lines analyzed using six preprocessing pipelines, eight normalization methods and seven batch-correction algorithms derived from four different sequencing platforms at different centers.

Inflammatory bowel disease (IBD) is characterized by a dysregulated intestinal epithelial barrier leading to breach of barrier immunity. Here we identified similar protein expression changes between IBD and Citrobacter rodentium-infected FVB mice with respect to dysregulation of solute transporters as well as components critical for intestinal barrier integrity. We attribute the disease associated changes in the model to the emergence of undifferentiated intermediate intestinal epithelial cells. Prophylactic treatment with IL-22.Fc in C. rodentium-infected FVB mice reduced disease severity and rescued the mice from lethality. Multi-omics and solute analyses revealed that IL-22.Fc treatment prevented disease-associated changes including disruption of the solute transporter machinery and restored proper physiological functions of the intestine, respectively. Taken together, we established the disease relevance of the C. rodentium-induced colitis model to IBD, demonstrated the protective role of IL-22 in amelioration of epithelial dysfunction and elucidated the molecular mechanisms with IL-22's effect on intestinal epithelial cells.

How mammalian neuronal identity is progressively acquired and reinforced during development is not understood. We have previously shown that loss of RP58 (ZNF238 or ZBTB18), a BTB/POZ-zinc finger-containing transcription factor, in the mouse brain leads to microcephaly, corpus callosum agenesis, and cerebellum hypoplasia and that it is required for normal neuronal differentiation. The transcriptional programs regulated by RP58 during this process are not known. Here, we report for the first time that in embryonic mouse neocortical neurons a complex set of genes normally expressed in other cell types, such as those from mesoderm derivatives, must be actively repressed in vivo and that RP58 is a critical regulator of these repressed transcriptional programs. Importantly, gene set enrichment analysis (GSEA) analyses of these transcriptional programs indicate that repressed genes include distinct sets of genes significantly associated with glioma progression and/or pluripotency. We also demonstrate that reintroducing RP58 in glioma stem cells leads not only to aspects of neuronal differentiation but also to loss of stem cell characteristics, including loss of stem cell markers and decrease in stem cell self-renewal capacities. Thus, RP58 acts as an in vivo master guardian of the neuronal identity transcriptome, and its function may be required to prevent brain disease development, including glioma progression.

Hidradenitis suppurativa (HS) is a highly prevalent, morbid inflammatory skin disease with limited treatment options. The major cell types and inflammatory pathways in skin of patients with HS are poorly understood, and which patients will respond to TNF-α blockade is currently unknown. We discovered that clinically and histologically healthy appearing skin (i.e., nonlesional skin) is dysfunctional in patients with HS with a relative loss of immune regulatory pathways. HS skin lesions were characterized by quantitative and qualitative dysfunction of type 2 conventional dendritic cells, relatively reduced regulatory T cells, an influx of memory B cells, and a plasma cell/plasmablast infiltrate predominantly in end-stage fibrotic skin. At the molecular level, there was a relative bias toward the IL-1 pathway and type 1 T cell responses when compared with both healthy skin and psoriatic patient skin. Anti–TNF-α therapy markedly attenuated B cell activation with minimal effect on other inflammatory pathways. Finally, we identified an immune activation signature in skin before anti–TNF-α treatment that correlated with subsequent lack of response to this modality. Our results reveal the fundamental immunopathogenesis of HS and provide a molecular foundation for future studies focused on stratifying patients based on likelihood of clinical response to TNF-α blockade., Hidradenitis suppurativa lesions are characterized by alterations in cDC2s, relatively reduced Tregs, an influx of memory B cells and plasma cells, and biases towards IL1 pathway activation and type 1 T cell responses.

Molecular stratification of cancer patients is driving the development of precision medicine-targeted therapies. Clustering of tumor samples based on gene expression profiles from high-throughput platforms, such as microarrays or NextGen sequencing, has resulted in distinct tumor subtypes for numerous cancers. However, the majority of the derived classifiers or gene signatures have not reached clinical utility. Therefore, informatics methods to accurately translate the derived gene-signature from the high-throughput platform to a clinically adaptable low-dimensional platform are critical. In this chapter, we discuss a workflow to derive and then transfer gene signatures from one analytical platform to another for cancer patient stratification. We summarize the results of the workflow on two different cancers. Finally we discuss the importance of data-discretization in dealing with the cross-platform data and incorporating the splice-variant or isoform-level gene expression profiles in the statistical analyses.

