Multiple Microarray Datasets Research Articles

A novel strategy of integrated microarray analysis identifies CENPA, CDK1 and CDC20 as a cluster of diagnostic biomarkers in lung adenocarcinoma

Lung adenocarcinoma (LAC) is the most lethal cancer and the leading cause of cancer-related death worldwide. The identification of meaningful clusters of co-expressed genes or representative biomarkers may help improve the accuracy of LAC diagnoses. Public databases, such as the Gene Expression Omnibus (GEO), provide rich resources of valuable information for clinics, however, the integration of multiple microarray datasets from various platforms and institutes remained a challenge. To determine potential indicators of LAC, we performed genome-wide relative significance (GWRS), genome-wide global significance (GWGS) and support vector machine (SVM) analyses progressively to identify robust gene biomarker signatures from 5 different microarray datasets that included 330 samples. The top 200 genes with robust signatures were selected for integrative analysis according to “guilt-by-association” methods, including protein-protein interaction (PPI) analysis and gene co-expression analysis. Of these 200 genes, only 10 genes showed both intensive PPI network and high gene co-expression correlation (r > 0.8). IPA analysis of this regulatory networks suggested that the cell cycle process is a crucial determinant of LAC. CENPA, as well as two linked hub genes CDK1 and CDC20, are determined to be potential indicators of LAC. Immunohistochemical staining showed that CENPA, CDK1 and CDC20 were highly expressed in LAC cancer tissue with co-expression patterns. A Cox regression model indicated that LAC patients with CENPA+/CDK1+ and CENPA+/CDC20+ were high-risk groups in terms of overall survival. In conclusion, our integrated microarray analysis demonstrated that CENPA, CDK1 and CDC20 might serve as novel cluster of prognostic biomarkers for LAC, and the cooperative unit of three genes provides a technically simple approach for identification of LAC patients.

Identification of potential transcriptomic markers in developing asthma: An integrative analysis of gene expression profiles

The goal of this study was to identify potential transcriptomic markers in developing asthma by an integrative analysis of multiple public microarray data sets. Using the R software and bioconductor packages, we performed a statistical analysis to identify differentially expressed (DE) genes in asthma, and further performed functional interpretation (enrichment analysis and co-expression network construction) and classification quality evaluation of the DE genes identified. 3 microarray datasets (192 cases and 91 controls in total) were collected for this analysis. 62 DE genes were identified in asthma, among which 43 genes were up-regulated and 19 genes were down-regulated. The up-regulated gene with the highest Log2 Fold Change (LFC) was CLCA1 (LFC=2.81). The down-regulated gene with the highest absolute LFC was BPIFA1 (LFC=−1.45). Enrichment analysis revealed that those DE genes strongly associated with proteolysis, retina homeostasis, humoral immune response, and salivary secretion. A support vector machine classifier (asthma versus healthy control) was also trained based on DE genes. In conclusion, the consistently DE genes identified in this study are suggested as candidate transcriptomic markers for asthma diagnosis, and provide novel insights into the pathogenesis of asthma.

RiceMetaSys for salt and drought stress responsive genes in rice: a web interface for crop improvement

