WGCNA: an R package for weighted correlation network analysis

Resting-state quantitative electroencephalography reveals increased neurophysiologic connectivity in depression

Symptoms of Major Depressive Disorder (MDD) are hypothesized to arise from dysfunction in brain networks linking the limbic system and cortical regions. Alterations in brain functional cortical connectivity in resting-state networks have been detected with functional imaging techniques, but neurophysiologic connectivity measures have not been systematically examined. We used weighted network analysis to examine resting state functional connectivity as measured by quantitative electroencephalographic (qEEG) coherence in 121 unmedicated subjects with MDD and 37 healthy controls. Subjects with MDD had significantly higher overall coherence as compared to controls in the delta (0.5-4 Hz), theta (4-8 Hz), alpha (8-12 Hz), and beta (12-20 Hz) frequency bands. The frontopolar region contained the greatest number of "hub nodes" (surface recording locations) with high connectivity. MDD subjects expressed higher theta and alpha coherence primarily in longer distance connections between frontopolar and temporal or parietooccipital regions, and higher beta coherence primarily in connections within and between electrodes overlying the dorsolateral prefrontal cortical (DLPFC) or temporal regions. Nearest centroid analysis indicated that MDD subjects were best characterized by six alpha band connections primarily involving the prefrontal region. The present findings indicate a loss of selectivity in resting functional connectivity in MDD. The overall greater coherence observed in depressed subjects establishes a new context for the interpretation of previous studies showing differences in frontal alpha power and synchrony between subjects with MDD and normal controls. These results can inform the development of qEEG state and trait biomarkers for MDD.

Coexpression network analysis in abdominal and gluteal adipose tissue reveals regulatory genetic loci for metabolic syndrome and related phenotypes

Metabolic Syndrome (MetS) is highly prevalent and has considerable public health impact, but its underlying genetic factors remain elusive. To identify gene networks involved in MetS, we conducted whole-genome expression and genotype profiling on abdominal (ABD) and gluteal (GLU) adipose tissue, and whole blood (WB), from 29 MetS cases and 44 controls. Co-expression network analysis for each tissue independently identified nine, six, and zero MetS–associated modules of coexpressed genes in ABD, GLU, and WB, respectively. Of 8,992 probesets expressed in ABD or GLU, 685 (7.6%) were expressed in ABD and 51 (0.6%) in GLU only. Differential eigengene network analysis of 8,256 shared probesets detected 22 shared modules with high preservation across adipose depots (D_ABD-GLU = 0.89), seven of which were associated with MetS (FDR P<0.01). The strongest associated module, significantly enriched for immune response–related processes, contained 94/620 (15%) genes with inter-depot differences. In an independent cohort of 145/141 twins with ABD and WB longitudinal expression data, median variability in ABD due to familiality was greater for MetS–associated versus un-associated modules (ABD: 0.48 versus 0.18, P = 0.08; GLU: 0.54 versus 0.20, P = 7.8×10⁻⁴). Cis-eQTL analysis of probesets associated with MetS (FDR P<0.01) and/or inter-depot differences (FDR P<0.01) provided evidence for 32 eQTLs. Corresponding eSNPs were tested for association with MetS–related phenotypes in two GWAS of >100,000 individuals; rs10282458, affecting expression of RARRES2 (encoding chemerin), was associated with body mass index (BMI) (P = 6.0×10⁻⁴); and rs2395185, affecting inter-depot differences of HLA-DRB1 expression, was associated with high-density lipoprotein (P = 8.7×10⁻⁴) and BMI–adjusted waist-to-hip ratio (P = 2.4×10⁻⁴). Since many genes and their interactions influence complex traits such as MetS, integrated analysis of genotypes and coexpression networks across multiple tissues relevant to clinical traits is an efficient strategy to identify novel associations.

Metabolic Syndrome (MetS) is a highly prevalent disorder with considerable public health concern, but its underlying genetic factors remain elusive. Given that most cellular components exert their functions through interactions with other cellular components, even the largest of genome-wide association (GWA) studies may often not detect their effects, nor necessarily provide insight into the complex molecular mechanisms of the disease. Rather than focusing on individual genes, the analysis of coexpression networks can be used for finding clusters (modules) of correlated expression levels across samples. In this study, we used a gene network–based approach for integrating clinical MetS, genotypic, and gene expression data from abdominal and gluteal adipose tissue and whole blood. We identified modules of genes related to MetS significantly enriched for immune response and oxidative phosphorylation pathways. We tested SNPs for association with MetS–associated expression (eSNPs), and tested prioritised eSNPs for association with MetS–related phenotypes in two large-scale GWA datasets. We identified two loci, neither of which had reached genome-wide significance levels in GWAs, associated with expression levels of RARRES2 and HLA-DRB1 and with MetS–related phenotypes, demonstrating that the integrated analysis of genotype and expression data from relevant multiple tissues can identify novel associations with complex traits such as MetS.

DNA methylation signatures in development and aging of the human prefrontal cortex

The human prefrontal cortex (PFC), a mastermind of the brain, is one of the last brain regions to mature. To investigate the role of epigenetics in the development of PFC, we examined DNA methylation in ∼14,500 genes at ∼27,000 CpG loci focused on 5' promoter regions in 108 subjects range in age from fetal to elderly. DNA methylation in the PFC shows unique temporal patterns across life. The fastest changes occur during the prenatal period, slow down markedly after birth and continue to slow further with aging. At the genome level, the transition from fetal to postnatal life is typified by a reversal of direction, from demethylation prenatally to increased methylation postnatally. DNA methylation is strongly associated with genotypic variants and correlates with expression of a subset of genes, including genes involved in brain development and in de novo DNA methylation. Our results indicate that promoter DNA methylation in the human PFC is a highly dynamic process modified by genetic variance and regulating gene transcription. Additional discovery is made possible with a stand-alone application, BrainCloudMethyl.

