Clusters Of Co-regulated Genes Research Articles

Identifying multiscale translational safety biomarkers using a network-based systems approach

Animal testing is the current standard for drug and chemicals safety assessment, but hazards translation to human is uncertain. Human invitro models can address the species translation but might not replicate invivo complexity. Herein, we propose a network-based method addressing these translational multiscale problems that derives invivo liver injury biomarkers applicable to invitro human early safety screening. We applied weighted correlation network analysis (WGCNA) toa large rat liver transcriptomic dataset to obtain co-regulated gene clusters (modules). We identified modules statistically associated with liver pathologies, including a module enriched for ATF4-regulated genes as associated with theoccurrence of hepatocellular single-cell necrosis, and as preserved in human liver invitro models. Within the module, we identified TRIB3 and MTHFD2 as a novel candidate stress biomarkers, and developed and used BAC-eGFPHepG2 reporters in a compound screening, identifying compounds showing ATF4-dependent stress response and potential early safety signals.

Identification of the key metabolites and related genes network modules highly associated with the nutrients and taste components among different Pepino (Solanum muricatum) cultivars

There is considerable knowledge about plant compounds that produce flavor, scent, and aroma. Aside from the similarities, however, groups of plant-produced nutrients and taste components have little in common with each other. Network analysis holds promise for metabolic gene discovery, which is especially important in plant systems where metabolic networks are not yet fully resolved. To bridge this gap, we propose a joint model of gene regulation and metabolic reactions in two different pepino varieties. Differential metabolomics analysis is carried out for detection of eventual interaction of compound. We adopted a multi-omics approach to profile the transcriptome and metabolome analyze differences in phenolic acids, flavonoids, organic acids, lipids, alkaloids, and sugars between LOF and SRF. The two most predominant classes of metabolites are phenolic acids and lipids in pepino. Overall results show enrichment in most DEGs was carbohydrate and biosynthesis of secondary metabolites pathway. Results of DEMs predominantly comprised N-p-coumaroyl agmatine and tryptamine, and significant differences were observed in their expression between LOF and SRF. Integrated DEMs and DEGs specific networks were constructed by combining two types of networks: transcriptional regulatory networks composed of interactions between DEMs and the regulated genes, and pepino metabolite–metabolite interaction networks. Newly discovered features, such as DEGs (USPA, UBE2 and DELLA) involved in the production of secondary metabolites are found in coregulated gene clusters. Moreover, lipid metabolites were most involved in DEMs correlations by OPLS-DA while identifying a significant number of DEGs co-regulated by SENP1, HMGCS et al. These results further that the metabolite discrepancies result from characterized the nutrients and taste components between two pepino genotype. Among the possible causes of the differences between species in pepino metabolite concentrations is co-regulated by these DEGs, continue to suggest that novel features of metabolite biosynthetic pathway remain to be uncovered. Finally, the integrated metabolome and transcriptome analyses have revealed that many important metabolic pathways are regulated at the transcriptional level. The metabolites content differences observed among varieties of the same species mainly originates from different regulated genes and enzymes expression. Overall, this study provides new insights into the underlying causes of differences in the plant metabolites and suggests that genetic data can be used to improve its nutrients and taste components.

Co-Regulated Genes and Gene Clusters.

There are many co-regulated genes in eukaryotic cells. The coordinated activation or repression of such genes occurs at specific stages of differentiation, or under the influence of external stimuli. As a rule, co-regulated genes are dispersed in the genome. However, there are also gene clusters, which contain paralogous genes that encode proteins with similar functions. In this aspect, they differ significantly from bacterial operons containing functionally linked genes that are not paralogs. In this review, we discuss the reasons for the existence of gene clusters in vertebrate cells and propose that clustering is necessary to ensure the possibility of selective activation of one of several similar genes.

Genetic factors contributing to extensive variability of sex-specific hepatic gene expression in Diversity Outbred mice.

