More accurate assessment of prognosis is important to further improve the choice of risk-related therapy in neuroblastoma (NB) patients. In this study, we aimed to establish and validate a prognostic miRNA signature for children with NB and tested it in both fresh frozen and archived formalin-fixed paraffin-embedded (FFPE) samples.
Four hundred-thirty human mature miRNAs were profiled in two patient subgroups with maximally divergent clinical courses. Univariate logistic regression analysis was used to select miRNAs correlating with NB patient survival. A 25-miRNA gene signature was built using 51 training samples, tested on 179 test samples, and validated on an independent set of 304 fresh frozen tumor samples and 75 archived FFPE samples.
The 25-miRNA signature significantly discriminates the test patients with respect to progression-free and overall survival (P < 0.0001), both in the overall population and in the cohort of high-risk patients. Multivariate analysis indicates that the miRNA signature is an independent predictor of patient survival after controlling for current risk factors. The results were confirmed in an external validation set. In contrast to a previously published mRNA classifier, the 25-miRNA signature was found to be predictive for patient survival in a set of 75 FFPE neuroblastoma samples.
In this study, we present the largest NB miRNA expression study so far, including more than 500 NB patients. We established and validated a robust miRNA classifier, able to identify a cohort of high-risk NB patients at greater risk for adverse outcome using both fresh frozen and archived material.
Bovine mastitis is an inflammation-driven disease of the bovine mammary gland that costs the global dairy industry several billion dollars per year. Because disease susceptibility is a multifactorial complex phenotype, an integrative biology approach is required to dissect the molecular networks involved. Here, we report such an approach using next-generation sequencing combined with advanced network and pathway biology methods to simultaneously profile mRNA and miRNA expression at multiple time points (0, 12, 24, 36 and 48 hr) in milk and blood FACS-isolated CD14+ monocytes from animals infected in vivo with Streptococcus uberis. More than 3700 differentially expressed (DE) genes were identified in milk-isolated monocytes (MIMs), a key immune cell recruited to the site of infection during mastitis. Upregulated genes were significantly enriched for inflammatory pathways, whereas downregulated genes were enriched for nonglycolytic metabolic pathways. Monocyte transcriptional changes in the blood, however, were more subtle but highlighted the impact of this infection systemically. Genes upregulated in blood-isolated monocytes (BIMs) showed a significant association with interferon and chemokine signaling. Furthermore, 26 miRNAs were DE in MIMs and three were DE in BIMs. Pathway analysis revealed that predicted targets of downregulated miRNAs were highly enriched for roles in innate immunity (FDR < 3.4E−8), particularly TLR signaling, whereas upregulated miRNAs preferentially targeted genes involved in metabolism. We conclude that during S. uberis infection miRNAs are key amplifiers of monocyte inflammatory response networks and repressors of several metabolic pathways.
infection; innate immunity; RNAseq; microRNA; transcriptional networks; complex genetics; tolerance; complex immunity; resistance
MicroRNAs (miRNAs) are small non-coding RNA molecules that regulate gene expression at a post-transcriptional level. An miRNA may target many messenger RNA (mRNA) transcripts, and each transcript may be targeted by multiple miRNAs. Our understanding of miRNA regulation is evolving to consider modules of miRNAs that regulate groups of functionally related mRNAs. Here we expand the model of miRNA functional modules and use it to guide the integration of miRNA and mRNA expression and target prediction data. We present evidence of cooperativity between miRNA classes within this integrated miRNA–mRNA association matrix. We then apply bicluster analysis to uncover miRNA functional modules within this integrated data set and develop a novel application to visualize and query these results. We show that this wholly unsupervised approach can discover a network of miRNA–mRNA modules that are enriched for both biological processes and miRNA classes. We apply this method to investigate the interplay of miRNAs and mRNAs in integrated data sets derived from neuroblastoma and human immune cells. This study is the first to apply the technique of biclustering to model functional modules within an integrated miRNA–mRNA association matrix. Results provide evidence of an extensive modular miRNA functional network and enable characterization of miRNA function and dysregulation in disease.
