IFN-γ Drives Human Monocyte Differentiation into Highly Proinflammatory Macrophages That Resemble a Phenotype Relevant to Psoriasis.

As key cells of the immune system, macrophages coordinate the activation and regulation of the immune response. Macrophages present a complex phenotype that can vary from homeostatic, proinflammatory, and profibrotic to anti-inflammatory phenotypes. The factors that drive the differentiation from monocyte to macrophage largely define the resultant phenotype, as has been shown by the differences found in M-CSF- and GM-CSF-derived macrophages. We explored alternative inflammatory mediators that could be used for in vitro differentiation of human monocytes into macrophages. IFN-γ is a potent inflammatory mediator produced by lymphocytes in disease and infections. We used IFN-γ to differentiate human monocytes into macrophages and characterized the cells at a functional and proteomic level. IFN-γ alone was sufficient to generate macrophages (IFN-γ Mϕ) that were phagocytic and responsive to polarization. We demonstrate that IFN-γ Mϕ are potent activators of T lymphocytes that produce IL-17 and IFN-γ. We identified potential markers (GBP-1, IP-10, IL-12p70, and IL-23) of IFN-γ Mϕ and demonstrate that these markers are enriched in the skin of patients with inflamed psoriasis. Collectively, we show that IFN-γ can drive human monocyte to macrophage differentiation, leading to bona fide macrophages with inflammatory characteristics.

Histone-fold domain protein NF-Y promotes chromatin accessibility for cell type-specific master transcription factors.

Cell type-specific master transcription factors (TFs) play vital roles in defining cell identity and function. However, the roles ubiquitous factors play in the specification of cell identity remain underappreciated. Here we show that the ubiquitous CCAAT-binding NF-Y complex is required for the maintenance of embryonic stem cell (ESC) identity and is an essential component of the core pluripotency network. Genome-wide studies in ESCs and neurons reveal that NF-Y regulates not only genes with housekeeping functions through cell type-invariant promoter-proximal binding, but also genes required for cell identity by binding to cell type-specific enhancers with master TFs. Mechanistically, NF-Y's distinct DNA-binding mode promotes master/pioneer TF binding at enhancers by facilitating a permissive chromatin conformation. Our studies unearth a conceptually unique function for histone-fold domain (HFD) protein NF-Y in promoting chromatin accessibility and suggest that other HFD proteins with analogous structural and DNA-binding properties may function in similar ways.

Assessment of rosacea symptom severity by genome-wide association study and expression analysis highlights immuno-inflammatory and skin pigmentation genes.

Rosacea is a common, chronic skin disease of variable severity with limited treatment options. The cause of rosacea is unknown, but it is believed to be due to a combination of hereditary and environmental factors. Little is known about the genetics of the disease. We performed a genome-wide association study (GWAS) of rosacea symptom severity with data from 73 265 research participants of European ancestry from the 23andMe customer base. Seven loci had variants associated with rosacea at the genome-wide significance level (P < 5 × 10-8). Further analyses highlighted likely gene regions or effector genes including IRF4 (P = 1.5 × 10-17), a human leukocyte antigen (HLA) region flanked by PSMB9 and HLA-DMB (P = 2.2 × 10-15), HERC2-OCA2 (P = 4.2 × 10-12), SLC45A2 (P = 1.7 × 10-10), IL13 (P = 2.8 × 10-9), a region flanked by NRXN3 and DIO2 (P = 4.1 × 10-9), and a region flanked by OVOL1and SNX32 (P = 1.2 × 10-8). All associations with rosacea were novel except for the HLA locus. Two of these loci (HERC-OCA2 and SLC45A2) and another precedented variant (rs1805007 in melanocortin 1 receptor) with an association P value just below the significance threshold (P = 1.3 × 10-7) have been previously associated with skin phenotypes and pigmentation, two of these loci are linked to immuno-inflammation phenotypes (IL13 and PSMB9-HLA-DMA) and one has been associated with both categories (IRF4). Genes within three loci (PSMB9-HLA-DMA, HERC-OCA2 and NRX3-DIO2) were differentially expressed in a previously published clinical rosacea transcriptomics study that compared lesional to non-lesional samples. The identified loci provide specificity of inflammatory mechanisms in rosacea, and identify potential pathways for therapeutic intervention.