Integrated cross-species transcriptional network analysis of metastatic susceptibility

Metastatic disease is the proximal cause of mortality for most cancers and remains a significant problem for the clinical management of neoplastic disease. Recent advances in global transcriptional analysis have enabled better prediction of individuals likely to progress to metastatic disease. However, minimal overlap between predictive signatures has precluded easy identification of key biological processes contributing to the prometastatic transcriptional state. To overcome this limitation, we have applied network analysis to two independent human breast cancer datasets and three different mouse populations developed for quantitative analysis of metastasis. Analysis of these datasets revealed that the gene membership of the networks is highly conserved within and between species, and that these networks predicted distant metastasis free survival. Furthermore these results suggest that susceptibility to metastatic disease is cell-autonomous in estrogen receptor-positive tumors and associated with the mitotic spindle checkpoint. In contrast, nontumor genetics and pathway activities-associated stromal biology are significant modifiers of the rate of metastatic spread of estrogen receptor-negative tumors. These results suggest that the application of network analysis across species may provide a robust method to identify key biological programs associated with human cancer progression.

Genome-wide analysis of a Wnt1-regulated transcriptional network implicates neurodegenerative pathways

Wnt proteins are critical to mammalian brain development and function. The canonical Wnt signaling pathway involves the stabilization and nuclear translocation of β-catenin; however, Wnt also signals through alternative, noncanonical pathways. To gain a systems-level, genome-wide view of Wnt signaling, we analyzed Wnt1-stimulated changes in gene expression by transcriptional microarray analysis in cultured human neural progenitor (hNP) cells at multiple time points over a 72-hour time course. We observed a widespread oscillatory-like pattern of changes in gene expression, involving components of both the canonical and the noncanonical Wnt signaling pathways. A higher-order, systems-level analysis that combined independent component analysis, waveform analysis, and mutual information-based network construction revealed effects on pathways related to cell death and neurodegenerative disease. Wnt effectors were tightly clustered with presenilin1 (PSEN1) and granulin (GRN), which cause dominantly inherited forms of Alzheimer's disease and frontotemporal dementia (FTD), respectively. We further explored a potential link between Wnt1 and GRN and found that Wnt1 decreased GRN expression by hNPs. Conversely, GRN knockdown increased WNT1 expression, demonstrating that Wnt and GRN reciprocally regulate each other. Finally, we provided in vivo validation of the in vitro findings by analyzing gene expression data from individuals with FTD. These unbiased and genome-wide analyses provide evidence for a connection between Wnt signaling and the transcriptional regulation of neurodegenerative disease genes.

Systems genetics of susceptibility to obesity-induced diabetes in mice

Inbred strains of mice are strikingly different in susceptibility to obesity-driven diabetes. For instance, deficiency in leptin receptor (db/db) leads to hyperphagia and obesity in both C57BL/6 and DBA/2 mice, but only on the DBA/2 background do the mice develop beta-cell loss leading to severe diabetes, while C57BL/6 mice are relatively resistant. To further investigate the genetic factors predisposing to diabetes, we have studied leptin receptor-deficient offspring of an F2 cross between C57BL/6J (db/+) males and DBA/2J females. The results show that the genetics of diabetes susceptibility are enormously complex and a number of quantitative trait loci (QTL) contributing to diabetes-related traits were identified, notably on chromosomes 4, 6, 7, 9, 10, 11, 12, and 19. The Chr. 4 locus is likely due to a disruption of the Zfp69 gene in C57BL/6J mice. To identify candidate genes and to model coexpression networks, we performed global expression array analysis in livers of the F2 mice. Expression QTL (eQTL) were identified and used to prioritize candidate genes at clinical trait QTL. In several cases, clusters of eQTLs colocalized with clinical trait QTLs, suggesting a common genetic basis. We constructed coexpression networks for both 5 and 12 wk old mice and identified several modules significantly associated with clinical traits. One module in 12 wk old mice was associated with several measures of hepatic fat content as well as with other lipid- and diabetes-related traits. These results add to the understanding of the complex genetic interactions contributing to obesity-induced diabetes.

Comparing statistical methods for constructing large scale gene networks

The gene regulatory network (GRN) reveals the regulatory relationships among genes and can provide a systematic understanding of molecular mechanisms underlying biological processes. The importance of computer simulations in understanding cellular processes is now widely accepted; a variety of algorithms have been developed to study these biological networks. The goal of this study is to provide a comprehensive evaluation and a practical guide to aid in choosing statistical methods for constructing large scale GRNs. Using both simulation studies and a real application in E. coli data, we compare different methods in terms of sensitivity and specificity in identifying the true connections and the hub genes, the ease of use, and computational speed. Our results show that these algorithms performed reasonably well, and each method has its own advantages: (1) GeneNet, WGCNA (Weighted Correlation Network Analysis), and ARACNE (Algorithm for the Reconstruction of Accurate Cellular Networks) performed well in constructing the global network structure; (2) GeneNet and SPACE (Sparse PArtial Correlation Estimation) performed well in identifying a few connections with high specificity.

Large-scale neuroinformatics for in situ hybridization data in the mouse brain

Large-scale databases of the brain are providing content to the neuroscience community through molecular, cellular, functional, and connectomic data. Organization, presentation, and maintenance requirements are substantial given the complexity, diverse modalities, resolution, and scale. In addition to microarrays, magnetic resonance imaging, and RNA sequencing, several in situ hybridization databases have been constructed due to their value in spatially localizing cellular expression. Scalable techniques for processing and presenting these data for maximum utility in viewing and analysis are key for end user value. We describe methods and use cases for the Allen Brain Atlas resources of the adult and developing mouse.