MicroRNAs contribute to the pathogenesis of many forms of cancer, including the pediatric cancer neuroblastoma, but the underlying mechanisms leading to altered miRNA expression are often unknown. Here, a novel integrated approach for analyzing DNA methylation coupled with miRNA and mRNA expression data sets identified 67 epigenetically regulated miRNA in neuroblastoma. A large proportion (42%) of these miRNAs were associated with poor patient survival when under-expressed in tumors. Moreover, we demonstrate that this panel of epigenetically silenced miRNAs targets a large set of genes that are over-expressed in tumors from patients with poor survival in a highly redundant manner. The genes targeted by the epigenetically regulated miRNAs are enriched for a number of biological processes, including regulation of cell differentiation. Functional studies involving ectopic over-expression of several of the epigenetically silenced miRNAs had a negative impact on neuroblastoma cell viability, providing further support to the concept that inactivation of these miRNAs is important for neuroblastoma disease pathogenesis. One locus, miR-340, induced either differentiation or apoptosis in a cell context dependent manner, indicating a tumor suppressive function for this miRNA. Intriguingly, it was determined that miR-340 is up-regulated by demethylation of an upstream genomic region that occurs during the process of neuroblastoma cell differentiation induced by all-trans retinoic acid (ATRA). Further biological studies of miR-340 revealed that it directly represses the SOX2 transcription factor by targeting of its 3’ UTR, explaining the mechanism by which SOX2 is down-regulated by ATRA. Although SOX2 contributes to the maintenance of stem cells in an undifferentiated state, we demonstrate that miR-340 mediated down-regulation of SOX2 is not required for ATRA induced differentiation to occur. In summary, our results exemplify the dynamic nature of the miRNA epigenome and identify a remarkable network of miRNA/mRNA interactions that significantly contribute to neuroblastoma disease pathogenesis.
miRNA; methylation; tumor suppressor; neuroblastoma; SOX2
In contrast to adult mutant gastrointestinal stromal tumors [GISTs], pediatric/wild-type GISTs remain poorly understood overall, given their lack of oncogenic activating tyrosine kinase mutations. These GISTs, with a predilection for gastric origin in female patients, show limited response to therapy with tyrosine kinase inhibitors and generally pursue a more indolent course, but still may prove fatal. Defective cellular respiration appears to underpin tumor development in these wild-type cases, which as a group lack expression of succinate dehydrogenase [SDH] B, a surrogate marker for respiratory chain metabolism. Yet, only a small subset of the wild-type tumors show mutations in the genes coding for the SDH subunits [SDHx]. To explore additional pathogenetic mechanisms in these wild-type GISTs, we elected to investigate post-transcriptional regulation of these tumors by conducting microRNA (miRNA) profiling of a mixed cohort of 73 cases including 18 gastric pediatric wild-type, 25 (20 gastric, 4 small bowel and 1 retroperitoneal) adult wild-type GISTs and 30 gastric adult mutant GISTs. By this approach we have identified distinct signatures for GIST subtypes which correlate tightly with clinico-pathological parameters. A cluster of miRNAs on 14q32 show strikingly different expression patterns amongst GISTs, a finding which appears to be explained at least in part by differential allelic methylation of this imprinted region. Small bowel and retroperitoneal wild-type GISTs segregate with adult mutant GISTs and express SDHB, while adult wild-type gastric GISTs are dispersed amongst adult mutant and pediatric wild-type cases, clustering in this situation on the basis of SDHB expression. Interestingly, global methylation analysis has recently similarly demonstrated that these wild-type, SDHB-immunonegative tumors show a distinct pattern compared with KIT and PDGFRA mutant tumors, which as a rule do express SDHB. All cases with Carney triad within our cohort cluster together tightly.