Uncovering new disease indications for G-protein coupled receptors and their endogenous ligands.

BACKGROUND: The Open Targets Platform integrates different data sources in order to facilitate identification of potential therapeutic drug targets to treat human diseases. It currently provides evidence for nearly 2.6 million potential target-disease pairs. G-protein coupled receptors are a drug target class of high interest because of the number of successful drugs being developed against them over many years. Here we describe a systematic approach utilizing the Open Targets Platform data to uncover and prioritize potential new disease indications for the G-protein coupled receptors and their ligands. RESULTS: Utilizing the data available in the Open Targets platform, potential G-protein coupled receptor and endogenous ligand disease association pairs were systematically identified. Intriguing examples such as GPR35 for inflammatory bowel disease and CXCR4 for viral infection are used as illustrations of how a systematic approach can aid in the prioritization of interesting drug discovery hypotheses. Combining evidences for G-protein coupled receptors and their corresponding endogenous peptidergic ligands increases confidence and provides supportive evidence for potential new target-disease hypotheses. Comparing such hypotheses to the global pharma drug discovery pipeline to validate the approach showed that more than 93% of G-protein coupled receptor-disease pairs with a high overall Open Targets score involved receptors with an existing drug discovery program. CONCLUSIONS: The Open Targets gene-disease score can be used to prioritize potential G-protein coupled receptors-indication hypotheses. In addition, availability of multiple different evidence types markedly increases confidence as does combining evidence from known receptor-ligand pairs. Comparing the top-ranked hypotheses to the current global pharma pipeline serves validation of our approach and identifies and prioritizes new therapeutic opportunities.

Fip1 regulates mRNA alternative polyadenylation to promote stem cell self-renewal.

mRNA alternative polyadenylation (APA) plays a critical role in post-transcriptional gene control and is highly regulated during development and disease. However, the regulatory mechanisms and functional consequences of APA remain poorly understood. Here, we show that an mRNA 3' processing factor, Fip1, is essential for embryonic stem cell (ESC) self-renewal and somatic cell reprogramming. Fip1 promotes stem cell maintenance, in part, by activating the ESC-specific APA profiles to ensure the optimal expression of a specific set of genes, including critical self-renewal factors. Fip1 expression and the Fip1-dependent APA program change during ESC differentiation and are restored to an ESC-like state during somatic reprogramming. Mechanistically, we provide evidence that the specificity of Fip1-mediated APA regulation depends on multiple factors, including Fip1-RNA interactions and the distance between APA sites. Together, our data highlight the role for post-transcriptional control in stem cell self-renewal, provide mechanistic insight on APA regulation in development, and establish an important function for APA in cell fate specification.

Network-Based Identification and Prioritization of Key Regulators of Coronary Artery Disease Loci.

OBJECTIVE: Recent genome-wide association studies of coronary artery disease (CAD) have revealed 58 genome-wide significant and 148 suggestive genetic loci. However, the molecular mechanisms through which they contribute to CAD and the clinical implications of these findings remain largely unknown. We aim to retrieve gene subnetworks of the 206 CAD loci and identify and prioritize candidate regulators to better understand the biological mechanisms underlying the genetic associations. APPROACH AND RESULTS: We devised a new integrative genomics approach that incorporated (1) candidate genes from the top CAD loci, (2) the complete genetic association results from the 1000 genomes-based CAD genome-wide association studies from the Coronary Artery Disease Genome Wide Replication and Meta-Analysis Plus the Coronary Artery Disease consortium, (3) tissue-specific gene regulatory networks that depict the potential relationship and interactions between genes, and (4) tissue-specific gene expression patterns between CAD patients and controls. The networks and top-ranked regulators according to these data-driven criteria were further queried against literature, experimental evidence, and drug information to evaluate their disease relevance and potential as drug targets. Our analysis uncovered several potential novel regulators of CAD such as LUM and STAT3, which possess properties suitable as drug targets. We also revealed molecular relations and potential mechanisms through which the top CAD loci operate. Furthermore, we found that multiple CAD-relevant biological processes such as extracellular matrix, inflammatory and immune pathways, complement and coagulation cascades, and lipid metabolism interact in the CAD networks. CONCLUSIONS: Our data-driven integrative genomics framework unraveled tissue-specific relations among the candidate genes of the CAD genome-wide association studies loci and prioritized novel network regulatory genes orchestrating biological processes relevant to CAD.