BackgroundGenome-wide microarray has enabled development of robust databases for functional genomics studies in rice. However, such databases do not directly cater to the needs of breeders. Here, we have attempted to develop a web interface which combines the information from functional genomic studies across different genetic backgrounds with DNA markers so that they can be readily deployed in crop improvement. In the current version of the database, we have included drought and salinity stress studies since these two are the major abiotic stresses in rice.ResultsRiceMetaSys, a user-friendly and freely available web interface provides comprehensive information on salt responsive genes (SRGs) and drought responsive genes (DRGs) across genotypes, crop development stages and tissues, identified from multiple microarray datasets. ‘Physical position search’ is an attractive tool for those using QTL based approach for dissecting tolerance to salt and drought stress since it can provide the list of SRGs and DRGs in any physical interval. To identify robust candidate genes for use in crop improvement, the ‘common genes across varieties’ search tool is useful. Graphical visualization of expression profiles across genes and rice genotypes has been enabled to facilitate the user and to make the comparisons more impactful. Simple Sequence Repeat (SSR) search in the SRGs and DRGs is a valuable tool for fine mapping and marker assisted selection since it provides primers for survey of polymorphism. An external link to intron specific markers is also provided for this purpose. Bulk retrieval of data without any limit has been enabled in case of locus and SSR search.ConclusionsThe aim of this database is to facilitate users with a simple and straight-forward search options for identification of robust candidate genes from among thousands of SRGs and DRGs so as to facilitate linking variation in expression profiles to variation in phenotype.Database URL: http://14.139.229.201

Identification of TGFβ-related genes regulated in murine osteoarthritis and chondrocyte hypertrophy by comparing multiple microarray datasets

Identification of TGFβ-related genes regulated in murine osteoarthritis and chondrocyte hypertrophy by comparing multiple microarray datasets

Gene network analysis reveals a novel 22-gene signature of carbon metabolism in hepatocellular carcinoma.

Although much progress has been made in understanding cancer cellular metabolism adaptation, the co-regulations between genes of metabolism and cancer pathways and their interactions remain poorly characterized. Here, we applied gene co-expression network analysis to 1509 metabolic gene expression data generated from 120 HCC and 180 non-tumor human liver tissues by microarray. Our analyses reveal that metabolism genes can be classified into different co-expression modules based on their associations with HCC related traits. The co-regulation mechanism of the carbon metabolism genes in normal liver tissues was interrupted during the processes of carcinogenesis. In parallel, we performed RNAseq analysis of HCC and non-tumor human liver tissues, and identified a unique 22-carbon-metabolism-gene-signature of increased expression. This gene signature was further verified in multiple microarray data sets, and its prognostic value was also proven by HCC patients' survival data from TCGA. Additionally, the tumorigenic function of two representative genes, CS and ACSS1, were validated experimentally by cell growth and spheroid formation assays. The current study provides evidence for the reprogramming of the co-regulation network between carbon metabolism and cancer pathway genes in HCC. In addition, this study also reveals a unique 22-carbon-metabolism-gene-expression-signature in HCC. Strategies targeting these genes may represent new therapeutic approaches for HCC treatment.

Meta-Analysis of EMT Datasets Reveals Different Types of EMT.

As a critical process during embryonic development, cancer progression and cell fate conversions, epithelial-mesenchymal transition (EMT) has been extensively studied over the last several decades. To further understand the nature of EMT, we performed meta-analysis of multiple microarray datasets to identify the related generic signature. In this study, 24 human and 17 mouse microarray datasets were integrated to identify conserved gene expression changes in different types of EMT. Our integrative analysis revealed that there is low agreement among the list of the identified signature genes and three other lists in previous studies. Since removing the datasets with weakly-induced EMT from the analysis did not significantly improve the overlapping in the signature-gene lists, we hypothesized the existence of different types of EMT. This hypothesis was further supported by the grouping of 74 human EMT-induction samples into five distinct clusters, and the identification of distinct pathways in these different clusters of EMT samples. The five clusters of EMT-induction samples also improves the understanding of the characteristics of different EMT types. Therefore, we concluded the existence of different types of EMT was the possible reason for its complex role in multiple biological processes.

MiR-539 inhibits prostate cancer progression by directly targeting SPAG5.