Comparing diverse single-cell RNA sequencing (scRNA-seq) datasets generated by different technologies and in different laboratories remains a major challenge. Here we address the need for guidance in choosing algorithms leading to accurate biological interpretations of varied data types acquired with different platforms. Using two well-characterized cellular reference samples (breast cancer cells and B cells), captured either separately or in mixtures, we compared different scRNA-seq platforms and several preprocessing, normalization and batch-effect correction methods at multiple centers. Although preprocessing and normalization contributed to variability in gene detection and cell classification, batch-effect correction was by far the most important factor in correctly classifying the cells. Moreover, scRNA-seq dataset characteristics (for example, sample and cellular heterogeneity and platform used) were critical in determining the optimal bioinformatic method. However, reproducibility across centers and platforms was high when appropriate bioinformatic methods were applied. Our findings offer practical guidance for optimizing platform and software selection when designing an scRNA-seq study.

Many psoriasis patients treated with biologics do not achieve total skin clearance. These patients possess residual plaques despite ongoing biologic treatment. To elucidate mechanisms of plaque persistence despite overall good drug response, we studied 50 subjects: psoriasis patients with residual plaques treated with one of three different biologics, untreated patients, and healthy controls. Skin biopsies from all subjects were characterized using three methods: mRNA expression, histology, and FACS of hematopoietic skin cells. Although all three methods provided evidence of drug effect, gene expression analysis revealed the persistence of key psoriasis pathways in treated plaques, including granulocyte adhesion and diapedesis, T helper type17 activation pathway, and interferon signaling with no novel pathways emerging. Focal decreases in parakeratosis and keratinocyte proliferation and differential reduction in IL-17 producing CD103– T cells, but no change in CD103+ tissue-resident memory T cells were observed. Of note, antitumor necrosis factor increased the interferon signaling pathway already present. Interestingly mast cells were the dominant source of IL-22 in all psoriasis subjects. These data suggest that while subtle differences can be observed in drug-treated plaques, underlying biologic mechanisms are similar to those present in untreated psoriatic lesions.

Protein ubiquitination is mediated sequentially by ubiquitin activating enzyme E1, ubiquitin conjugating enzyme E2 and ubiquitin ligase E3. Uba1 was thought to be the only E1 until the recent identification of Uba6. To differentiate the biological functions of Uba1 and Uba6, we applied an orthogonal ubiquitin transfer (OUT) technology to profile their ubiquitination targets in mammalian cells. By expressing pairs of an engineered ubiquitin and engineered Uba1 or Uba6 that were generated for exclusive interactions, we identified 697 potential Uba6 targets and 527 potential Uba1 targets with 258 overlaps. Bioinformatics analysis reveals substantial differences in pathways involving Uba1- and Uba6-specific targets. We demonstrate that polyubiquitination and proteasomal degradation of ezrin and CUGBP1 require Uba6, but not Uba1, and that Uba6 is involved in the control of ezrin localization and epithelial morphogenesis. These data suggest that distinctive substrate pools exist for Uba1 and Uba6 that reflect non-redundant biological roles for Uba6., The transfer of ubiquitin (UB) to cellular targets is mediated sequentially by three groups of enzymes, UB activating enzyme (E1), UB conjugating enzyme (E2) and UB ligase (E3). Here the authors provide evidence that the two mammalian E1 enzymes, Uba1 and Uba6, exert biologically distinct functions.