Sex-specific transcription characterizes hundreds of genes in mouse liver, many implicated in sex-differential drug and lipid metabolism and disease susceptibility. While the regulation of liver sex differences by growth hormone-activated STAT5 is well established, little is known about autosomal genetic factors regulating the sex-specific liver transcriptome. Here we show, using genotyping and expression data from a large population of Diversity Outbred mice, that genetic factors work in tandem with growth hormone to control the individual variability of hundreds of sex-biased genes, including many long non-coding RNA genes. Significant associations between single nucleotide polymorphisms and sex-specific gene expression were identified as expression quantitative trait loci (eQTLs), many of which showed strong sex-dependent associations. Remarkably, autosomal genetic modifiers of sex-specific genes were found to account for more than 200 instances of gain or loss of sex-specificity across eight Diversity Outbred mouse founder strains. Sex-biased STAT5 binding sites and open chromatin regions with strain-specific variants were significantly enriched at eQTL regions regulating correspondingly sex-specific genes, supporting the proposed functional regulatory nature of the eQTL regions identified. Binding of the male-biased, growth hormone-regulated repressor BCL6 was most highly enriched at trans-eQTL regions controlling female-specific genes. Co-regulated gene clusters defined by overlapping eQTLs included sets of highly correlated genes from different chromosomes, further supporting trans-eQTL action. These findings elucidate how an unexpectedly large number of autosomal factors work in tandem with growth hormone signaling pathways to regulate the individual variability associated with sex differences in liver metabolism and disease.

Spermiogenesis and Male Fertility Require the Function of Suppressor of Hairy-Wing in Somatic Cyst Cells of Drosophila.

Drosophila Suppressor of Hairy-wing [Su(Hw)] protein is an example of a multivalent transcription factor. Although best known for its role in establishing the chromatin insulator of the gypsy retrotransposon, Su(Hw) functions as an activator and repressor at non-gypsy genomic sites. It remains unclear how the different regulatory activities of Su(Hw) are utilized during development. Motivated from observations of spatially restricted expression of Su(Hw) in the testis, we investigated the role of Su(Hw) in spermatogenesis to advance an understanding of its developmental contributions as an insulator, repressor, and activator protein. We discovered that Su(Hw) is required for sustained male fertility. Although dynamics of Su(Hw) expression coincide with changes in nuclear architecture and activation of coregulated testis-specific gene clusters, we show that loss of Su(Hw) does not disrupt meiotic chromosome pairing or transcription of testis-specific genes, suggesting that Su(Hw) has minor architectural or insulator functions in the testis. Instead, Su(Hw) has a prominent role as a repressor of neuronal genes, consistent with suggestions that Su(Hw) is a functional homolog of mammalian REST, a repressor of neuronal genes in non-neuronal tissues. We show that Su(Hw) regulates transcription in both germline and somatic cells. Surprisingly, the essential spermatogenesis function of Su(Hw) resides in somatic cyst cells, implying context-specific consequences due to loss of this transcription factor. Together, our studies highlight that Su(Hw) has a major developmental function as a transcriptional repressor, with the effect of its loss dependent upon the cell-specific factors.

Genome-wide Analysis of Transcriptional Variability in a Large Maize-Teosinte Population

Gene expression regulation plays an important role in controlling plant phenotypes and adaptation. Here, we report a comprehensive assessment of gene expression variation through the transcriptome analyses of a large maize-teosinte experimental population. Genome-wide mapping identified 25 660 expression quantitative trait loci (eQTL) for 17 311 genes, capturing an unprecedented range of expression variation. We found that local eQTL were more frequently mapped to adjacent genes, displaying a mode of expression piggybacking, which consequently created co-regulated gene clusters. Genes within the co-regulated gene clusters tend to have relevant functions and shared chromatin modifications. Distant eQTL formed 125 significant distant eQTL hotspots with their targets significantly enriched in specific functional categories. By integrating different sources of information, we identified putative trans- regulators for a variety of metabolic pathways. We demonstrated that the bHLH transcription factor R1 and hexokinase HEX9 might act as crucial regulators for flavonoid biosynthesis and glycolysis, respectively. Moreover, we showed that domestication or improvement has significantly affected global gene expression, with many genes targeted by selection. Of particular interest, the Bx genes for benzoxazinoid biosynthesis may have undergone coordinated cis-regulatory divergence between maize and teosinte, and a transposon insertion that inactivates Bx12 was under strong selection as maize spread into temperate environments with a distinct herbivore community.