Retinoic acid-related orphan receptor γ (RORγ): a novel participant in the diurnal regulation of hepatic gluconeogenesis and insulin sensitivity.

The hepatic circadian clock plays a key role in the daily regulation of glucose metabolism, but the precise molecular mechanisms that coordinate these two biological processes are not fully understood. In this study, we identify a novel connection between the regulation of RORγ by the clock machinery and the diurnal regulation of glucose metabolic networks. We demonstrate that particularly at daytime, mice deficient in RORγ exhibit improved insulin sensitivity and glucose tolerance due to reduced hepatic gluconeogenesis. This is associated with a reduced peak expression of several glucose metabolic genes critical in the control of gluconeogenesis and glycolysis. Genome-wide cistromic profiling, promoter and mutation analysis support the concept that RORγ regulates the transcription of several glucose metabolic genes directly by binding ROREs in their promoter regulatory region. Similar observations were made in liver-specific RORγ-deficient mice suggesting that the changes in glucose homeostasis were directly related to the loss of hepatic RORγ expression. Altogether, our study shows that RORγ regulates several glucose metabolic genes downstream of the hepatic clock and identifies a novel metabolic function for RORγ in the diurnal regulation of hepatic gluconeogenesis and insulin sensitivity. The inhibition of the activation of several metabolic gene promoters by an RORγ antagonist suggests that antagonists may provide a novel strategy in the management of metabolic diseases, including type 2 diabetes.

Integrative framework for identification of key cell identity genes uncovers determinants of ES cell identity and homeostasis.

Identification of genes associated with specific biological phenotypes is a fundamental step toward understanding the molecular basis underlying development and pathogenesis. Although RNAi-based high-throughput screens are routinely used for this task, false discovery and sensitivity remain a challenge. Here we describe a computational framework for systematic integration of published gene expression data to identify genes defining a phenotype of interest. We applied our approach to rank-order all genes based on their likelihood of determining ES cell (ESC) identity. RNAi-mediated loss-of-function experiments on top-ranked genes unearthed many novel determinants of ESC identity, thus validating the derived gene ranks to serve as a rich and valuable resource for those working to uncover novel ESC regulators. Underscoring the value of our gene ranks, functional studies of our top-hit Nucleolin (Ncl), abundant in stem and cancer cells, revealed Ncl's essential role in the maintenance of ESC homeostasis by shielding against differentiation-inducing redox imbalance-induced oxidative stress. Notably, we report a conceptually novel mechanism involving a Nucleolin-dependent Nanog-p53 bistable switch regulating the homeostatic balance between self-renewal and differentiation in ESCs. Our findings connect the dots on a previously unknown regulatory circuitry involving genes associated with traits in both ESCs and cancer and might have profound implications for understanding cell fate decisions in cancer stem cells. The proposed computational framework, by helping to prioritize and preselect candidate genes for tests using complex and expensive genetic screens, provides a powerful yet inexpensive means for identification of key cell identity genes.

The epigenetic signature of systemic insulin resistance in obese women.