Glioblastoma multiforme remains the deadliest malignant brain tumor, with glioma stem cells (GSC) contributing to treatment resistance and tumor recurrence. We have identified MAPK-interacting kinases (MNK) as potential targets for the GSC population in glioblastoma multiforme. Isoform-level subtyping using The Cancer Genome Atlas revealed that both MNK genes (MKNK1 and MKNK2) are upregulated in mesenchymal glioblastoma multiforme as compared with other subtypes. Expression of MKNK1 is associated with increased glioma grade and correlated with the mesenchymal GSC marker, CD44, and coexpression of MKNK1 and CD44 predicts poor survival in glioblastoma multiforme. In established and patient-derived cell lines, pharmacologic MNK inhibition using LY2801653 (merestinib) inhibited phosphorylation of the eukaryotic translation initiation factor 4E, a crucial effector for MNK-induced mRNA translation in cancer cells and a marker of transformation. Importantly, merestinib inhibited growth of GSCs grown as neurospheres as determined by extreme limiting dilution analysis. When the effects of merestinib were assessed in vivo using an intracranial xenograft mouse model, improved overall survival was observed in merestinib-treated mice. Taken together, these data provide strong preclinical evidence that pharmacologic MNK inhibition targets mesenchymal glioblastoma multiforme and its GSC population. Implications: These findings raise the possibility of MNK inhibition as a viable therapeutic approach to target the mesenchymal subtype of glioblastoma multiforme. Mol Cancer Res; 14(10); 984–93. ©2016 AACR.

Background Crohn's disease (CD) and ulcerative colitis (UC) are intestinal chronic inflammatory conditions characterized by altered epithelial barrier function and tissue damage. Despite significant efforts to understanding the biological mechanisms responsible for gut inflammation, the pathophysiology of CD and UC remains poorly understood. Methods To help elucidate the potential mechanisms responsible for gut inflammation in CD and UC, transcriptomic and proteomic profiling of human colon biopsy specimens was performed. Dysregulated genes and proteins in disease tissues compared with normal tissues were characterized from the expression profiles and further subjected to pathway analysis to identify altered biological processes and signaling pathways. Results Sample analysis showed 4250 genes with matched protein expression and a wide range of correlation of RNA-protein abundance across samples. Pathway analysis of dysregulated genes and proteins in CD and UC showed alterations in immune and inflammatory responses, complement cascade, and the suppression of metabolic processes and PPAR signaling. In CD, increased T-helper cell differentiation and elevated toll-like receptor and JAK/STAT signaling were observed. Interestingly, increased MAPK signaling was only observed in UC. Weighted gene co-expression network analysis suggested a possible role of epigenetic regulation in UC. Of note, a large discrepancy between regulation of RNA and protein levels in inflamed colon samples was detected for previously identified biomarkers including MMP14 and LAMP1. Conclusions With the analysis of dysregulated genes and pathways, the present study unravels key mechanisms contributing to CD and UC pathogenesis and emphasizes that integrative analysis of multi-omics data sets can provide more insight into understanding complex disease mechanisms.

Background Previous studies have identified genetic variants associated with bronchopulmonary dysplasia (BPD) in extremely preterm infants. However, findings with genome-wide significance have been rare, and not replicated. We hypothesized that whole exome sequencing (WES) of premature subjects with extremely divergent phenotypic outcomes could facilitate the identification of genetic variants or gene networks contributing disease risk. Results The Prematurity and Respiratory Outcomes Program (PROP) recruited a cohort of > 765 extremely preterm infants for the identification of markers of respiratory morbidity. We completed WES on 146 PROP subjects (85 affected, 61 unaffected) representing extreme phenotypes of early respiratory morbidity. We tested for association between disease status and individual common variants, screened for rare variants exclusive to either affected or unaffected subjects, and tested the combined association of variants across gene loci. Pathway analysis was performed and disease-related expression patterns were assessed. Marginal association with BPD was observed for numerous common and rare variants. We identified 345 genes with variants unique to BPD-affected preterm subjects, and 292 genes with variants unique to our unaffected preterm subjects. Of these unique variants, 28 (19 in the affected cohort and 9 in unaffected cohort) replicate a prior WES study of BPD-associated variants. Pathway analysis of sets of variants, informed by disease-related gene expression, implicated protein kinase A, MAPK and Neuregulin/epidermal growth factor receptor signaling. Conclusions We identified novel genes and associated pathways that may play an important role in susceptibility/resilience for the development of lung disease in preterm infants. Electronic supplementary material The online version of this article (10.1186/s12863-018-0679-7) contains supplementary material, which is available to authorized users.