BackgroundWe conducted multiple microarray datasets analyses from clinical and xenograft tumor tissues to search for disease progression-driving oncogenes in prostate cancer (PCa). Sperm-associated antigen 5 (SPAG5) attracted our attention. SPAG5 was recently identified as an oncogene participating in lung cancer and cervical cancer progression. However, the roles of SPAG5 in PCa progression remain unknown.MethodsSPAG5 expression level in clinical primary PCa, metastatic PCa, castration resistant PCa, neuroendocrine PCa, and normal prostate tissues was investigated. We established multiple in vivo xenografts models using patient-derived tissues and investigated SPAG5 expression trend in these models. We also investigated the functions of SPAG5 in vivo and in vitro studies. Luciferase reporter assays were performed to investigate potential miRNAs that can regulate SPAG5.ResultsWe identified that SPAG5 expression was gradually increased in PCa progression and its level was significantly associated with lymph node metastasis, clinical stage, Gleason score, and biochemical recurrence. Our results indicated that SPAG5 knockdown can drastically inhibit PCa cell proliferation, migration, and invasion in vitro and supress tumor growth and metastasis in vivo. We identified that miR-539 can directly target SPAG5. Ectopic overexpression of miR-539 can drastically inhibit SPAG5 expression and the restoration of SPAG5 expression can reverse the inhibitory effects of miR-539 on PCa cell proliferation and metastasis.ConclusionOur results collectively showed a progression-driving role of SPAG5 in PCa which can be regulated by miR-539, suggesting that miR-539/SPAG5 can serve as a potential therapeutic target for PCa.Electronic supplementary materialThe online version of this article (doi:10.1186/s13046-016-0337-8) contains supplementary material, which is available to authorized users.

Handling batch effects on cross-platform classification of microarray data

Gene-set-based microarray analysis is commonly applied in the classification of complex diseases. However, the robustness of a classifier is normally limited by the small number of samples in many microarray datasets. Although a merged dataset from multiple experiments may improve classification performance, batch effects or technical/biological variations among these experiments may eventually confound the analysis. Besides the batch effects, merging multiple microarray datasets from different platforms can generate missing values, due to a different number of covered genes. In this work, we extend previous works that focused on the missing value incident by further exploring the impact of batch effects on cross-platform classification. Two quality measures of data purity are proposed and two data imputation methods are compared. The results show that by doing batch correction the quality of the merged data is improved significantly. Furthermore, the classification performance is high when the normalised purity is above a certain threshold.

Integration of multiple microarray datasets to generate a robust genomic signature to predict recurrence in oral squamous cell carcinoma (OSCC)

Integration of multiple microarray datasets to generate a robust genomic signature to predict recurrence in oral squamous cell carcinoma (OSCC)

FoxM1 transactivates PTTG1 and promotes colorectal cancer cell migration and invasion.

BackgroundMetastasis is the major cause of cancer-related death. Forkhead Box M1 (FoxM1) is a master regulator of tumor metastasis. This study aims to identify new FoxM1 targets in regulating tumor metastasis using bioinformatics tools as well as biological experiments.MethodsIllumina microarray was used to profile WT and PTTG1 knockout HCT116 cells. R2 Genomics Analysis was used to identify PTTG1 as a potential FoxM1 targeted gene. Luciferase reporter array, EMSA and Chromatin Immunoprecipitation (ChIP) were used to determine the binding of FoxM1 to PTTG1 promoter. Boyden chamber assay was used to evaluate the effects of FoxM1-PTTG1 on cell migration and invasion. Splenic-injection induced liver metastasis model was used to evaluate the effects of FoxM1-PTTG1 on liver metastasis of colorectal cancer.ResultsAnalyses of multiple microarray datasets derived from human colorectal cancer indicated that correlation levels of FoxM1 and pituitary tumor transforming gene (PTTG1) are highly concordant (R = 0.68 ~ 0.89, p = 2.1E-226 ~ 9.6E-86). FoxM1 over-expression increased and knock-down decreased PTTG1 expression. Luciferase reporter assay identified that the −600 to −300 bp region of PTTG1 promoter is important for FoxM1 to enhance PTTG1 promoter activity. EMSA and ChIP assays confirmed that FoxM1 directly binds to PTTG1 promoter at the −391 to −385 bp region in colorectal cancer cells. Boyden chamber assay indicated that both FoxM1 and PTTG1 regulate migration and invasion of HCT116 and SW620 colorectal cancer cells. Further in vivo assays indicated that PTTG1 knock out decreased the liver metastasis of FoxM1 over-expressing HCT116 cells. Microarray analyses identified 662 genes (FDR < 0.05) differentially expressed between WT and PTTG1−/− HCT116 cells. Among them, dickkopf homolog 1 (DKK1), a known WNT pathway inhibitor, was suppressed by PTTG1 and FoxM1.ConclusionsPTTG1 is a FoxM1 targeted gene. FoxM1 binds to PTTG1 promoter to enhance PTTG1 transcription, and FoxM1-PTTG1 pathway promotes colorectal cancer migration and invasion.Electronic supplementary materialThe online version of this article (doi:10.1186/s12920-015-0126-9) contains supplementary material, which is available to authorized users.