AIMS/HYPOTHESIS: Insulin resistance (IR) links obesity to type 2 diabetes. The aim of this study was to explore whether white adipose tissue (WAT) epigenetic dysregulation is associated with systemic IR by genome-wide CG dinucleotide (CpG) methylation and gene expression profiling in WAT from insulin-resistant and insulin-sensitive women. A secondary aim was to determine whether the DNA methylation signature in peripheral blood mononuclear cells (PBMCs) reflects WAT methylation and, if so, can be used as a marker for systemic IR. METHODS: From 220 obese women, we selected a total of 80 individuals from either of the extreme ends of the distribution curve of HOMA-IR, an indirect measure of systemic insulin sensitivity. Genome-wide transcriptome and DNA CpG methylation profiling by array was performed on subcutaneous (SAT) and visceral (omental) adipose tissue (VAT). CpG methylation in PBMCs was assayed in the same cohort. RESULTS: There were 647 differentially expressed genes (false discovery rate [FDR] 10%) in SAT, all of which displayed directionally consistent associations in VAT. This suggests that IR is associated with dysregulated expression of a common set of genes in SAT and VAT. The average degree of DNA methylation did not differ between the insulin-resistant and insulin-sensitive group in any of the analysed tissues/cells. There were 223 IR-associated genes in SAT containing a total of 336 nominally significant differentially methylated sites (DMS). The 223 IR-associated genes were over-represented in pathways related to integrin cell surface interactions and insulin signalling and included COL5A1, GAB1, IRS2, PFKFB3 and PTPRJ. In VAT there were a total of 51 differentially expressed genes (FDR 10%); 18 IR-associated genes contained a total of 29 DMS. CONCLUSIONS/INTERPRETATION: In individuals discordant for insulin sensitivity, the average DNA CpG methylation in SAT and VAT is similar, although specific genes, particularly in SAT, display significantly altered expression and DMS in IR, possibly indicating that epigenetic regulation of these genes influences metabolism.

Ldb1-nucleated transcription complexes function as primary mediators of global erythroid gene activation.

Erythropoiesis is dependent on the lineage-specific transcription factors Gata1, Tal1, and Klf1. Several erythroid genes have been shown to require all 3 factors for their expression, suggesting that they function synergistically; however, there is little direct evidence for widespread cooperation. Gata1 and Tal1 can assemble within higher-order protein complexes (Ldb1 complexes) that include the adapter molecules Lmo2 and Ldb1. Ldb1 proteins are capable of coassociation, and long-range Ldb1-mediated oligomerization of enhancer- and promoter-bound Ldb1 complexes has been shown to be required for ß-globin gene expression. In this study, we generated a genomewide map of Ldb1 complex binding sites that revealed widespread binding at erythroid genes and at known erythroid enhancer elements. Ldb1 complex binding sites frequently colocalized with Klf1 binding sites and with consensus binding motifs for other erythroid transcription factors. Transcriptomic analysis demonstrated a strong correlation between Ldb1 complex binding and Ldb1 dependency for gene expression and identified a large cohort of genes coregulated by Ldb1 complexes and Klf1. Together, these results provide a foundation for defining the mechanism and scope of Ldb1 complex activity during erythropoiesis.

Diverse stresses dramatically alter genome-wide p53 binding and transactivation landscape in human cancer cells.

The effects of diverse stresses on promoter selectivity and transcription regulation by the tumor suppressor p53 are poorly understood. We have taken a comprehensive approach to characterizing the human p53 network that includes p53 levels, binding, expression and chromatin changes under diverse stresses. Human osteosarcoma U2OS cells treated with anti-cancer drugs Doxorubicin (DXR) or Nutlin-3 (Nutlin) led to strikingly different p53 gene binding patterns based on chromatin immunoprecipitation with high-throughput sequencing experiments. Although two contiguous RRRCWWGYYY decamers is the consensus binding motif, p53 can bind a single decamer and function in vivo. Although the number of sites bound by p53 was six times greater for Nutlin than DXR, expression changes induced by Nutlin were much less dramatic compared with DXR. Unexpectedly, the solvent dimethylsulphoxide (DMSO) alone induced p53 binding to many sites common to DXR; however, this binding had no effect on target gene expression. Together, these data imply a two-stage mechanism for p53 transactivation where p53 binding only constitutes the first stage. Furthermore, both p53 binding and transactivation were associated with increased active histone modification histone H3 lysine 4 trimethylation. We discovered 149 putative new p53 target genes including several that are relevant to tumor suppression, revealing potential new targets for cancer therapy and expanding our understanding of the p53 regulatory network.