Using genome-wide expression profiling to define gene networks relevant to the study of complex traits: from RNA integrity to network topology

Postgenomic studies of the function of genes and their role in disease have now become an area of intense study since efforts to define the raw sequence material of the genome have largely been completed. The use of whole-genome approaches such as microarray expression profiling and, more recently, RNA-sequence analysis of transcript abundance has allowed an unprecedented look at the workings of the genome. However, the accurate derivation of such high-throughput data and their analysis in terms of biological function has been critical to truly leveraging the postgenomic revolution. This chapter will describe an approach that focuses on the use of gene networks to both organize and interpret genomic expression data. Such networks, derived from statistical analysis of large genomic datasets and the application of multiple bioinformatics data resources, potentially allow the identification of key control elements for networks associated with human disease, and thus may lead to derivation of novel therapeutic approaches. However, as discussed in this chapter, the leveraging of such networks cannot occur without a thorough understanding of the technical and statistical factors influencing the derivation of genomic expression data. Thus, while the catch phrase may be "it's the network … stupid," the understanding of factors extending from RNA isolation to genomic profiling technique, multivariate statistics, and bioinformatics are all critical to defining fully useful gene networks for study of complex biology.

Reverse engineering: a key component of systems biology to unravel global abiotic stress cross-talk

Understanding the global abiotic stress response is an important stepping stone for the development of universal stress tolerance in plants in the era of climate change. Although co-occurrence of several stress factors (abiotic and biotic) in nature is found to be frequent, current attempts are poor to understand the complex physiological processes impacting plant growth under combinatory factors. In this review article, we discuss the recent advances of reverse engineering approaches that led to seminal discoveries of key candidate regulatory genes involved in cross-talk of abiotic stress responses and summarized the available tools of reverse engineering and its relevant application. Among the universally induced regulators involved in various abiotic stress responses, we highlight the importance of (i) abscisic acid (ABA) and jasmonic acid (JA) hormonal cross-talks and (ii) the central role of WRKY transcription factors (TF), potentially mediating both abiotic and biotic stress responses. Such interactome networks help not only to derive hypotheses but also play a vital role in identifying key regulatory targets and interconnected hormonal responses. To explore the full potential of gene network inference in the area of abiotic stress tolerance, we need to validate hypotheses by implementing time-dependent gene expression data from genetically engineered plants with modulated expression of target genes. We further propose to combine information on gene-by-gene interactions with data from physical interaction platforms such as protein-protein or TF-gene networks.

Identifying gene networks underlying the neurobiology of ethanol and alcoholism

For complex disorders such as alcoholism, identifying the genes linked to these diseases and their specific roles is difficult. Traditional genetic approaches, such as genetic association studies (including genome-wide association studies) and analyses of quantitative trait loci (QTLs) in both humans and laboratory animals already have helped identify some candidate genes. However, because of technical obstacles, such as the small impact of any individual gene, these approaches only have limited effectiveness in identifying specific genes that contribute to complex diseases. The emerging field of systems biology, which allows for analyses of entire gene networks, may help researchers better elucidate the genetic basis of alcoholism, both in humans and in animal models. Such networks can be identified using approaches such as high-throughput molecular profiling (e.g., through microarray-based gene expression analyses) or strategies referred to as genetical genomics, such as the mapping of expression QTLs (eQTLs). Characterization of gene networks can shed light on the biological pathways underlying complex traits and provide the functional context for identifying those genes that contribute to disease development.

Structural measures for network biology using QuACN

BACKGROUND

Structural measures for networks have been extensively developed, but many of them have not yet demonstrated their sustainably. That means, it remains often unclear whether a particular measure is useful and feasible to solve a particular problem in network biology. Exemplarily, the classification of complex biological networks can be named, for which structural measures are used leading to a minimal classification error. Hence, there is a strong need to provide freely available software packages to calculate and demonstrate the appropriate usage of structural graph measures in network biology.

RESULTS

Here, we discuss topological network descriptors that are implemented in the R-package QuACN and demonstrate their behavior and characteristics by applying them to a set of example graphs. Moreover, we show a representative application to illustrate their capabilities for classifying biological networks. In particular, we infer gene regulatory networks from microarray data and classify them by methods provided by QuACN. Note that QuACN is the first freely available software written in R containing a large number of structural graph measures.

CONCLUSION

The R package QuACN is under ongoing development and we add promising groups of topological network descriptors continuously. The package can be used to answer intriguing research questions in network biology, e.g., classifying biological data or identifying meaningful biological features, by analyzing the topology of biological networks.

Genetic architecture of gene expression in ovine skeletal muscle

Background

In livestock populations the genetic contribution to muscling is intensively monitored in the progeny of industry sires and used as a tool in selective breeding programs. The genes and pathways conferring this genetic merit are largely undefined. Genetic variation within a population has potential, amongst other mechanisms, to alter gene expression via cis- or trans-acting mechanisms in a manner that impacts the functional activities of specific pathways that contribute to muscling traits. By integrating sire-based genetic merit information for a muscling trait with progeny-based gene expression data we directly tested the hypothesis that there is genetic structure in the gene expression program in ovine skeletal muscle.