Individualized identification of disease-associated pathways with disrupted coordination of gene expression.

Current pathway analysis approaches are primarily dedicated to capturing deregulated pathways at the population level and cannot provide patient-specific pathway deregulation information. In this article, the authors present a simple approach, called individPath, to detect pathways with significantly disrupted intra-pathway relative expression orderings for each disease sample compared with the stable, normal intra-pathway relative expression orderings pre-determined in previously accumulated normal samples. Through the analysis of multiple microarray data sets for lung and breast cancer, the authors demonstrate individPath's effectiveness for detecting cancer-associated pathways with disrupted relative expression orderings at the individual level and dissecting the heterogeneity of pathway deregulation among different patients. The portable use of this simple approach in clinical contexts is exemplified by the identification of prognostic intra-pathway gene pair signatures to predict overall survival of resected early-stage lung adenocarcinoma patients and signatures to predict relapse-free survival of estrogen receptor-positive breast cancer patients after tamoxifen treatment.

Identification of potential transcriptomic markers in developing ankylosing spondylitis: a meta-analysis of gene expression profiles.

The goal of this study was to identify potential transcriptomic markers in developing ankylosing spondylitis by a meta-analysis of multiple public microarray datasets. Using the INMEX (integrative meta-analysis of expression data) program, we performed the meta-analysis to identify consistently differentially expressed (DE) genes in ankylosing spondylitis and further performed functional interpretation (gene ontology analysis and pathway analysis) of the DE genes identified in the meta-analysis. Three microarray datasets (26 cases and 29 controls in total) were collected for meta-analysis. 905 consistently DE genes were identified in ankylosing spondylitis, among which 482 genes were upregulated and 423 genes were downregulated. The upregulated gene with the smallest combined rank product (RP) was GNG11 (combined RP = 299.64). The downregulated gene with the smallest combined RP was S100P (combined RP = 335.94). In the gene ontology (GO) analysis, the most significantly enriched GO term was “immune system process” (P = 3.46 × 10−26). The most significant pathway identified in the pathway analysis was antigen processing and presentation (P = 8.40 × 10−5). The consistently DE genes in ankylosing spondylitis and biological pathways associated with those DE genes identified provide valuable information for studying the pathophysiology of ankylosing spondylitis.

β-Catenin is central to DUX4-driven network rewiring in facioscapulohumeral muscular dystrophy.