Genome urbanization: clusters of topologically co-regulated genes delineate functional compartments in the genome of Saccharomyces cerevisiae.

The eukaryotic genome evolves under the dual constraint of maintaining coordinated gene transcription and performing effective DNA replication and cell division, the coupling of which brings about inevitable DNA topological tension. DNA supercoiling is resolved and, in some cases, even harnessed by the genome through the function of DNA topoisomerases, as has been shown in the concurrent transcriptional activation and suppression of genes upon transient deactivation of topoisomerase II (topoII). By analyzing a genome-wide transcription run-on experiment upon thermal inactivation of topoII in Saccharomyces cerevisiae we were able to define 116 gene clusters of consistent response (either positive or negative) to topological stress. A comprehensive analysis of these topologically co-regulated gene clusters reveals pronounced preferences regarding their functional, regulatory and structural attributes. Genes that negatively respond to topological stress, are positioned in gene-dense pericentromeric regions, are more conserved and associated to essential functions, while upregulated gene clusters are preferentially located in the gene-sparse nuclear periphery, associated with secondary functions and under complex regulatory control. We propose that genome architecture evolves with a core of essential genes occupying a compact genomic ‘old town’, whereas more recently acquired, condition-specific genes tend to be located in a more spacious ‘suburban’ genomic periphery.

Biologically Active Secondary Metabolites from the Fungi.

Many Fungi have a well-developed secondary metabolism. The diversity of fungal species and the diversification of biosynthetic gene clusters underscores a nearly limitless potential for metabolic variation and an untapped resource for drug discovery and synthetic biology. Much of the ecological success of the filamentous fungi in colonizing the planet is owed to their ability to deploy their secondary metabolites in concert with their penetrative and absorptive mode of life. Fungal secondary metabolites exhibit biological activities that have been developed into life-saving medicines and agrochemicals. Toxic metabolites, known as mycotoxins, contaminate human and livestock food and indoor environments. Secondary metabolites are determinants of fungal diseases of humans, animals, and plants. Secondary metabolites exhibit a staggering variation in chemical structures and biological activities, yet their biosynthetic pathways share a number of key characteristics. The genes encoding cooperative steps of a biosynthetic pathway tend to be located contiguously on the chromosome in coregulated gene clusters. Advances in genome sequencing, computational tools, and analytical chemistry are enabling the rapid connection of gene clusters with their metabolic products. At least three fungal drug precursors, penicillin K and V, mycophenolic acid, and pleuromutilin, have been produced by synthetic reconstruction and expression of respective gene clusters in heterologous hosts. This review summarizes general aspects of fungal secondary metabolism and recent developments in our understanding of how and why fungi make secondary metabolites, how these molecules are produced, and how their biosynthetic genes are distributed across the Fungi. The breadth of fungal secondary metabolite diversity is highlighted by recent information on the biosynthesis of important fungus-derived metabolites that have contributed to human health and agriculture and that have negatively impacted crops, food distribution, and human environments.

Identification and characterization of large DNA deletions affecting oil quality traits in soybean seeds through transcriptome sequencing analysis.

Key messageIdentification and characterization of a 254-kb genomic deletion on a duplicated chromosome segment that resulted in a low level of palmitic acid in soybean seeds using transcriptome sequencing.A large number of soybean genotypes varying in seed oil composition and content have been identified. Understanding the molecular mechanisms underlying these variations is important for breeders to effectively utilize them as a genetic resource. Through design and application of a bioinformatics approach, we identified nine co-regulated gene clusters by comparing seed transcriptomes of nine soybean genotypes varying in oil composition and content. We demonstrated that four gene clusters in the genotypes M23, Jack and N0304-303-3 coincided with large-scale genome rearrangements. The co-regulated gene clusters in M23 and Jack mapped to a previously described 164-kb deletion and a copy number amplification of the Rhg1 locus, respectively. The coordinately down-regulated gene clusters in N0304-303-3 were caused by a 254-kb deletion containing 19 genes including a fatty acyl-ACP thioesterase B gene (FATB1a). This deletion was associated with reduced palmitic acid content in seeds and was the molecular cause of a previously reported nonfunctional FATB1a allele, fapnc. The M23 and N0304-304-3 deletions were located in duplicated genome segments retained from the Glycine-specific whole genome duplication that occurred 13 million years ago. The homoeologous genes in these duplicated regions shared a strong similarity in both their encoded protein sequences and transcript accumulation levels, suggesting that they may have conserved and important functions in seeds. The functional conservation of homoeologous genes may result in genetic redundancy and gene dosage effects for their associated seed traits, explaining why the large deletion did not cause lethal effects or completely eliminate palmitic acid in N0304-303-3.Electronic supplementary materialThe online version of this article (doi:10.1007/s00122-016-2725-z) contains supplementary material, which is available to authorized users.