Stratification of cancer patients into different molecular groups is essential for developing targeted therapies. High-throughput technologies, such as microarrays and next-generation sequencing, have been extensively used for generating multi-omics data. Indeed, The Cancer Genome Atlas (TCGA) consortium, one of the most comprehensive and popular databases of cancer, has been accumulating large volumes of invaluable data for more than 30 cancer types, offering unprecedented opportunity to attain new insights in cancer biology. Despite the technological advances, analyzing, integrating and translating the gene signatures across different platforms remains a computationally challenging task. Here, we developed a novel computational framework that integrates genomic and clinical data to stratify cancer patients into different molecular subgroups and predict clinically applicable phenotypes, such as survival. Application of this user-friendly framework derives platform-independent isoform-level gene signatures that can be translated from high-dimensional platforms (e.g., RNA-Seq) to clinically adaptable low-dimensional molecular assays (e.g., RT-PCR) for prognostic stratification. We applied the pipeline on two TCGA lung cancer datasets—Lung Adenocarcinoma (LUAD) and Lung Squamous Cell Carcinoma (LUSC). Using independent test data, we achieved about 93% (LUAD) and 98% (LUSC) classification accuracy using less than 70 isoform-level gene signature. The proposed informatics platform is applicable to other cancer types in TCGA data portal. Citation Format: Segun Jung, Yingtao Bi, Ramana V. Davuluri. Informatics framework for clustering and deriving gene signatures for prognostic stratification of cancer patients. [abstract]. In: Proceedings of the AACR Special Conference on Translation of the Cancer Genome; Feb 7-9, 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 1):Abstract nr A2-47.

Engineering the ubiquitin transfer cascades by phage display enables an efficient way to profile E3 substrates., E3 ubiquitin (UB) ligases E4B and carboxyl terminus of Hsc70-interacting protein (CHIP) use a common U-box motif to transfer UB from E1 and E2 enzymes to their substrate proteins and regulate diverse cellular processes. To profile their ubiquitination targets in the cell, we used phage display to engineer E2-E4B and E2-CHIP pairs that were free of cross-reactivity with the native UB transfer cascades. We then used the engineered E2-E3 pairs to construct “orthogonal UB transfer (OUT)” cascades so that a mutant UB (xUB) could be exclusively used by the engineered E4B or CHIP to label their substrate proteins. Purification of xUB-conjugated proteins followed by proteomics analysis enabled the identification of hundreds of potential substrates of E4B and CHIP in human embryonic kidney 293 cells. Kinase MAPK3 (mitogen-activated protein kinase 3), methyltransferase PRMT1 (protein arginine N-methyltransferase 1), and phosphatase PPP3CA (protein phosphatase 3 catalytic subunit alpha) were identified as the shared substrates of the two E3s. Phosphatase PGAM5 (phosphoglycerate mutase 5) and deubiquitinase OTUB1 (ovarian tumor domain containing ubiquitin aldehyde binding protein 1) were confirmed as E4B substrates, and β-catenin and CDK4 (cyclin-dependent kinase 4) were confirmed as CHIP substrates. On the basis of the CHIP-CDK4 circuit identified by OUT, we revealed that CHIP signals CDK4 degradation in response to endoplasmic reticulum stress.