Genome-wide signatures of transcription factor activity: connecting transcription factors, disease, and small molecules.

Identifying transcription factors (TF) involved in producing a genome-wide transcriptional profile is an essential step in building mechanistic model that can explain observed gene expression data. We developed a statistical framework for constructing genome-wide signatures of TF activity, and for using such signatures in the analysis of gene expression data produced by complex transcriptional regulatory programs. Our framework integrates ChIP-seq data and appropriately matched gene expression profiles to identify True REGulatory (TREG) TF-gene interactions. It provides genome-wide quantification of the likelihood of regulatory TF-gene interaction that can be used to either identify regulated genes, or as genome-wide signature of TF activity. To effectively use ChIP-seq data, we introduce a novel statistical model that integrates information from all binding "peaks" within 2 Mb window around a gene's transcription start site (TSS), and provides gene-level binding scores and probabilities of regulatory interaction. In the second step we integrate these binding scores and regulatory probabilities with gene expression data to assess the likelihood of True REGulatory (TREG) TF-gene interactions. We demonstrate the advantages of TREG framework in identifying genes regulated by two TFs with widely different distribution of functional binding events (ERα and E2f1). We also show that TREG signatures of TF activity vastly improve our ability to detect involvement of ERα in producing complex diseases-related transcriptional profiles. Through a large study of disease-related transcriptional signatures and transcriptional signatures of drug activity, we demonstrate that increase in statistical power associated with the use of TREG signatures makes the crucial difference in identifying key targets for treatment, and drugs to use for treatment. All methods are implemented in an open-source R package treg. The package also contains all data used in the analysis including 494 TREG binding profiles based on ENCODE ChIP-seq data. The treg package can be downloaded at http://GenomicsPortals.org.

Acute depletion of Tet1-dependent 5-hydroxymethylcytosine levels impairs LIF/Stat3 signaling and results in loss of embryonic stem cell identity.

The TET family of FE(II) and 2-oxoglutarate-dependent enzymes (Tet1/2/3) promote DNA demethylation by converting 5-methylcytosine to 5-hydroxymethylcytosine (5hmC), which they further oxidize into 5-formylcytosine and 5-carboxylcytosine. Tet1 is robustly expressed in mouse embryonic stem cells (mESCs) and has been implicated in mESC maintenance. Here we demonstrate that, unlike genetic deletion, RNAi-mediated depletion of Tet1 in mESCs led to a significant reduction in 5hmC and loss of mESC identity. The differentiation phenotype due to Tet1 depletion positively correlated with the extent of 5hmC loss. Meta-analyses of genomic data sets suggested interaction between Tet1 and leukemia inhibitory factor (LIF) signaling. LIF signaling is known to promote self-renewal and pluripotency in mESCs partly by opposing MAPK/ERK-mediated differentiation. Withdrawal of LIF leads to differentiation of mESCs. We discovered that Tet1 depletion impaired LIF-dependent Stat3-mediated gene activation by affecting Stat3's ability to bind to its target sites on chromatin. Nanog overexpression or inhibition of MAPK/ERK signaling, both known to maintain mESCs in the absence of LIF, rescued Tet1 depletion, further supporting the dependence of LIF/Stat3 signaling on Tet1. These data support the conclusion that analysis of mESCs in the hours/days immediately following efficient Tet1 depletion reveals Tet1's normal physiological role in maintaining the pluripotent state that may be subject to homeostatic compensation in genetic models.

Cnot1, Cnot2, and Cnot3 maintain mouse and human ESC identity and inhibit extraembryonic differentiation.