Facioscapulohumeral muscular dystrophy (FSHD) is an incurable disease, characterized by skeletal muscle weakness and wasting. Genetically, FSHD is characterized by contraction or hypomethylation of repeat D4Z4 units on chromosome 4, which causes aberrant expression of the transcription factor DUX4 from the last repeat. Many genes have been implicated in FSHD pathophysiology, but an integrated molecular model is currently lacking. We developed a novel differential network methodology, Interactome Sparsification and Rewiring (InSpiRe), which detects network rewiring between phenotypes by integrating gene expression data with known protein interactions. Using InSpiRe, we performed a meta-analysis of multiple microarray datasets from FSHD muscle biopsies, then removed secondary rewiring using non-FSHD datasets, to construct a unified network of rewired interactions. Our analysis identified β-catenin as the main coordinator of FSHD-associated protein interaction signalling, with pathways including canonical Wnt, HIF1-α and TNF-α clearly perturbed. To detect transcriptional changes directly elicited by DUX4, gene expression profiling was performed using microarrays on murine myoblasts. This revealed that DUX4 significantly modified expression of the genes in our FSHD network. Furthermore, we experimentally confirmed that Wnt/β-catenin signalling is affected by DUX4 in murine myoblasts. Thus, we provide the first unified molecular map of FSHD signalling, capable of uncovering pathomechanisms and guiding therapeutic development.

Reliable pre-eclampsia pathways based on multiple independent microarray data sets.

Pre-eclampsia is a multifactorial disorder characterized by heterogeneous clinical manifestations. Gene expression profiling of preeclamptic placenta have provided different and even opposite results, partly due to data compromised by various experimental artefacts. Here we aimed to identify reliable pre-eclampsia-specific pathways using multiple independent microarray data sets. Gene expression data of control and preeclamptic placentas were obtained from Gene Expression Omnibus. Single-sample gene-set enrichment analysis was performed to generate gene-set activation scores of 9707 pathways obtained from the Molecular Signatures Database. Candidate pathways were identified by t-test-based screening using data sets, GSE10588, GSE14722 and GSE25906. Additionally, recursive feature elimination was applied to arrive at a further reduced set of pathways. To assess the validity of the pre-eclampsia pathways, a statistically-validated protocol was executed using five data sets including two independent other validation data sets, GSE30186, GSE44711. Quantitative real-time PCR was performed for genes in a panel of potential pre-eclampsia pathways using placentas of 20 women with normal or severe preeclamptic singleton pregnancies (n = 10, respectively). A panel of ten pathways were found to discriminate women with pre-eclampsia from controls with high accuracy. Among these were pathways not previously associated with pre-eclampsia, such as the GABA receptor pathway, as well as pathways that have already been linked to pre-eclampsia, such as the glutathione and CDKN1C pathways. mRNA expression of GABRA3 (GABA receptor pathway), GCLC and GCLM (glutathione metabolic pathway), and CDKN1C was significantly reduced in the preeclamptic placentas. In conclusion, ten accurate and reliable pre-eclampsia pathways were identified based on multiple independent microarray data sets. A pathway-based classification may be a worthwhile approach to elucidate the pathogenesis of pre-eclampsia.

Abstract 473: TRIM59, a novel prognostic marker for gastric cancer, promotes tumorigenesis via ubiquitination and degradation of the p53 tumor suppressor

Abstract As a key tumor suppressor, p53 plays crucial roles in gastric cancer initiation and progression. However, the underlying mechanisms for the down-regulation of p53 in gastric cancer patients are not fully understood. Several tripartite motif (TRIM) family proteins are identified to be involved in the tumorigenesis of multiple cancer types, but the significance and biological function of a new TRIM family protein, TRIM59, in terms of whether and how it affects p53 signaling in gastric cancer are largely unknown. We analyzed multiple microarray datasets from the Oncomine database and tested the expression level of TRIM59 in gastric tumors vs. their matched normal tissues and in gastric carcinoma cell lines using quantitative-polymerase chain reactions (Q-PCR) and immunohistochemical staining. We found Expression of TRIM59 was significantly up-regulated in gastric cancer and strongly associated with poor prognosis. TRIM59 knockdown led to significantly attenuated gastric cancer cell proliferation, clone formation, migration, and in vivo xenograft tumorigenicity, whereas over-expression of TRIM59 showed the opposite effects. Mechanistically, TRIM59 interacted with p53 and enhanced ubiquitination and degradation of the p53 protein. In addition, elevated expression of TRIM59 correlated with down-regulation of p53 target genes in human gastric tumors. In summary, TRIM59 is a novel prognostic marker for gastric cancer and promotes tumorigenesis through facilitating p53 ubiquitination and degradation. Citation Format: Zhicheng Zhou, You Wang, Jian Li, Hui Cao, Helen He Zhu, Wei-Qiang Gao. TRIM59, a novel prognostic marker for gastric cancer, promotes tumorigenesis via ubiquitination and degradation of the p53 tumor suppressor. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 473. doi:10.1158/1538-7445.AM2014-473