CNS Primitive Neuroectodermal tumors (CNS-PNETs) are members of the embryonal family of malignant childhood brain tumors, which remain refractory to current therapeutic treatments. Current paradigm of brain tumorigenesis implicates brain tumor-initiating cells (BTIC) in the onset of tumorigenesis and tumor maintenance. However, despite their significance, there is currently no comprehensive characterization of CNS-PNETs BTICs. Recently, we described an animal model of CNS-PNET generated by orthotopic transplantation of human Radial Glial (RG) cells - the progenitor cells for adult neural stem cells (NSC) - into NOD-SCID mice brain and proposed that BTICs may play a role in the maintenance of these tumors. Here we report the characterization of BTIC lines derived from this CNS-PNET animal model. BTIC's orthotopic transplantation generated highly aggressive tumors also characterized as CNS-PNETs. The BTICs have the hallmarks of NSCs as they demonstrate self-renewing capacity and have the ability to differentiate into astrocytes and early migrating neurons. Moreover, the cells demonstrate aberrant accumulation of wild type tumor-suppressor protein p53, indicating its functional inactivation, highly up-regulated levels of onco-protein cMYC and the BTIC marker OCT3/4, along with metabolic switch to glycolysis - suggesting that these changes occurred in the early stages of tumorigenesis. Furthermore, based on RNA- and DNA-seq data, the BTICs did not acquire any transcriptome-changing genomic alterations indicating that the onset of tumorigenesis may be epigenetically driven. The study of these BTIC self-renewing cells in our model may enable uncovering the molecular alterations that are responsible for the onset and maintenance of the malignant PNET phenotype.

Oncogenic mutations in two isocitrate dehydrogenase (IDH)-encoding genes (IDH1 and IDH2) have been identified in acute myelogenous leukemia, low-grade glioma, and secondary glioblastoma (GBM). Our in silico and wet-bench analyses indicate that non-mutated IDH1 mRNA and protein are commonly overexpressed in primary GBMs. We show that genetic and pharmacologic inactivation of IDH1 decreases GBM cell growth, promotes a more differentiated tumor cell state, increases apoptosis in response to targeted therapies, and prolongs the survival of animal subjects bearing patient-derived xenografts (PDXs). On a molecular level, diminished IDH1 activity results in reduced α-ketoglutarate (αKG) and NADPH production, paralleled by deficient carbon flux from glucose or acetate into lipids, exhaustion of reduced glutathione, increased levels of reactive oxygen species (ROS), and enhanced histone methylation and differentiation marker expression. These findings suggest that IDH1 upregulation represents a common metabolic adaptation by GBMs to support macromolecular synthesis, aggressive growth, and therapy resistance.

Recently, we described a new animal model of CNS primitive neuroectodermal tumors (CNS-PNET), which was generated by orthotopic transplantation of human Radial Glial (RG) cells into NOD-SCID mice’s brain sub-ventricular zone. In the current study we conducted comprehensive RNA-Seq analyses to gain insights on the mechanisms underlying tumorigenesis in this mouse model of CNS-PNET. Here we show that the RNA-Seq profiles derived from these tumors cluster with those reported for patients’ PNETs. Moreover, we found that (i) stabilization of HIF-1α and HIF-2α, which are involved in mediation of the hypoxic responses in the majority of cell types, (ii) up-regulation of MYCC, a key onco-protein whose dysregulation occurs in ~70% of human tumors, and (iii) accumulation of stabilized p53, which is commonly altered in human cancers, constitute hallmarks of our tumor model, and might represent the basis for CNS-PNET tumorigenesis in this model. We discuss the possibility that these three events might be interconnected. These results indicate that our model may prove invaluable to uncover the molecular events leading to MYCC and TP53 alterations, which would be of broader interest considering their relevance to many human malignancies. Lastly, this mouse model might prove useful for drug screening targeting MYCC and related members of its protein interaction network.