Embryonic stem cell (ESC) identity and self-renewal is maintained by extrinsic signaling pathways and intrinsic gene regulatory networks. Here, we show that three members of the Ccr4-Not complex, Cnot1, Cnot2, and Cnot3, play critical roles in maintaining mouse and human ESC identity as a protein complex and inhibit differentiation into the extraembryonic lineages. Enriched in the inner cell mass of blastocysts, these Cnot genes are highly expressed in ESC and downregulated during differentiation. In mouse ESCs, Cnot1, Cnot2, and Cnot3 are important for maintenance in both normal conditions and the 2i/LIF medium that supports the ground state pluripotency. Genetic analysis indicated that they do not act through known self-renewal pathways or core transcription factors. Instead, they repress the expression of early trophectoderm (TE) transcription factors such as Cdx2. Importantly, these Cnot genes are also necessary for the maintenance of human ESCs, and silencing them mainly lead to TE and primitive endoderm differentiation. Together, our results indicate that Cnot1, Cnot2, and Cnot3 represent a novel component of the core self-renewal and pluripotency circuitry conserved in mouse and human ESCs.

A Hidradenitis Suppurativa molecular disease signature derived from patient samples by high-throughput RNA sequencing and re-analysis of previously reported transcriptomic data sets.

Hidradenitis suppurativa (HS) is a common, debilitating inflammatory skin disease linked to immune dysregulation and abnormalities in follicular structure and function. Several studies have characterized the transcriptomic profile of affected and unaffected skin in small populations. In this study of 20 patients, RNA from lesional and matching non-lesional skin biopsies in 20 subjects were used to identify an expression-based HS disease signature. This was followed by differential expression and pathway enrichment analyses, as well as jointly reanalyzing our findings with previously published transcriptomic profiles. We establish an RNA-Seq based HS expression disease signature that is mostly consistent with previous reports. Bulk-RNA profiles from 104 subjects in 7 previously reported data sets identified a disease signature of 118 differentially regulated genes compared to three control data sets from non-lesional skin. We confirmed previously reported expression profiles and further characterized dysregulation in complement activation and host response to bacteria in disease pathogenesis. Changes in the transcriptome of lesional skin in this cohort of HS patients is consistent with smaller previously reported populations. The findings further support the significance of immune dysregulation, in particular with regard to bacterial response mechanisms. Joint analysis of this and previously reported cohorts indicate a remarkably consistent expression profile.

Generalized random set framework for functional enrichment analysis using primary genomics datasets.

MOTIVATION: Functional enrichment analysis using primary genomics datasets is an emerging approach to complement established methods for functional enrichment based on predefined lists of functionally related genes. Currently used methods depend on creating lists of 'significant' and 'non-significant' genes based on ad hoc significance cutoffs. This can lead to loss of statistical power and can introduce biases affecting the interpretation of experimental results. RESULTS: We developed and validated a new statistical framework, generalized random set (GRS) analysis, for comparing the genomic signatures in two datasets without the need for gene categorization. In our tests, GRS produced correct measures of statistical significance, and it showed dramatic improvement in the statistical power over other methods currently used in this setting. We also developed a procedure for identifying genes driving the concordance of the genomics profiles and demonstrated a dramatic improvement in functional coherence of genes identified in such analysis. AVAILABILITY: GRS can be downloaded as part of the R package CLEAN from http://ClusterAnalysis.org/. An online implementation is available at http://GenomicsPortals.org/.

Integrated analysis of dermal blister fluid proteomics and genome-wide skin gene expression in systemic sclerosis: an observational study.

Background: Skin fibrosis is a hallmark feature of systemic sclerosis. Skin biopsy transcriptomics and blister fluid proteomics give insight into the local environment of the skin. We have integrated these modalities with the aim of developing a surrogate for the modified Rodnan skin score (mRSS), using candidate genes and proteins from the skin and blister fluid as anchors to identify key analytes in the plasma. Methods: In this single-centre, prospective observational study at the Royal Free Campus, University College London, London, UK, transcriptional and proteomic analyses of blood and skin were performed in a cohort of patients with systemic sclerosis (n=52) and healthy controls (n=16). Weighted gene co-expression network analysis was used to explore the association of skin transcriptomics data, clinical traits, and blister fluid proteomic results. Candidate hub analytes were identified as those present in both blister and skin gene sets (modules), and which correlated with plasma (module membership >0·7 and gene significance >0·6). Hub analytes were confirmed using RNA transcript data obtained from skin biopsy samples from patients with early diffuse cutaneous systemic sclerosis at 12 months. Findings: We identified three modules in the skin, and two in blister fluid, which correlated with a diagnosis of early diffuse cutaneous systemic sclerosis. From these modules, 11 key hub analytes were identified, present in both skin and blister fluid modules, whose transcript and protein levels correlated with plasma protein concentrations, mRSS, and showed statistically significant correlation on repeat transcriptomic samples taken at 12 months. Multivariate analysis identified four plasma analytes as correlates of mRSS (COL4A1, COMP, SPON1, and TNC), which can be used to differentiate disease subtype. Interpretation: This unbiased approach has identified potential biological candidates that might be drivers of local skin pathogenesis in systemic sclerosis. By focusing on measurable analytes in the plasma, we generated a promising composite plasma protein biomarker that could be used for assessment of skin severity, case stratification, and as a potential outcome measure for clinical trials and practice. Once fully validated, the biomarker score could replace a clinical score such as the mRSS, which carries substantial variability. Funding: GlaxoSmithKline and UK Medical Research Council.