Results

The genetic performance of six sires for a well defined muscling trait, longissimus lumborum muscle depth, was measured using extensive progeny testing and expressed as an Estimated Breeding Value by comparison with contemporary sires. Microarray gene expression data were obtained for longissimus lumborum samples taken from forty progeny of the six sires (4-8 progeny/sire). Initial unsupervised hierarchical clustering analysis revealed strong genetic architecture to the gene expression data, which also discriminated the sire-based Estimated Breeding Value for the trait. An integrated systems biology approach was then used to identify the major functional pathways contributing to the genetics of enhanced muscling by using both Estimated Breeding Value weighted gene co-expression network analysis and a differential gene co-expression network analysis. The modules of genes revealed by these analyses were enriched for a number of functional terms summarised as muscle sarcomere organisation and development, protein catabolism (proteosome), RNA processing, mitochondrial function and transcriptional regulation.

Conclusions

This study has revealed strong genetic structure in the gene expression program within ovine longissimus lumborum muscle. The balance between muscle protein synthesis, at the levels of both transcription and translation control, and protein catabolism mediated by regulated proteolysis is likely to be the primary determinant of the genetic merit for the muscling trait in this sheep population. There is also evidence that high genetic merit for muscling is associated with a fibre type shift toward fast glycolytic fibres. This study provides insight into mechanisms, presumably subject to strong artificial selection, that underpin enhanced muscling in sheep populations.

An integrative functional genomics approach for discovering biomarkers in schizophrenia

Schizophrenia (SZ) is a complex disorder resulting from both genetic and environmental causes with a lifetime prevalence world-wide of 1%; however, there are no specific, sensitive and validated biomarkers for SZ. A general unifying hypothesis has been put forward that disease-associated single nucleotide polymorphisms (SNPs) from genome-wide association study (GWAS) are more likely to be associated with gene expression quantitative trait loci (eQTL). We will describe this hypothesis and review primary methodology with refinements for testing this paradigmatic approach in SZ. We will describe biomarker studies of SZ and testing enrichment of SNPs that are associated both with eQTLs and existing GWAS of SZ. SZ-associated SNPs that overlap with eQTLs can be placed into gene-gene expression, protein-protein and protein-DNA interaction networks. Further, those networks can be tested by reducing/silencing the gene expression levels of critical nodes. We present pilot data to support these methods of investigation such as the use of eQTLs to annotate GWASs of SZ, which could be applied to the field of biomarker discovery. Those networks that have association with SNP markers, especially cis-regulated expression, might lead to a more clear understanding of important candidate genes that predispose to disease and alter expression. This method has general application to many complex disorders.

Correlation network analysis applied to complex biofilm communities

The complexity of the human microbiome makes it difficult to reveal organizational principles of the community and even more challenging to generate testable hypotheses. It has been suggested that in the gut microbiome species such as Bacteroides thetaiotaomicron are keystone in maintaining the stability and functional adaptability of the microbial community. In this study, we investigate the interspecies associations in a complex microbial biofilm applying systems biology principles. Using correlation network analysis we identified bacterial modules that represent important microbial associations within the oral community. We used dental plaque as a model community because of its high diversity and the well known species-species interactions that are common in the oral biofilm. We analyzed samples from healthy individuals as well as from patients with periodontitis, a polymicrobial disease. Using results obtained by checkerboard hybridization on cultivable bacteria we identified modules that correlated well with microbial complexes previously described. Furthermore, we extended our analysis using the Human Oral Microbe Identification Microarray (HOMIM), which includes a large number of bacterial species, among them uncultivated organisms present in the mouth. Two distinct microbial communities appeared in healthy individuals while there was one major type in disease. Bacterial modules in all communities did not overlap, indicating that bacteria were able to effectively re-associate with new partners depending on the environmental conditions. We then identified hubs that could act as keystone species in the bacterial modules. Based on those results we then cultured a not-yet-cultivated microorganism, Tannerella sp. OT286 (clone BU063). After two rounds of enrichment by a selected helper (Prevotella oris OT311) we obtained colonies of Tannerella sp. OT286 growing on blood agar plates. This system-level approach would open the possibility of manipulating microbial communities in a targeted fashion as well as associating certain bacterial modules to clinical traits (e.g.: obesity, Crohn's disease, periodontal disease, etc).

Liverome: a curated database of liver cancer-related gene signatures with self-contained context information

Background

Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide. A number of molecular profiling studies have investigated the changes in gene and protein expression that are associated with various clinicopathological characteristics of HCC and generated a wealth of scattered information, usually in the form of gene signature tables. A database of the published HCC gene signatures would be useful to liver cancer researchers seeking to retrieve existing differential expression information on a candidate gene and to make comparisons between signatures for prioritization of common genes. A challenge in constructing such database is that a direct import of the signatures as appeared in articles would lead to a loss or ambiguity of their context information that is essential for a correct biological interpretation of a gene’s expression change. This challenge arises because designation of compared sample groups is most often abbreviated, ad hoc, or even missing from published signature tables. Without manual curation, the context information becomes lost, leading to uninformative database contents. Although several databases of gene signatures are available, none of them contains informative form of signatures nor shows comprehensive coverage on liver cancer. Thus we constructed Liverome, a curated database of liver cancer-related gene signatures with self-contained context information.

Description

Liverome’s data coverage is more than three times larger than any other signature database, consisting of 143 signatures taken from 98 HCC studies, mostly microarray and proteome, and involving 6,927 genes. The signatures were post-processed into an informative and uniform representation and annotated with an itemized summary so that all context information is unambiguously self-contained within the database. The signatures were further informatively named and meaningfully organized according to ten functional categories for guided browsing. Its web interface enables a straightforward retrieval of known differential expression information on a query gene and a comparison of signatures to prioritize common genes. The utility of Liverome-collected data is shown by case studies in which useful biological insights on HCC are produced.