IntroductionBreast cancer being a multifaceted disease constitutes a wide spectrum of histological and molecular variability in tumors. However, the task for the identification of these variances is complicated by the interplay between inherited genetic and epigenetic aberrations. Therefore, this study provides an extrapolate outlook to the sinister partnership between DNA methylation and single-nucleotide polymorphisms (SNPs) in relevance to the identification of prognostic markers in breast cancer. The effect of these SNPs on methylation is defined as methylation quantitative trait loci (meQTL).Materialsand MethodsWe developed a novel method to identify prognostic gene signatures for breast cancer by integrating genomic and epigenomic data. This is based on the hypothesis that multiple sources of evidence pointing to the same gene or pathway are likely to lead to reduced false positives. We also apply random resampling to reduce overfitting noise by dividing samples into training and testing data sets. Specifically, the common samples between Illumina 450 DNA methylation, Affymetrix SNP array, and clinical data sets obtained from the Cancer Genome Atlas (TCGA) for breast invasive carcinoma (BRCA) were randomly divided into training and test models. An intensive statistical analysis based on log-rank test and Cox proportional hazard model has established a significant association between differential methylation and the stratification of breast cancer patients into high- and low-risk groups, respectively.ResultsThe comprehensive assessment based on the conjoint effect of CpG–SNP pair has guided in delaminating the breast cancer patients into the high- and low-risk groups. In particular, the most significant association was found with respect to cg05370838–rs2230576, cg00956490–rs940453, and cg11340537–rs2640785 CpG–SNP pairs. These CpG–SNP pairs were strongly associated with differential expression of ADAM8, CREB5, and EXPH5 genes, respectively. Besides, the exclusive effect of SNPs such as rs10101376, rs140679, and rs1538146 also hold significant prognostic determinant.ConclusionsThus, the analysis based on DNA methylation and SNPs have resulted in the identification of novel susceptible loci that hold prognostic relevance in breast cancer.

Author(s): Macyszyn, Luke; Akbari, Hamed; Pisapia, Jared M; Da, Xiao; Attiah, Mark; Pigrish, Vadim; Bi, Yingtao; Pal, Sharmistha; Davuluri, Ramana V; Roccograndi, Laura; Dahmane, Nadia; Martinez-Lage, Maria; Biros, George; Wolf, Ronald L; Bilello, Michel; O'Rourke, Donald M; Davatzikos, Christos | Abstract: BackgroundMRI characteristics of brain gliomas have been used to predict clinical outcome and molecular tumor characteristics. However, previously reported imaging biomarkers have not been sufficiently accurate or reproducible to enter routine clinical practice and often rely on relatively simple MRI measures. The current study leverages advanced image analysis and machine learning algorithms to identify complex and reproducible imaging patterns predictive of overall survival and molecular subtype in glioblastoma (GB).MethodsOne hundred five patients with GB were first used to extract approximately 60 diverse features from preoperative multiparametric MRIs. These imaging features were used by a machine learning algorithm to derive imaging predictors of patient survival and molecular subtype. Cross-validation ensured generalizability of these predictors to new patients. Subsequently, the predictors were evaluated in a prospective cohort of 29 new patients.ResultsSurvival curves yielded a hazard ratio of 10.64 for predicted long versus short survivors. The overall, 3-way (long/medium/short survival) accuracy in the prospective cohort approached 80%. Classification of patients into the 4 molecular subtypes of GB achieved 76% accuracy.ConclusionsBy employing machine learning techniques, we were able to demonstrate that imaging patterns are highly predictive of patient survival. Additionally, we found that GB subtypes have distinctive imaging phenotypes. These results reveal that when imaging markers related to infiltration, cell density, microvascularity, and blood-brain barrier compromise are integrated via advanced pattern analysis methods, they form very accurate predictive biomarkers. These predictive markers used solely preoperative images, hence they can significantly augment diagnosis and treatment of GB patients.

A heteroscedastic linear regression model is developed from plausible assumptions that describe the time evolution of performance metrics for equipment. The inherited motivation for the related weighted least squares analysis of the model is an essential and attractive selling point to engineers with interest in equipment surveillance methodologies. A simple test for the significance of the heteroscedasticity suggested by a data set is derived and a simulation study is used to evaluate the power of the test and compare it with several other applicable tests that were designed under different contexts. Tolerance intervals within the context of the model are derived, thus generalizing well-known tolerance intervals for ordinary least squares regression. Use of the model and its associated analyses is illustrated with an aerospace application where hundreds of electronic components are continuously monitored by an automated system that flags components that are suspected of unusual degradation patterns.