TOP2Ahigh is the phenotype of recurrence and metastasis whereas TOP2Aneg cells represent cancer stem cells in prostate cancer.

Recurrence and metastasis are the main causes of death for prostate cancer patients and cancer stem cells (CSCs) are proposed to play important roles in cancer recurrence and metastasis. It is generally thought that genes upregulated in recurrent/metastatic disease are likely biomarkers of CSCs. Hence we analyzed multiple microarray datasets on prostate tumor tissues to identify upregulated genes associated with cancer recurrence/metastasis, and tried to explore whether those genes were true biomarkers of prostate CSCs. Our results indicated that TOP2A was the most highly upregulated gene in recurrent/metastatic prostate cancer, and its high expression was positively correlated with poor prognosis in patients. Using a promoter reporter system, we unexpectedly discovered enrichment of CSCs in TOP2Aneg cells. Compared to TOP2Ahigh cells, TOP2Aneg cells formed spheres and tumors more efficiently, and became enriched in the presence of stresses. Analysis of cell divisions by time lapse imaging indicated that more slow-cycling cells were observed in TOP2Aneg cells while the proportion of abnormal divisions was higher in TOP2Ahigh cells. Our studies demonstrate that TOP2Ahigh is the phenotype of recurrence/metastasis but TOP2Aneg cells show slow cycling and have CSCs properties in prostate cancer, which has significant implications for prostate cancer therapy.

Integration of multiple microarray datasets generating an eight gene signature as a validated biomarker for the detection of oral squamous cell carcinoma

s / British Journal of Oral and Maxillofacial Surgery 52 (2014) e43–e74 e61 bone from a previous project. Multidirectional mechanical tests were performed on the various designs of titanium cell structures (n=80). The predictability and quality of manufacturing was assessed statistically and also with scanning electron microscope (SEM). The results demonstrated structures matching the mechanical properties of bone and even anisotropy as our results suggest 3 GPa elasticity. This allows the possibility to build regenerating bone with predictable properties. In addition, predictable patterning unlike etching and sandblasting of microscopic (nano) features found to be significant and non homogenous simple repetitive patterns provide better cellular response. The benefit that tissue engineering techniques offer is decreased morbidity, relative independence from donor site, with a highly specific and customized shape. Titanium based reconstruction constructs seems to offer an alternative future for bony reconstruction. http://dx.doi.org/10.1016/j.bjoms.2014.07.059

TRIM59 Is Up-regulated in Gastric Tumors, Promoting Ubiquitination and Degradation of p53