Conclusion

Liverome database provides a comprehensive collection of well-curated HCC gene signatures and straightforward interfaces for gene search and signature comparison as well. Liverome is available at http://liverome.kobic.re.kr.

Interferon regulatory factor 7 is a major hub connecting interferon-mediated responses in virus-induced asthma exacerbations in vivo

BACKGROUND

Exacerbations are responsible for a substantial burden of morbidity and health care use in children with asthma. Most asthma exacerbations are triggered by viral infections; however, the underlying mechanisms have not been systematically investigated.

OBJECTIVE

The objective of this study was to elucidate the molecular networks that underpin virus-induced exacerbations in asthmatic children in vivo.

METHODS

We followed exacerbation-prone asthmatic children prospectively and profiled global patterns of gene expression in nasal lavage samples obtained during an acute, moderate, picornavirus-induced exacerbation and 7 to 14 days later. Coexpression network analysis and prior knowledge was used to reconstruct the underlying gene networks.

RESULTS

The data showed that an intricate modular program consisting of more than 1000 genes was upregulated during acute exacerbations in comparison with 7 to 14 days later. The modules were enriched for coherent cellular processes, including interferon-induced antiviral responses, innate pathogen sensing, response to wounding, small nucleolar RNAs, and the ubiquitin-proteosome and lysosome degradation pathways. Reconstruction of the wiring diagram of the modules revealed the presence of hyperconnected hub nodes, most notably interferon regulatory factor 7, which was identified as a major hub linking interferon-mediated antiviral responses.

CONCLUSIONS

This study provides an integrated view of the inflammatory networks that are upregulated during virus-induced asthma exacerbations in vivo. A series of innate signaling hubs were identified that could be novel therapeutic targets for asthma attacks.

Gene networks and haloperidol-induced catalepsy

The current study examined the changes in striatal gene network structure induced by short-term selective breeding from a heterogeneous stock for haloperidol response. Brain (striatum) gene expression data were obtained using the Illumina WG 8.2 array, and the datasets from responding and non-responding selected lines were independently interrogated using a weighted gene coexpression network analysis (WGCNA). We detected several gene modules (groups of coexpressed genes) in each dataset; the membership of the modules was found to be largely concordant, and a consensus network was constructed. Further validation of the network topology showed that using approximately 35 samples is sufficient to reliably infer the transcriptome network. An in-depth analysis showed significant changes in network structure and gene connectivity associated with the selected lines; these changes were validated using a bootstrapping procedure. The most dramatic changes were associated with a gene module richly annotated with neurobehavioral traits. The changes in network connectivity were concentrated in the links between this module and the rest of the network, in addition to changes within the module; this observation is consistent with recent results in protein and metabolic networks. These results suggest that a network-based strategy will help identify the genetic factors associated with haloperidol response.

The transcriptome of the reference potato genome Solanum tuberosum Group Phureja clone DM1-3 516R44

Advances in molecular breeding in potato have been limited by its complex biological system, which includes vegetative propagation, autotetraploidy, and extreme heterozygosity. The availability of the potato genome and accompanying gene complement with corresponding gene structure, location, and functional annotation are powerful resources for understanding this complex plant and advancing molecular breeding efforts. Here, we report a reference for the potato transcriptome using 32 tissues and growth conditions from the doubled monoploid Solanum tuberosum Group Phureja clone DM1-3 516R44 for which a genome sequence is available. Analysis of greater than 550 million RNA-Seq reads permitted the detection and quantification of expression levels of over 22,000 genes. Hierarchical clustering and principal component analyses captured the biological variability that accounts for gene expression differences among tissues suggesting tissue-specific gene expression, and genes with tissue or condition restricted expression. Using gene co-expression network analysis, we identified 18 gene modules that represent tissue-specific transcriptional networks of major potato organs and developmental stages. This information provides a powerful resource for potato research as well as studies on other members of the Solanaceae family.

Analysis of transcriptome complexity through RNA sequencing in normal and failing murine hearts

RATIONALE

Accurate and comprehensive de novo transcriptome profiling in heart is a central issue to better understand cardiac physiology and diseases. Although significant progress has been made in genome-wide profiling for quantitative changes in cardiac gene expression, current knowledge offers limited insights to the total complexity in cardiac transcriptome at individual exon level.

OBJECTIVE

To develop more robust bioinformatic approaches to analyze high-throughput RNA sequencing (RNA-Seq) data, with the focus on the investigation of transcriptome complexity at individual exon and transcript levels.

METHODS AND RESULTS

In addition to overall gene expression analysis, the methods developed in this study were used to analyze RNA-Seq data with respect to individual transcript isoforms, novel spliced exons, novel alternative terminal exons, novel transcript clusters (ie, novel genes), and long noncoding RNA genes. We applied these approaches to RNA-Seq data obtained from mouse hearts after pressure-overload-induced by transaortic constriction. Based on experimental validations, analyses of the features of the identified exons/transcripts, and expression analyses including previously published RNA-Seq data, we demonstrate that the methods are highly effective in detecting and quantifying individual exons and transcripts. Novel insights inferred from the examined aspects of the cardiac transcriptome open ways to further experimental investigations.

CONCLUSIONS

Our work provided a comprehensive set of methods to analyze mouse cardiac transcriptome complexity at individual exon and transcript levels. Applications of the methods may infer important new insights to gene regulation in normal and disease hearts in terms of exon utilization and potential involvement of novel components of cardiac transcriptome.