Transforming growth factor-β (TGF-β) signaling regulates many diverse cellular activities through both canonical (SMAD-dependent) and non-canonical branches, which includes the mitogen-activated protein kinase (MAPK), Rho-like guanosine triphosphatase and phosphatidylinositol-3-kinase/AKT pathways. Here, we demonstrate that miR-335 directly targets and downregulates genes in the TGF-β non-canonical pathways, including the Rho-associated coiled-coil containing protein (ROCK1) and MAPK1, resulting in reduced phosphorylation of downstream pathway members. Specifically, inhibition of ROCK1 and MAPK1 reduces phosphorylation levels of the motor protein myosin light chain (MLC) leading to a significant inhibition of the invasive and migratory potential of neuroblastoma cells. Additionally, miR-335 targets the leucine-rich alpha-2-glycoprotein 1 (LRG1) messenger RNA, which similarly results in a significant reduction in the phosphorylation status of MLC and a decrease in neuroblastoma cell migration and invasion. Thus, we link LRG1 to the migratory machinery of the cell, altering its activity presumably by exerting its effect within the non-canonical TGF-β pathway. Moreover, we demonstrate that the MYCN transcription factor, whose coding sequence is highly amplified in a particularly clinically aggressive neuroblastoma tumor subtype, directly binds to a region immediately upstream of the miR-335 transcriptional start site, resulting in transcriptional repression. We conclude that MYCN contributes to neuroblastoma cell migration and invasion, by directly downregulating miR-335, resulting in the upregulation of the TGF-β signaling pathway members ROCK1, MAPK1 and putative member LRG1, which positively promote this process. Our results provide novel insight into the direct regulation of TGF-β non-canonical signaling by miR-335, which in turn is downregulated by MYCN.
Ultra-conserved regions (UCRs) are segments of the genome (≥ 200 bp) that exhibit 100% DNA sequence conservation between human, mouse and rat. Transcribed UCRs (T-UCRs) have been shown to be differentially expressed in cancers versus normal tissue, indicating a possible role in carcinogenesis. All-trans-retinoic acid (ATRA) causes some neuroblastoma (NB) cell lines to undergo differentiation and leads to a significant decrease in the oncogenic transcription factor MYCN. Here, we examine the impact of ATRA treatment on T-UCR expression and investigate the biological significance of these changes.
We designed a custom tiling microarray to profile the expression of 481 T-UCRs in sense and anti-sense orientation (962 potential transcripts) in untreated and ATRA-treated neuroblastoma cell lines (SH-SY5Y, SK-N-BE, LAN-5). Following identification of significantly differentially expressed T-UCRs, we carried out siRNA knockdown and gene expression microarray analysis to investigate putative functional roles for selected T-UCRs.
Following ATRA-induced differentiation, 32 T-UCRs were differentially expressed (16 up-regulated, 16 down-regulated) across all three cell lines. Further insight into the possible role of T-UC.300A, an independent transcript whose expression is down-regulated following ATRA was achieved by siRNA knockdown, resulting in the decreased viability and invasiveness of ATRA-responsive cell lines. Gene expression microarray analysis following knockdown of T-UC.300A revealed a number of genes whose expression was altered by changing T-UC.300A levels and that might play a role in the increased proliferation and invasion of NB cells prior to ATRA-treatment.
Our results indicate that significant numbers of T-UCRs have altered expression levels in response to ATRA. While the precise roles that T-UCRs might play in cancer or in normal development are largely unknown and an important area for future study, our findings strongly indicate that the function of non-coding RNA T-UC.300A is connected with proliferation, invasion and the inhibition of differentiation of neuroblastoma cell lines prior to ATRA treatment.
ATRA; neuroblastoma; Transcribed ultra-conserved regions; Differentiation
The current SIOP treatment protocol for Wilms’ tumor involves pre-operative chemotherapy followed by nephrectomy. Not all patients benefit equally from such chemotherapy. The aim of this study was to generate a miRNA profile of chemo resistant blastemal cells in high risk Wilms’ tumors which might serve as predictive markers of therapeutic response at the pre-treatment biopsy stage. We have shown here that unsupervised hierarchical clustering of genome-wide miRNA expression profiles can clearly separate intermediate risk tumors from high risk tumors. A total of 29 miRNAs were significantly differentially expressed between post-treatment intermediate risk and high risk groups, including miRNAs that have been previously linked to chemo resistance in other cancer types. Furthermore, 7 of these 29 miRNAs were already at the pre-treatment biopsy stage differentially expressed between cases ultimately deemed intermediate risk compared to high risk. These miRNA alterations include down-regulation in high risk cases of miR-193a.5p, miR-27a and the up-regulation of miR-483.5p, miR-628.5p, miR-590.5p, miR-302a and miR-367. The demonstration of such miRNA markers at the pre-treatment biopsy stage could permit stratification of patients to more tailored treatment regimens.