FunGeneClusterS: Predicting fungal gene clusters from genome and transcriptome data

IntroductionSecondary metabolites of fungi are receiving an increasing amount of interest due to their prolific bioactivities and the fact that fungal biosynthesis of secondary metabolites often occurs from co-regulated and co-located gene clusters. This makes the gene clusters attractive for synthetic biology and industrial biotechnology applications. We have previously published a method for accurate prediction of clusters from genome and transcriptome data, which could also suggest cross-chemistry, however, this method was limited both in the number of parameters which could be adjusted as well as in user-friendliness. Furthermore, sensitivity to the transcriptome data required manual curation of the predictions. In the present work, we have aimed at improving these features. ResultsFunGeneClusterS is an improved implementation of our previous method with a graphical user interface for off- and on-line use. The new method adds options to adjust the size of the gene cluster(s) being sought as well as an option for the algorithm to be flexible with genes in the cluster which may not seem to be co-regulated with the remainder of the cluster. We have benchmarked the method using data from the well-studied Aspergillus nidulans and found that the method is an improvement over the previous one. In particular, it makes it possible to predict clusters with more than 10 genes more accurately, and allows identification of co-regulated gene clusters irrespective of the function of the genes. It also greatly reduces the need for manual curation of the prediction results. We furthermore applied the method to transcriptome data from A. niger. Using the identified best set of parameters, we were able to identify clusters for 31 out of 76 previously predicted secondary metabolite synthases/synthetases. Furthermore, we identified additional putative secondary metabolite gene clusters. In total, we predicted 432 co-transcribed gene clusters in A. niger (spanning 1.323 genes, 12% of the genome). Some of these had functions related to primary metabolism, e.g. we have identified a cluster for biosynthesis of biotin, as well as several for degradation of aromatic compounds. The data identifies that suggests that larger parts of the fungal genome than previously anticipated operates as gene clusters. This includes both primary and secondary metabolism as well as other cellular maintenance functions. ConclusionWe have developed FunGeneClusterS in a graphical implementation and made the method capable of adjustments to different datasets and target clusters. The method is versatile in that it can predict co-regulated clusters not limited to secondary metabolism. Our analysis of data has shown not only the validity of the method, but also strongly suggests that large parts of fungal primary metabolism and cellular functions are both co-regulated and co-located.

The Cryptosporidium parvum ApiAP2 gene family: insights into the evolution of apicomplexan AP2 regulatory systems.

We provide the first comprehensive analysis of any transcription factor family in Cryptosporidium, a basal-branching apicomplexan that is the second leading cause of infant diarrhea globally. AP2 domain-containing proteins have evolved to be the major regulatory family in the phylum to the exclusion of canonical regulators. We show that apicomplexan and perkinsid AP2 domains cluster distinctly from other chromalveolate AP2s. Protein-binding specificity assays of C. parvum AP2 domains combined with motif conservation upstream of co-regulated gene clusters allowed the construction of putative AP2 regulons across the in vitro life cycle. Orthologous Apicomplexan AP2 (ApiAP2) expression has been rearranged relative to the malaria parasite P. falciparum, suggesting ApiAP2 network rewiring during evolution. C. hominis orthologs of putative C. parvum ApiAP2 proteins and target genes show greater than average variation. C. parvum AP2 domains display reduced binding diversity relative to P. falciparum, with multiple domains binding the 5′-TGCAT-3′, 5′-CACACA-3′ and G-box motifs (5′-G[T/C]GGGG-3′). Many overrepresented motifs in C. parvum upstream regions are not AP2 binding motifs. We propose that C. parvum is less reliant on ApiAP2 regulators in part because it utilizes E2F/DP1 transcription factors. C. parvum may provide clues to the ancestral state of apicomplexan transcriptional regulation, pre-AP2 domination.