Weighted correlation network analysis (WGCNA) applied to the tomato fruit metabolome

BACKGROUND

Advances in "omics" technologies have revolutionized the collection of biological data. A matching revolution in our understanding of biological systems, however, will only be realized when similar advances are made in informatic analysis of the resulting "big data." Here, we compare the capabilities of three conventional and novel statistical approaches to summarize and decipher the tomato metabolome.

METHODOLOGY

Principal component analysis (PCA), batch learning self-organizing maps (BL-SOM) and weighted gene co-expression network analysis (WGCNA) were applied to a multivariate NMR dataset collected from developmentally staged tomato fruits belonging to several genotypes. While PCA and BL-SOM are appropriate and commonly used methods, WGCNA holds several advantages in the analysis of highly multivariate, complex data.

CONCLUSIONS

PCA separated the two major genetic backgrounds (AC and NC), but provided little further information. Both BL-SOM and WGCNA clustered metabolites by expression, but WGCNA additionally defined "modules" of co-expressed metabolites explicitly and provided additional network statistics that described the systems properties of the tomato metabolic network. Our first application of WGCNA to tomato metabolomics data identified three major modules of metabolites that were associated with ripening-related traits and genetic background.

High-density kinetic analysis of the metabolomic and transcriptomic response of Arabidopsis to eight environmental conditions

The time-resolved response of Arabidopsis thaliana towards changing light and/or temperature at the transcriptome and metabolome level is presented. Plants grown at 21°C with a light intensity of 150 μE m⁻² sec⁻¹ were either kept at this condition or transferred into seven different environments (4°C, darkness; 21°C, darkness; 32°C, darkness; 4°C, 85 μE m⁻² sec⁻¹; 21 °C, 75 μE m⁻² sec⁻¹; 21°C, 300 μE m⁻² sec⁻¹ ; 32°C, 150 μE m⁻² sec⁻¹). Samples were taken before (0 min) and at 22 time points after transfer resulting in (8×) 22 time points covering both a linear and a logarithmic time series totaling 177 states. Hierarchical cluster analysis shows that individual conditions (defined by temperature and light) diverge into distinct trajectories at condition-dependent times and that the metabolome follows different kinetics from the transcriptome. The metabolic responses are initially relatively faster when compared with the transcriptional responses. Gene Ontology over-representation analysis identifies a common response for all changed conditions at the transcriptome level during the early response phase (5-60 min). Metabolic networks reconstructed via metabolite-metabolite correlations reveal extensive environment-specific rewiring. Detailed analysis identifies conditional connections between amino acids and intermediates of the tricarboxylic acid cycle. Parallel analysis of transcriptional changes strongly support a model where in the absence of photosynthesis at normal/high temperatures protein degradation occurs rapidly and subsequent amino acid catabolism serves as the main cellular energy supply. These results thus demonstrate the engagement of the electron transfer flavoprotein system under short-term environmental perturbations.

Comparative physiology and transcriptional networks underlying the heat shock response in Populus trichocarpa, Arabidopsis thaliana and Glycine max

The heat shock response continues to be layered with additional complexity as interactions and crosstalk among heat shock proteins (HSPs), the reactive oxygen network and hormonal signalling are discovered. However, comparative analyses exploring variation in each of these processes among species remain relatively unexplored. In controlled environment experiments, photosynthetic response curves were conducted from 22 to 42 °C and indicated that temperature optimum of light-saturated photosynthesis was greater for Glycine max relative to Arabidopsis thaliana or Populus trichocarpa. Transcript profiles were taken at defined states along the temperature response curves, and inferred pathway analysis revealed species-specific variation in the abiotic stress and the minor carbohydrate raffinose/galactinol pathways. A weighted gene co-expression network approach was used to group individual genes into network modules linking biochemical measures of the antioxidant system to leaf-level photosynthesis among P. trichocarpa, G. max and A. thaliana. Network-enabled results revealed an expansion in the G. max HSP17 protein family and divergence in the regulation of the antioxidant and heat shock modules relative to P. trichocarpa and A. thaliana. These results indicate that although the heat shock response is highly conserved, there is considerable species-specific variation in its regulation.

Host regulatory network response to infection with highly pathogenic H5N1 avian influenza virus

During the last decade, more than half of humans infected with highly pathogenic avian influenza (HPAI) H5N1 viruses have died, yet virus-induced host signaling has yet to be clearly elucidated. Airway epithelia are known to produce inflammatory mediators that contribute to HPAI H5N1-mediated pathogenicity, but a comprehensive analysis of the host response in this cell type is lacking. Here, we leveraged a system approach to identify and statistically validate signaling subnetworks that define the dynamic transcriptional response of human bronchial epithelial cells after infection with influenza A/Vietnam/1203/2004 (H5N1, VN1203). Importantly, we validated a subset of transcripts from one subnetwork in both Calu-3 cells and mice. A more detailed examination of two subnetworks involved in the immune response and keratinization processes revealed potential novel mediators of HPAI H5N1 pathogenesis and host response signaling. Finally, we show how these results compare to those for a less virulent strain of influenza virus. Using emergent network properties, we provide fresh insight into the host response to HPAI H5N1 virus infection and identify novel avenues for perturbation studies and potential therapeutic interventions for fatal HPAI H5N1 disease.

Strategies for aggregating gene expression data: the collapseRows R function

Background

Genomic and other high dimensional analyses often require one to summarize multiple related variables by a single representative. This task is also variously referred to as collapsing, combining, reducing, or aggregating variables. Examples include summarizing several probe measurements corresponding to a single gene, representing the expression profiles of a co-expression module by a single expression profile, and aggregating cell-type marker information to de-convolute expression data. Several standard statistical summary techniques can be used, but network methods also provide useful alternative methods to find representatives. Currently few collapsing functions are developed and widely applied.