Several studies have implicated the dysregulation of microRNAs in neuroblastoma pathogenesis, an often fatal paediatric cancer arising from precursor cells of the sympathetic nervous system. Our group and others have demonstrated that lower expression of miR-542-5p is highly associated with poor patient survival, indicating a potential tumor suppressive function. Here, we demonstrate that ectopic over-expression of this miRNA decreases the invasive potential of neuroblastoma cell lines in vitro, along with primary tumor growth and metastases in an orthotopic mouse xenograft model, providing the first functional evidence for the involvement of miR-542-5p as a tumor suppressor in any type of cancer.
MicroRNAs; neuroblastoma; miR-542-5p; orthotopic mouse model
MicroRNAs function as negative regulators of post-transcriptional gene expression, playing major roles in cellular differentiation. Several neuroblastoma cell lines can be induced to undergo differentiation by all-trans-retinoic acid (ATRA) and are used for modelling signalling pathways involved in this process. To identify miRNAs contributing to differentiation, we profiled 364 loci following ATRA treatment of neuroblastoma cell lines and found miR-10a and miR-10b to be highly over-expressed in SK-N-BE, LAN5, and SHSY-5Y. Ectopic over-expression of these miRNAs led to a major reprogramming of the transcriptome and a differentiated phenotype that was similar to that induced by ATRA in each of these cell lines. One of the predicted down-regulated miR-10a/b targets was nuclear receptor co-repressor 2 (NCOR2), a co-repressor of gene transcription which is known to suppress neurite outgrowth. NCOR2 was experimentally validated as a direct target of miR-10a/b, and siRNA mediated inhibition of this mRNA alone resulted in neural cell differentiation. Moreover, induction of differentiation could be blocked by ectopic up-regulation of NCOR2 using an expression construct lacking the miR-10a/b 3’ UTR target site. We conclude that miR-10a/b play major roles in the process of neural cell differentiation through direct targeting of NCOR2, which in turn induces a cascade of primary and secondary transcriptional alterations, including the down-regulation of MYCN.
ATRA; MYCN; neuroblastoma; miR-10a; miR-10b; NCOR2; differentiation
MicroRNAs are small molecules which regulate gene expression post-transcriptionally and aberrant expression of several miRNAs is associated with neuroblastoma, a childhood cancer arising from precursor cells of the sympathetic nervous system. Amplification of the MYCN transcription factor characterizes the most clinically aggressive subtype of this disease, and although alteration of p53 signaling is not commonly found in primary tumors, deregulation of proteins involved in this pathway frequently arise in recurrent disease after pharmacological treatment. TH-MYCN is a well-characterized transgenic model of MYCN-driven neuroblastoma which recapitulates many clinicopathologic features of the human disease. Here, we evaluate the dysregulation of miRNAs in tumors from TH-MYCN mice that are either wild-type (TH-MYCN) or deficient (TH-MYCN/p53ERTAM) for the p53 tumor suppressor gene.
We analyzed the expression of 591 miRNAs in control (adrenal) and neuroblastoma tumor tissues derived from either TH-MYCN or TH-MYCN/p53ERTAM mice, respectively wild-type or deficient in p53. Comparing miRNA expression in tumor and control samples, we identified 159 differentially expressed miRNAs. Using data previously obtained from human neuroblastoma samples, we performed a comparison of miRNA expression between murine and human tumors to assess the concordance between murine and human expression data. Notably, the miR-17-5p-92 oncogenic polycistronic cluster, which is over-expressed in human MYCN amplified tumors, was over-expressed in mouse tumors. Moreover, analyzing miRNAs expression in a mouse model (TH-MYCN/p53ERTAM) possessing a transgenic p53 allele that drives the expression of an inactive protein, we identified miR-125b-3p and miR-676 as directly or indirectly regulated by the level of functional p53.
Our study represents the first miRNA profiling of an important mouse model of neuroblastoma. Similarities and differences in miRNAs expression between human and murine neuroblastoma were identified, providing important insight into the efficacy of this mouse model for assessing miRNA involvement in neuroblastoma and their potential effectiveness as therapeutic targets.