Modeling forest ecosystem responses to elevated carbon dioxide and ozone using artificial neural networks

Rising atmospheric levels of carbon dioxide and ozone will impact productivity and carbon sequestration in forest ecosystems. The scale of this process and the potential economic consequences provide an incentive for the development of models to predict the types and rates of ecosystem responses and feedbacks that result from and influence of climate change. In this paper, we use phenotypic and molecular data derived from the Aspen Free Air CO2 Enrichment site (Aspen-FACE) to evaluate modeling approaches for ecosystem responses to changing conditions. At FACE, it was observed that different aspen clones exhibit clone-specific responses to elevated atmospheric levels of carbon dioxide and ozone. To identify the molecular basis for these observations, we used artificial neural networks (ANN) to examine above and below-ground community phenotype responses to elevated carbon dioxide, elevated ozone and gene expression profiles. The aspen community models generated using this approach identified specific genes and subnetworks of genes associated with variable sensitivities for aspen clones. The ANN model also predicts specific co-regulated gene clusters associated with differential sensitivity to elevated carbon dioxide and ozone in aspen species. The results suggest ANN is an effective approach to predict relevant gene expression changes resulting from environmental perturbation and provides useful information for the rational design of future biological experiments.

Topology and Control of the Cell-Cycle-Regulated Transcriptional Circuitry

Nearly 20% of the budding yeast genome is transcribed periodically during the cell division cycle. The precise temporal execution of this large transcriptional program is controlled by a large interacting network of transcriptional regulators, kinases, and ubiquitin ligases. Historically, this network has been viewed as a collection of four coregulated gene clusters that are associated with each phase of the cell cycle. Although the broad outlines of these gene clusters were described nearly 20 years ago, new technologies have enabled major advances in our understanding of the genes comprising those clusters, their regulation, and the complex regulatory interplay between clusters. More recently, advances are being made in understanding the roles of chromatin in the control of the transcriptional program. We are also beginning to discover important regulatory interactions between the cell-cycle transcriptional program and other cell-cycle regulatory mechanisms such as checkpoints and metabolic networks. Here we review recent advances and contemporary models of the transcriptional network and consider these models in the context of eukaryotic cell-cycle controls.

Genome-wide analysis of coordinated transcript abundance during seed development in different Brassica rapa morphotypes

BackgroundBrassica seeds are important as basic units of plant growth and sources of vegetable oil. Seed development is regulated by many dynamic metabolic processes controlled by complex networks of spatially and temporally expressed genes. We conducted a global microarray gene co-expression analysis by measuring transcript abundance of developing seeds from two diverse B. rapa morphotypes: a pak choi (leafy-type) and a yellow sarson (oil-type), and two of their doubled haploid (DH) progenies, (1) to study the timing of metabolic processes in developing seeds, (2) to explore the major transcriptional differences in developing seeds of the two morphotypes, and (3) to identify the optimum stage for a genetical genomics study in B. rapa seed.ResultsSeed developmental stages were similar in developing seeds of pak choi and yellow sarson of B. rapa; however, the colour of embryo and seed coat differed among these two morphotypes. In this study, most transcriptional changes occurred between 25 and 35 DAP, which shows that the timing of seed developmental processes in B. rapa is at later developmental stages than in the related species B. napus. Using a Weighted Gene Co-expression Network Analysis (WGCNA), we identified 47 “gene modules”, of which 27 showed a significant association with temporal and/or genotypic variation. An additional hierarchical cluster analysis identified broad spectra of gene expression patterns during seed development. The predominant variation in gene expression was according to developmental stages rather than morphotype differences. Since lipids are the major storage compounds of Brassica seeds, we investigated in more detail the regulation of lipid metabolism. Four co-regulated gene clusters were identified with 17 putative cis-regulatory elements predicted in their 1000 bp upstream region, either specific or common to different lipid metabolic pathways.ConclusionsThis is the first study of genome-wide profiling of transcript abundance during seed development in B. rapa. The identification of key physiological events, major expression patterns, and putative cis-regulatory elements provides useful information to construct gene regulatory networks in B. rapa developing seeds and provides a starting point for a genetical genomics study of seed quality traits.Electronic supplementary materialThe online version of this article (doi:10.1186/1471-2164-14-840) contains supplementary material, which is available to authorized users.