Results

We introduce the R function collapseRows that implements several collapsing methods and evaluate its performance in three applications. First, we study a crucial step of the meta-analysis of microarray data: the merging of independent gene expression data sets, which may have been measured on different platforms. Toward this end, we collapse multiple microarray probes for a single gene and then merge the data by gene identifier. We find that choosing the probe with the highest average expression leads to best between-study consistency. Second, we study methods for summarizing the gene expression profiles of a co-expression module. Several gene co-expression network analysis applications show that the optimal collapsing strategy depends on the analysis goal. Third, we study aggregating the information of cell type marker genes when the aim is to predict the abundance of cell types in a tissue sample based on gene expression data ("expression deconvolution"). We apply different collapsing methods to predict cell type abundances in peripheral human blood and in mixtures of blood cell lines. Interestingly, the most accurate prediction method involves choosing the most highly connected "hub" marker gene. Finally, to facilitate biological interpretation of collapsed gene lists, we introduce the function userListEnrichment, which assesses the enrichment of gene lists for known brain and blood cell type markers, and for other published biological pathways.

Conclusions

The R function collapseRows implements several standard and network-based collapsing methods. In various genomic applications we provide evidence that both types of methods are robust and biologically relevant tools.

Integrative network biology: graph prototyping for co-expression cancer networks

Network-based analysis has been proven useful in biologically-oriented areas, e.g., to explore the dynamics and complexity of biological networks. Investigating a set of networks allows deriving general knowledge about the underlying topological and functional properties. The integrative analysis of networks typically combines networks from different studies that investigate the same or similar research questions. In order to perform an integrative analysis it is often necessary to compare the properties of matching edges across the data set. This identification of common edges is often burdensome and computational intensive. Here, we present an approach that is different from inferring a new network based on common features. Instead, we select one network as a graph prototype, which then represents a set of comparable network objects, as it has the least average distance to all other networks in the same set. We demonstrate the usefulness of the graph prototyping approach on a set of prostate cancer networks and a set of corresponding benign networks. We further show that the distances within the cancer group and the benign group are statistically different depending on the utilized distance measure.

Association between plasma metabolites and gene expression profiles in five porcine endocrine tissues

Background

Endocrine tissues play a fundamental role in maintaining homeostasis of plasma metabolites such as non-esterified fatty acids and glucose, the levels of which reflect the energy balance or the health status of animals. However, the relationship between the transcriptome of endocrine tissues and plasma metabolites has been poorly studied.

Methods

We determined the blood levels of 12 plasma metabolites in 27 pigs belonging to five breeds, each breed consisting of both females and males. The transcriptome of five endocrine tissues i.e. hypothalamus, adenohypophysis, thyroid gland, gonads and backfat tissues from 16 out of the 27 pigs was also determined. Sex and breed effects on the 12 plasma metabolites were investigated and associations between genes expressed in the five endocrine tissues and the 12 plasma metabolites measured were analyzed. A probeset was defined as a quantitative trait transcript (QTT) when its association with a particular metabolic trait achieved a nominal P value < 0.01.

Results

A larger than expected number of QTT was found for non-esterified fatty acids and alanine aminotransferase in at least two tissues. The associations were highly tissue-specific. The QTT within the tissues were divided into co-expression network modules enriched for genes in Kyoto Encyclopedia of Genes and Genomes or gene ontology categories that are related to the physiological functions of the corresponding tissues. We also explored a multi-tissue co-expression network using QTT for non-esterified fatty acids from the five tissues and found that a module, enriched in hypothalamus QTT, was positioned at the centre of the entire multi-tissue network.

Conclusions

These results emphasize the relationships between endocrine tissues and plasma metabolites in terms of gene expression. Highly tissue-specific association patterns suggest that candidate genes or gene pathways should be investigated in the context of specific tissues.

A systems genetic analysis of high density lipoprotein metabolism and network preservation across mouse models

We report a systems genetic analysis of high density lipoprotein (HDL) levels in an F2 intercross between inbred strains CAST/EiJ and C57BL/6J. We previously showed that there are dramatic differences in HDL metabolism in a cross between these strains, and we now report co-expression network analysis of HDL that integrates global expression data from liver and adipose with relevant metabolic traits. Using data from a total of 293 F2 intercross mice, we constructed weighted gene co-expression networks and identified modules (subnetworks) associated with HDL and clinical traits. These were examined for genes implicated in HDL levels based on large human genome-wide associations studies (GWAS) and examined with respect to conservation between tissue and sexes in a total of 9 data sets. We identify genes that are consistently ranked high by association with HDL across the 9 data sets. We focus in particular on two genes, Wfdc2 and Hdac3, that are located in close proximity to HDL QTL peaks where causal testing indicates that they may affect HDL. Our results provide a rich resource for studies of complex metabolic interactions involving HDL. This article is part of a Special Issue entitled Advances in High Density Lipoprotein Formation and Metabolism: A Tribute to John F. Oram (1945-2010).