Neuroblastoma is an often fatal pediatric cancer arising from precursor cells of the sympathetic nervous system. 13-Cis retinoic acid is included in the treatment regime for patients with high-risk disease, and a similar derivative, all-trans retinoic acid (ATRA) causes neuroblastoma cell lines to undergo differentiation. The molecular signaling pathways involved with ATRA induced differentiation are complex, and the role that DNA methylation changes might play are unknown. The purpose of this study was to evaluate the genome-wide effects of ATRA on DNA methylation using methylated DNA immunoprecipitation applied to microarrays representing all known promoter and CpG islands. 402 gene promoters became demethylated, while 88 were hypermethylated post-ATRA. mRNA expression microarrays revealed that 82 of the demethylated genes were over-expressed by >2 fold, while 13 of the hyper methylated genes were under-expressed. Gene ontology analysis indicated that de-methylated and re-expressed genes were enriched for signal transduction pathways, including NOS1, which is required for neural cell differentiation. As a potential mechanism for the DNA methylation changes, we demonstrate the down-regulation of methyltransferases, DNMT1 and DNMT3B, along with the up-regulation of endogenous microRNAs targeting them. Ectopic over-expression of miR-152, targeting DNMT1, also negatively impacted cell invasiveness and anchorage independent growth, contributing in part to the differentiated phenotype. We conclude that functionally important, miRNA-mediated DNA de-methylation changes contribute to the process of ATRA induced differentiation resulting in the activation of NOS1, a critical determinant of neural cell differentiation. Our findings illustrate the plasticity and dynamic nature of the epigenome during cancer cell differentiation.
DNA Hypermethylation; MYCN; ATRA; Neuroblastoma; miRNA
MYCN is a transcription factor that is expressed during the development of the neural crest and its dysregulation plays a major role in the pathogenesis of pediatric cancers such as neuroblastoma, medulloblastoma and rhabdomyosarcoma. MeCP2 is a CpG methyl binding protein which has been associated with a number of cancers and developmental disorders, particularly Rett syndrome.
Methods and Findings
Using an integrative global genomics approach involving chromatin immunoprecipitation applied to microarrays, we have determined that MYCN and MeCP2 co-localize to gene promoter regions, as well as inter/intragenic sites, within the neuroblastoma genome (MYCN amplified Kelly cells) at high frequency (70.2% of MYCN sites were also positive for MeCP2). Intriguingly, the frequency of co-localization was significantly less at promoter regions exhibiting substantial hypermethylation (8.7%), as determined by methylated DNA immunoprecipitation (MeDIP) applied to the same microarrays. Co-immunoprecipitation of MYCN using an anti-MeCP2 antibody indicated that a MYCN/MeCP2 interaction occurs at protein level. mRNA expression profiling revealed that the median expression of genes with promoters bound by MYCN was significantly higher than for genes bound by MeCP2, and that genes bound by both proteins had intermediate expression. Pathway analysis was carried out for genes bound by MYCN, MeCP2 or MYCN/MeCP2, revealing higher order functions.
Our results indicate that MYCN and MeCP2 protein interact and co-localize to similar genomic sites at very high frequency, and that the patterns of binding of these proteins can be associated with significant differences in transcriptional activity. Although it is not yet known if this interaction contributes to neuroblastoma disease pathogenesis, it is intriguing that the interaction occurs at the promoter regions of several genes important for the development of neuroblastoma, including ALK, AURKA and BDNF.
The purpose of this study was to further define the biology of the 11q− neuroblastoma tumor subgroup by the integration of aCGH with miRNA expression profiling data to determine if improved patient stratification is possible.
A set of primary neuroblastoma (n=160) which was broadly representative of all genetic subtypes was analyzed by aCGH and for the expression of 430 miRNAs. A 15 miRNA expression signature previously demonstrated to be predictive of clinical outcome was used to analyze an independent cohort of 11q− tumors (n=37).
Loss of 4p and gain of 7q occurred at a significantly higher frequency in the 11q−tumors, further defining the genetic characteristics of this subtype. The 11q− tumors could be split into two subgroups using a miRNA expression survival signature which differed significantly in both clinical outcome and the overall frequency of large scale genomic imbalances, with the poor survival subgroup having significantly more imbalances. MiRNAs from the expression signature which were up-regulated in unfavorable tumors were predicted to target down-regulated genes from a published mRNA expression classifier of clinical outcome at a higher than expected frequency, indicating the miRNAs might contribute to the regulation of genes within the signature.