A semi-parametric Bayesian model for unsupervised differential co-expression analysis.

BACKGROUND: Differential co-expression analysis is an emerging strategy for characterizing disease related dysregulation of gene expression regulatory networks. Given pre-defined sets of biological samples, such analysis aims at identifying genes that are co-expressed in one, but not in the other set of samples. RESULTS: We developed a novel probabilistic framework for jointly uncovering contexts (i.e. groups of samples) with specific co-expression patterns, and groups of genes with different co-expression patterns across such contexts. In contrast to current clustering and bi-clustering procedures, the implicit similarity measure in this model used for grouping biological samples is based on the clustering structure of genes within each sample and not on traditional measures of gene expression level similarities. Within this framework, biological samples with widely discordant expression patterns can be placed in the same context as long as the co-clustering structure of genes is concordant within these samples. To the best of our knowledge, this is the first method to date for unsupervised differential co-expression analysis in this generality. When applied to the problem of identifying molecular subtypes of breast cancer, our method identified reproducible patterns of differential co-expression across several independent expression datasets. Sample groupings induced by these patterns were highly informative of the disease outcome. Expression patterns of differentially co-expressed genes provided new insights into the complex nature of the ERalpha regulatory network. CONCLUSIONS: We demonstrated that the use of the co-clustering structure as the similarity measure in the unsupervised analysis of sample gene expression profiles provides valuable information about expression regulatory networks.

CLEAN: CLustering Enrichment ANalysis.

BACKGROUND: Integration of biological knowledge encoded in various lists of functionally related genes has become one of the most important aspects of analyzing genome-wide functional genomics data. In the context of cluster analysis, functional coherence of clusters established through such analyses have been used to identify biologically meaningful clusters, compare clustering algorithms and identify biological pathways associated with the biological process under investigation. RESULTS: We developed a computational framework for analytically and visually integrating knowledge-based functional categories with the cluster analysis of genomics data. The framework is based on the simple, conceptually appealing, and biologically interpretable gene-specific functional coherence score (CLEAN score). The score is derived by correlating the clustering structure as a whole with functional categories of interest. We directly demonstrate that integrating biological knowledge in this way improves the reproducibility of conclusions derived from cluster analysis. The CLEAN score differentiates between the levels of functional coherence for genes within the same cluster based on their membership in enriched functional categories. We show that this aspect results in higher reproducibility across independent datasets and produces more informative genes for distinguishing different sample types than the scores based on the traditional cluster-wide analysis. We also demonstrate the utility of the CLEAN framework in comparing clusterings produced by different algorithms. CLEAN was implemented as an add-on R package and can be downloaded at http://Clusteranalysis.org. The package integrates routines for calculating gene specific functional coherence scores and the open source interactive Java-based viewer Functional TreeView (FTreeView). CONCLUSION: Our results indicate that using the gene-specific functional coherence score improves the reproducibility of the conclusions made about clusters of co-expressed genes over using the traditional cluster-wide scores. Using gene-specific coherence scores also simplifies the comparisons of clusterings produced by different clustering algorithms and provides a simple tool for selecting genes with a "functionally coherent" expression profile.