Background Many supervised learning algorithms have been applied in deriving gene signatures for patient stratification from gene expression data. However, transferring the multi-gene signatures from one analytical platform to another without loss of classification accuracy is a major challenge. Here, we compared three unsupervised data discretization methods--Equal-width binning, Equal-frequency binning, and k-means clustering--in accurately classifying the four known subtypes of glioblastoma multiforme (GBM) when the classification algorithms were trained on the isoform-level gene expression profiles from exon-array platform and tested on the corresponding profiles from RNA-seq data. Results We applied an integrated machine learning framework that involves three sequential steps; feature selection, data discretization, and classification. For models trained and tested on exon-array data, the addition of data discretization step led to robust and accurate predictive models with fewer number of variables in the final models. For models trained on exon-array data and tested on RNA-seq data, the addition of data discretization step dramatically improved the classification accuracies with Equal-frequency binning showing the highest improvement with more than 90% accuracies for all the models with features chosen by Random Forest based feature selection. Overall, SVM classifier coupled with Equal-frequency binning achieved the best accuracy (> 95%). Without data discretization, however, only 73.6% accuracy was achieved at most. Conclusions The classification algorithms, trained and tested on data from the same platform, yielded similar accuracies in predicting the four GBM subgroups. However, when dealing with cross-platform data, from exon-array to RNA-seq, the classifiers yielded stable models with highest classification accuracies on data transformed by Equal frequency binning. The approach presented here is generally applicable to other cancer types for classification and identification of molecular subgroups by integrating data across different gene expression platforms.

Given that the majority of multi-exon genes generate diverse functional products, it is important to evaluate expression at the isoform level. Previous studies have demonstrated strong gene-level correlations between RNA sequencing (RNA-seq) and microarray platforms, but have not studied their concordance at the isoform level. We performed transcript abundance estimation on raw RNA-seq and exon-array expression profiles available for common glioblastoma multiforme samples from The Cancer Genome Atlas using different analysis pipelines, and compared both the isoform- and gene-level expression estimates between programs and platforms. The results showed better concordance between RNA-seq/exon-array and reverse transcription-quantitative polymerase chain reaction (RT-qPCR) platforms for fold change estimates than for raw abundance estimates, suggesting that fold change normalization against a control is an important step for integrating expression data across platforms. Based on RT-qPCR validations, eXpress and Multi-Mapping Bayesian Gene eXpression (MMBGX) programs achieved the best performance for RNA-seq and exon-array platforms, respectively, for deriving the isoform-level fold change values. While eXpress achieved the highest correlation with the RT-qPCR and exon-array (MMBGX) results overall, RSEM was more highly correlated with MMBGX for the subset of transcripts that are highly variable across the samples. eXpress appears to be most successful in discriminating lowly expressed transcripts, but IsoformEx and RSEM correlate more strongly with MMBGX for highly expressed transcripts. The results also reinforce how potentially important isoform-level expression changes can be masked by gene-level estimates, and demonstrate that exon arrays yield comparable results to RNA-seq for evaluating isoform-level expression changes.

BACKGROUND: Recent progress in medulloblastoma biology has confirmed the presence of 4 clinically and molecularly distinct subgroups. MicroRNAs are dysregulated in cancer, including pediatric central nervous system tumors. We aim to determine if microRNA expression can aid in the molecular classification of medulloblastoma in tissue and cerebrospinal fluid (CSF). METHODS: RNA purification from 41 formalin-fixed paraffin embedded primary medulloblastoma samples was performed. TaqMan® miRNA assays were used to quantify the levels of 762 mature miRNAs from each sample using the Applied Biosystems 7900HT Fast Real-Time PCR system. Molecular subgrouping using established gene sets was performed using the Nanostring® platform. Clinical, pathologic and outcome data was collected on all patients. In a separate cohort of patients with medulloblastoma, expression of circulating miRNA was quantified in CSF at the time of diagnosis. RESULTS: Molecular subgrouping by gene expression classified tumors into WNT (n = 6), Shh (n = 5), Group 3 (n = 11) and Group 4 (n = 19) tumors. Clinical and outcome data was consistent with previously published cohorts. Consensus clustering by microRNA expression is not sufficient for molecular subgrouping. Independent of subgroup, the presence of metastatic disease at diagnosis was associated with high levels of miR-373 (p = 0.003) and miR-213 (p = 0.003) and expression of miR-497 was associated with poor overall survival (p =