We demonstrate that two distinct biological subtypes of neuroblastoma with loss of 11q occur which differ in their miRNA expression profiles, frequency of segmental imbalances and clinical outcome. A miRNA expression signature, combined with an analysis of segmental imbalances, provides greater prediction of EFS and OS outcomes than 11q status by itself, improving patient stratification.
aCGH; MYCN; neuroblastoma; miRNA
Neuroblastoma is a paediatric cancer of the sympathetic nervous system. The single most important genetic indicator of poor clinical outcome is amplification of the MYCN transcription factor. One of many down-stream MYCN targets is miR-184, which is either directly or indirectly repressed by this transcription factor, possibly due to its pro-apoptotic effects when ectopically over-expressed in neuroblastoma cells. The purpose of this study was to elucidate the molecular mechanism by which miR-184 conveys pro-apoptotic effects.
We demonstrate that the knock-down of endogenous miR-184 has the opposite effect of ectopic up-regulation, leading to enhanced neuroblastoma cell numbers. As a mechanism of how miR-184 causes apoptosis when over-expressed, and increased cell numbers when inhibited, we demonstrate direct targeting and degradation of AKT2, a major downstream effector of the phosphatidylinositol 3-kinase (PI3K) pathway, one of the most potent pro-survival pathways in cancer. The pro-apoptotic effects of miR-184 ectopic over-expression in neuroblastoma cell lines is reproduced by siRNA inhibition of AKT2, while a positive effect on cell numbers similar to that obtained by the knock-down of endogenous miR-184 can be achieved by ectopic up-regulation of AKT2. Moreover, co-transfection of miR-184 with an AKT2 expression vector lacking the miR-184 target site in the 3'UTR rescues cells from the pro-apoptotic effects of miR-184.
MYCN contributes to tumorigenesis, in part, by repressing miR-184, leading to increased levels of AKT2, a direct target of miR-184. Thus, two important genes with positive effects on cell growth and survival, MYCN and AKT2, can be linked into a common genetic pathway through the actions of miR-184. As an inhibitor of AKT2, miR-184 could be of potential benefit in miRNA mediated therapeutics of MYCN amplified neuroblastoma and other forms of cancer.
Neuroblastoma, a cancer derived from precursor cells of the sympathetic nervous system, is a major cause of childhood cancer related deaths. The single most important prognostic indicator of poor clinical outcome in this disease is genomic amplification of MYCN, a member of a family of oncogenic transcription factors.
We applied MYCN chromatin immunoprecipitation to microarrays (ChIP-chip) using MYCN amplified/non-amplified cell lines as well as a conditional knockdown cell line to determine the distribution of MYCN binding sites within all annotated promoter regions.
Assessment of E-box usage within consistently positive MYCN binding sites revealed a predominance for the CATGTG motif (p<0.0016), with significant enrichment of additional motifs CATTTG, CATCTG, CAACTG in the MYCN amplified state. For cell lines over-expressing MYCN, gene ontology analysis revealed enrichment for the binding of MYCN at promoter regions of numerous molecular functional groups including DNA helicases and mRNA transcriptional regulation. In order to evaluate MYCN binding with respect to other genomic features, we determined the methylation status of all annotated CpG islands and promoter sequences using methylated DNA immunoprecipitation (MeDIP). The integration of MYCN ChIP-chip and MeDIP data revealed a highly significant positive correlation between MYCN binding and DNA hypermethylation. This association was also detected in regions of hemizygous loss, indicating that the observed association occurs on the same homologue. In summary, these findings suggest that MYCN binding occurs more commonly at CATGTG as opposed to the classic CACGTG E-box motif, and that disease associated over expression of MYCN leads to aberrant binding to additional weaker affinity E-box motifs in neuroblastoma. The co-localization of MYCN binding and DNA hypermethylation further supports the dual role of MYCN, namely that of a classical transcription factor affecting the activity of individual genes, and that of a mediator of global chromatin structure.