Inactivation of Intergenic Enhancers by EBNA3A Initiates and Maintains Polycomb Signatures across a Chromatin Domain Encoding CXCL10 and CXCL9

Epstein-Barr virus (EBV) causes a persistent infection in human B cells by establishing specific transcription programs to control B cell activation and differentiation. Transcriptional reprogramming of EBV infected B cells is predominantly driven by the action of EBV nuclear antigens, among them the transcriptional repressor EBNA3A. By comparing gene expression profiles of wt and EBNA3A negative EBV infected B cells, we have previously identified a broad array of cellular genes controlled by EBNA3A. We now find that genes repressed by EBNA3A in these cells are significantly enriched for the repressive histone mark H3K27me3, which is installed by Polycomb group (PcG) proteins. This PcG-controlled subset of genes also carries H3K27me3 marks in a variety of other tissues, suggesting that the commitment to PcG silencing is an intrinsic feature of these gene loci that can be used by EBNA3A. In addition, EBNA3A targets frequently reside in co-regulated gene clusters. To study the mechanism of gene repression by EBNA3A and to evaluate the relative contribution of PcG proteins during this process, we have selected the genomic neighbors CXCL10 and CXCL9 as a model for co-repressed and PcG-controlled genes. We show that EBNA3A binds to CBF1 occupied intergenic enhancers located between CXCL10 and CXCL9 and displaces the transactivator EBNA2. This impairs enhancer activity, resulting in a rapid transcriptional shut-down of both genes in a CBF1-dependent manner and initiation of a delayed gain of H3K27me3 marks covering an extended chromatin domain. H3K27me3 marks increase gradually and are maintained by EBNA3A. Our study provides direct evidence that repression by EBNA3A requires CBF1 and that EBNA3A and EBNA2 compete for access to CBF1 at identical genomic sites. Most importantly, our results demonstrate that transcriptional silencing by EBNA3A precedes the appearance of repressive PcG marks and indicate that both events are triggered by loss of enhancer activity.

Correlating transcriptional networks to breast cancer survival: a large-scale coexpression analysis

Weighted gene coexpression network analysis (WGCNA) is a powerful 'guilt-by-association'-based method to extract coexpressed groups of genes from large heterogeneous messenger RNA expression data sets. We have utilized WGCNA to identify 11 coregulated gene clusters across 2342 breast cancer samples from 13 microarray-based gene expression studies. A number of these transcriptional modules were found to be correlated to clinicopathological variables (e.g. tumor grade), survival endpoints for breast cancer as a whole (disease-free survival, distant disease-free survival and overall survival) and also its molecular subtypes (luminal A, luminal B, HER2+ and basal-like). Examples of findings arising from this work include the identification of a cluster of proliferation-related genes that when upregulated correlated to increased tumor grade and were associated with poor survival in general. The prognostic potential of novel genes, for example, ubiquitin-conjugating enzyme E2S (UBE2S) within this group was confirmed in an independent data set. In addition, gene clusters were also associated with survival for breast cancer molecular subtypes including a cluster of genes that was found to correlate with prognosis exclusively for basal-like breast cancer. The upregulation of several single genes within this coexpression cluster, for example, the potassium channel, subfamily K, member 5 (KCNK5) was associated with poor outcome for the basal-like molecular subtype. We have developed an online database to allow user-friendly access to the coexpression patterns and the survival analysis outputs uncovered in this study (available at http://glados.ucd.ie/Coexpression/).