Little is known about factors that promote gastric carcinogenesis. We analyzed multiple microarray data sets for messenger RNAs (mRNAs) that were increased significantly in human gastric tumor samples, compared with the adjacent normal gastric tissue. We foundexpression of tripartite motif 59 (TRIM59), which encodes a putative ubiquitin ligase, to be increased, and investigated itseffects in gastric cancer cell lines. We analyzed microarray data sets from the Oncomine database. We used quantitative polymerase chain reaction and immunoblotting to measure levels of TRIM59 mRNA and protein in50 human gastric cancer and paired normal tissues, obtained from Renji Hospital and the First Affiliated Hospital of Nanchang University, in China. We also measured protein levels in the gastric epithelial cell line GES-1; the cancer celllines MKN45, AGS, SGC7901, BGC823, Snu5, N87, and Snu1; and in tissue arrays of 108 human gastric tumors. TRIM59 was knocked down and overexpressed in gastric cancer cell lines, and the effects on proliferation, clone formation, migration, and growth of xenograft tumors in nude mice were assessed. TRIM59-related signaling pathways were examined by immunoblotting and quantitative polymerase chain reaction. We analyzed interactions among TRIM59, P53, and ubiquitin in immunoprecipitation studies. Levels of TRIM59 mRNA and protein were increased significantly in gastric tumors compared with nontumor tissues; increased levels were associated with advanced tumor stage and shorter patient survival times. TRIM59 knockdown reduced proliferation, clone formation, and migration of gastric cancer cell lines, as well as growth of xenograft tumors in nude mice; overexpression of TRIM59 had the opposite effects. TRIM59 interacted physically with P53, increasing its ubiquitination and degradation. Increased levels of TRIM59 in human gastric tumors correlated withreduced expression of P53 target genes. The putative ubiquitin ligase TRIM59 is up-regulated in human gastric tumors compared with nontumor tissues. Levels of TRIM59 correlate with tumor progression and patient survival times. TRIM59 interacts with P53, promoting its ubiquitination and degradation, and TRIM59 might promote gastric carcinogenesis via this mechanism.

Prognostic markers in lung cancer: is it ready for prime time?

Non-small cell lung cancer (NSCLC) is a heterogeneity disease and to date, specific clinical factors and tumor stage are established as prognostic markers. Nevertheless, prognosis within stage may vary significantly. During the last 3 decades, genes/proteins that drive tumor initiation and progression, such as oncogenes and tumor suppressor genes have been studied as additional potential prognostic markers. The protein markers as evaluated by immunohistochemistry (IHC) have previously dominated these studies. However, with the development of high-throughput techniques to interrogate genome wide genetic or gene expression changes, DNA (copy number and mutation) and RNA (mRNA and microRNA) based markers have more recently been studied as prognostic markers. Largely due to the heterogeneity and complexity of NSCLC, single gene markers including KRAS mutation has not been validated as strong prognostic markers. In contrast, several gene expression signatures representing mRNA levels of multiple genes have been developed and validated in multiple microarray datasets of independent patient cohorts. The salient features of these gene signatures and their potential value to predict benefit from adjuvant chemotherapy is discussed.

Application of the Bi-CoPaM Method to Five Escherichia Coli Datasets Generated under Various Biological Conditions

The increasing amounts of high-throughput biological datasets stimulate the information engineering and machine learning research community to direct more studies towards designing and applying novel methods which are sophisticated and specialised to tackle the problems that are specific in such datasets. The recently proposed binarisation of consensus partition matrices (Bi-CoPaM) method tackles the problem of scrutinising multiple gene expression microarray datasets to identify the subsets of genes which are consistently co-expressed across them. It allows for clustering results which better reflect the biological fact that most of the genes in any cell are expected to be irrelevant to the specific context in hand, as well as the fact that many genes might participate in multiple processes. This has been achieved by clustering the given set of genes while allowing any gene to have any of the three eventualities, to be exclusively assigned to a single cluster, to be simultaneously assigned to multiple clusters, or not to be assigned to any of the clusters. In this study, we expand the scope of application of the Bi-CoPaM method by applying it, for the first time, to bacterial datasets, namely to a set of five Escherichia coli bacterial datasets generated under different biological conditions, in order to identify the subsets of genes which are consistently co-expressed, i.e. well correlated with each other. We identify two clusters with such consistent co-expression, and interestingly, they themselves are consistently negatively correlated with each other. The first cluster is enriched with genes participating in protein synthesis and DNA repair while the second is enriched with transporting genes. Consequently, we draw biological hypotheses that relate some of the genes with currently unknown biological processes to their potential processes. These hypotheses can serve as pilots for focused future gene discovery studies.