MiRNAs regulate gene expression at a post-transcriptional level and their dysregulation can play major roles in the pathogenesis of many different forms of cancer, including neuroblastoma, an often fatal paediatric cancer originating from precursor cells of the sympathetic nervous system. We have analyzed a set of neuroblastoma (n = 145) that is broadly representative of the genetic subtypes of this disease for miRNA expression (430 loci by stem-loop RT qPCR) and for DNA copy number alterations (array CGH) to assess miRNA involvement in disease pathogenesis. The tumors were stratified and then randomly split into a training set (n = 96) and a validation set (n = 49) for data analysis. Thirty-seven miRNAs were significantly over- or under-expressed in MYCN amplified tumors relative to MYCN single copy tumors, indicating a potential role for the MYCN transcription factor in either the direct or indirect dysregulation of these loci. In addition, we also determined that there was a highly significant correlation between miRNA expression levels and DNA copy number, indicating a role for large-scale genomic imbalances in the dysregulation of miRNA expression. In order to directly assess whether miRNA expression was predictive of clinical outcome, we used the Random Forest classifier to identify miRNAs that were most significantly associated with poor overall patient survival and developed a 15 miRNA signature that was predictive of overall survival with 72.7% sensitivity and 86.5% specificity in the validation set of tumors. We conclude that there is widespread dysregulation of miRNA expression in neuroblastoma tumors caused by both over-expression of the MYCN transcription factor and by large-scale chromosomal imbalances. MiRNA expression patterns are also predicative of clinical outcome, highlighting the potential for miRNA mediated diagnostics and therapeutics.
Metabolomics, or metabonomics, refers to the quantitative analysis of all metabolites present within a biological sample and is generally carried out using NMR spectroscopy or Mass Spectrometry. Such analysis produces a set of peaks, or features, indicative of the metabolic composition of the sample and may be used as a basis for sample classification. Feature selection may be employed to improve classification accuracy or aid model explanation by establishing a subset of class discriminating features. Factors such as experimental noise, choice of technique and threshold selection may adversely affect the set of selected features retrieved. Furthermore, the high dimensionality and multi-collinearity inherent within metabolomics data may exacerbate discrepancies between the set of features retrieved and those required to provide a complete explanation of metabolite signatures. Given these issues, the latter in particular, we present the MetaFIND application for 'post-feature selection' correlation analysis of metabolomics data.
In our evaluation we show how MetaFIND may be used to elucidate metabolite signatures from the set of features selected by diverse techniques over two metabolomics datasets. Importantly, we also show how MetaFIND may augment standard feature selection and aid the discovery of additional significant features, including those which represent novel class discriminating metabolites. MetaFIND also supports the discovery of higher level metabolite correlations.
Standard feature selection techniques may fail to capture the full set of relevant features in the case of high dimensional, multi-collinear metabolomics data. We show that the MetaFIND 'post-feature selection' analysis tool may aid metabolite signature elucidation, feature discovery and inference of metabolic correlations.
Microarrays have the capacity to measure the expressions of thousands of genes in parallel over many experimental samples. The unsupervised classification technique of bicluster analysis has been employed previously to uncover gene expression correlations over subsets of samples with the aim of providing a more accurate model of the natural gene functional classes. This approach also has the potential to aid functional annotation of unclassified open reading frames (ORFs). Until now this aspect of biclustering has been under-explored. In this work we illustrate how bicluster analysis may be extended into a 'semi-supervised' ORF annotation approach referred to as BALBOA.
The efficacy of the BALBOA ORF classification technique is first assessed via cross validation and compared to a multi-class k-Nearest Neighbour (kNN) benchmark across three independent gene expression datasets. BALBOA is then used to assign putative functional annotations to unclassified yeast ORFs. These predictions are evaluated using existing experimental and protein sequence information. Lastly, we employ a related semi-supervised method to predict the presence of novel functional modules within yeast.
In this paper we demonstrate how unsupervised classification methods, such as bicluster analysis, may be extended using of available annotations to form semi-supervised approaches within the gene expression analysis domain. We show that such methods have the potential to improve upon supervised approaches and shed new light on the functions of unclassified ORFs and their co-regulation.