CoBi: Pattern Based Co-Regulated Biclustering of Gene Expression Data

Co-regulation is a common phenomenon in gene expression. Finding positively and negatively co-regulated gene clusters from gene expression data is a real need. Existing techniques based on global similarity are unable to detect true up- and down-regulated gene clusters. This paper presents an expression pattern based biclustering technique, CoBi, for grouping both positively and negatively regulated genes from microarray expression data. Regulation pattern and similarity in degree of fluctuation are accounted for while computing similarity between two genes. Unlike traditional biclustering techniques, which use greedy iterative approaches, it uses a BiClust tree that needs single pass over the entire dataset to find a set of biologically relevant biclusters. Biclusters determined from different gene expression datasets by the technique show highly enriched functional categories.

The Limits of De Novo DNA Motif Discovery

A major challenge in molecular biology is reverse-engineering the cis-regulatory logic that plays a major role in the control of gene expression. This program includes searching through DNA sequences to identify “motifs” that serve as the binding sites for transcription factors or, more generally, are predictive of gene expression across cellular conditions. Several approaches have been proposed for de novo motif discovery–searching sequences without prior knowledge of binding sites or nucleotide patterns. However, unbiased validation is not straightforward. We consider two approaches to unbiased validation of discovered motifs: testing the statistical significance of a motif using a DNA “background” sequence model to represent the null hypothesis and measuring performance in predicting membership in gene clusters. We demonstrate that the background models typically used are “too null,” resulting in overly optimistic assessments of significance, and argue that performance in predicting TF binding or expression patterns from DNA motifs should be assessed by held-out data, as in predictive learning. Applying this criterion to common motif discovery methods resulted in universally poor performance, although there is a marked improvement when motifs are statistically significant against real background sequences. Moreover, on synthetic data where “ground truth” is known, discriminative performance of all algorithms is far below the theoretical upper bound, with pronounced “over-fitting” in training. A key conclusion from this work is that the failure of de novo discovery approaches to accurately identify motifs is basically due to statistical intractability resulting from the fixed size of co-regulated gene clusters, and thus such failures do not necessarily provide evidence that unfound motifs are not active biologically. Consequently, the use of prior knowledge to enhance motif discovery is not just advantageous but necessary. An implementation of the LR and ALR algorithms is available at http://code.google.com/p/likelihood-ratio-motifs/.

A Single Cluster of Coregulated Genes Encodes the Biosynthesis of the Mycotoxins Roquefortine C and Meleagrin in Penicillium chrysogenum

A single gene cluster of Penicillium chrysogenum contains genes involved in the biosynthesis and secretion of the mycotoxins roquefortine C and meleagrin. Five of these genes have been silenced by RNAi. Pc21g15480 (rds) encodes a nonribosomal cyclodipeptide synthetase for the biosynthesis of both roquefortine C and meleagrin. Pc21g15430 (rpt) encodes a prenyltransferase also required for the biosynthesis of both mycotoxins. Silencing of Pc21g15460 or Pc21g15470 led to a decrease in roquefortine C and meleagrin, whereas silencing of the methyltransferase gene (Pc21g15440; gmt) resulted in accumulation of glandicolin B, indicating that this enzyme catalyzes the conversion of glandicolin B to meleagrin. All these genes are transcriptionally coregulated. Our results prove that roquefortine C and meleagrin derive from a single pathway.

Formation of plant metabolic gene clusters within dynamic chromosomal regions

In bacteria, genes with related functions often are grouped together in operons and are cotranscribed as a single polycistronic mRNA. In eukaryotes, functionally related genes generally are scattered across the genome. Notable exceptions include gene clusters for catabolic pathways in yeast, synthesis of secondary metabolites in filamentous fungi, and the major histocompatibility complex in animals. Until quite recently it was thought that gene clusters in plants were restricted to tandem duplicates (for example, arrays of leucine-rich repeat disease-resistance genes). However, operon-like clusters of coregulated nonhomologous genes are an emerging theme in plant biology, where they may be involved in the synthesis of certain defense compounds. These clusters are unlikely to have arisen by horizontal gene transfer, and the mechanisms behind their formation are poorly understood. Previously in thale cress (Arabidopsis thaliana) we identified an operon-like gene cluster that is required for the synthesis and modification of the triterpene thalianol. Here we characterize a second operon-like triterpene cluster (the marneral cluster) from A. thaliana, compare the features of these two clusters, and investigate the evolutionary events that have led to cluster formation. We conclude that common mechanisms are likely to underlie the assembly and control of operon-like gene clusters in plants.