Gene coexpression network analysis as a source of functional annotation for rice genes

With the existence of large publicly available plant gene expression data sets, many groups have undertaken data analyses to construct gene coexpression networks and functionally annotate genes. Often, a large compendium of unrelated or condition-independent expression data is used to construct gene networks. Condition-dependent expression experiments consisting of well-defined conditions/treatments have also been used to create coexpression networks to help examine particular biological processes. Gene networks derived from either condition-dependent or condition-independent data can be difficult to interpret if a large number of genes and connections are present. However, algorithms exist to identify modules of highly connected and biologically relevant genes within coexpression networks. In this study, we have used publicly available rice (Oryza sativa) gene expression data to create gene coexpression networks using both condition-dependent and condition-independent data and have identified gene modules within these networks using the Weighted Gene Coexpression Network Analysis method. We compared the number of genes assigned to modules and the biological interpretability of gene coexpression modules to assess the utility of condition-dependent and condition-independent gene coexpression networks. For the purpose of providing functional annotation to rice genes, we found that gene modules identified by coexpression analysis of condition-dependent gene expression experiments to be more useful than gene modules identified by analysis of a condition-independent data set. We have incorporated our results into the MSU Rice Genome Annotation Project database as additional expression-based annotation for 13,537 genes, 2,980 of which lack a functional annotation description. These results provide two new types of functional annotation for our database. Genes in modules are now associated with groups of genes that constitute a collective functional annotation of those modules. Additionally, the expression patterns of genes across the treatments/conditions of an expression experiment comprise a second form of useful annotation.

Differential C3NET reveals disease networks of direct physical interactions

BACKGROUND

Genes might have different gene interactions in different cell conditions, which might be mapped into different networks. Differential analysis of gene networks allows spotting condition-specific interactions that, for instance, form disease networks if the conditions are a disease, such as cancer, and normal. This could potentially allow developing better and subtly targeted drugs to cure cancer. Differential network analysis with direct physical gene interactions needs to be explored in this endeavour.

RESULTS

C3NET is a recently introduced information theory based gene network inference algorithm that infers direct physical gene interactions from expression data, which was shown to give consistently higher inference performances over various networks than its competitors. In this paper, we present, DC3net, an approach to employ C3NET in inferring disease networks. We apply DC3net on a synthetic and real prostate cancer datasets, which show promising results. With loose cutoffs, we predicted 18583 interactions from tumor and normal samples in total. Although there are no reference interactions databases for the specific conditions of our samples in the literature, we found verifications for 54 of our predicted direct physical interactions from only four of the biological interaction databases. As an example, we predicted that RAD50 with TRF2 have prostate cancer specific interaction that turned out to be having validation from the literature. It is known that RAD50 complex associates with TRF2 in the S phase of cell cycle, which suggests that this predicted interaction may promote telomere maintenance in tumor cells in order to allow tumor cells to divide indefinitely. Our enrichment analysis suggests that the identified tumor specific gene interactions may be potentially important in driving the growth in prostate cancer. Additionally, we found that the highest connected subnetwork of our predicted tumor specific network is enriched for all proliferation genes, which further suggests that the genes in this network may serve in the process of oncogenesis.

CONCLUSIONS

Our approach reveals disease specific interactions. It may help to make experimental follow-up studies more cost and time efficient by prioritizing disease relevant parts of the global gene network.

Network for activation of human endothelial cells by oxidized phospholipids: a critical role of heme oxygenase 1

RATIONALE

Oxidized palmitoyl arachidonyl phosphatidylcholine (Ox-PAPC) accumulates in atherosclerotic lesions, is proatherogenic, and influences the expression of more than 1000 genes in endothelial cells.

OBJECTIVE

To elucidate the major pathways involved in Ox-PAPC action, we conducted a systems analysis of endothelial cell gene expression after exposure to Ox-PAPC.

METHODS AND RESULTS

We used the variable responses of primary endothelial cells from 149 individuals exposed to Ox-PAPC to construct a network that consisted of 11 groups of genes, or modules. Modules were enriched for a broad range of Gene Ontology pathways, some of which have not been identified previously as major Ox-PAPC targets. Further validating our method of network construction, modules were consistent with relationships established by cell biology studies of Ox-PAPC effects on endothelial cells. This network provides novel hypotheses about molecular interactions, as well as candidate molecular regulators of inflammation and atherosclerosis. We validated several hypotheses based on network connections and genomic association. Our network analysis predicted that the hub gene CHAC1 (cation transport regulator homolog 1) was regulated by the ATF4 (activating transcription factor 4) arm of the unfolded protein response pathway, and here we showed that ATF4 directly activates an element in the CHAC1 promoter. We showed that variation in basal levels of heme oxygenase 1 (HMOX1) contribute to the response to Ox-PAPC, consistent with its position as a hub in our network. We also identified G-protein-coupled receptor 39 (GPR39) as a regulator of HMOX1 levels and showed that it modulates the promoter activity of HMOX1. We further showed that OKL38/OSGN1 (oxidative stress-induced growth inhibitor), the hub gene in the blue module, is a key regulator of both inflammatory and antiinflammatory molecules.

CONCLUSIONS

Our systems genetics approach has provided a broad view of the pathways involved in the response of endothelial cells to Ox-PAPC and also identified novel regulatory mechanisms.

Gene coexpression network alignment and conservation of gene modules between two grass species: maize and rice

One major objective for plant biology is the discovery of molecular subsystems underlying complex traits. The use of genetic and genomic resources combined in a systems genetics approach offers a means for approaching this goal. This study describes a maize (Zea mays) gene coexpression network built from publicly available expression arrays. The maize network consisted of 2,071 loci that were divided into 34 distinct modules that contained 1,928 enriched functional annotation terms and 35 cofunctional gene clusters. Of note, 391 maize genes of unknown function were found to be coexpressed within modules along with genes of known function. A global network alignment was made between this maize network and a previously described rice (Oryza sativa) coexpression network. The IsoRankN tool was used, which incorporates both gene homology and network topology for the alignment. A total of 1,173 aligned loci were detected between the two grass networks, which condensed into 154 conserved subgraphs that preserved 4,758 coexpression edges in rice and 6,105 coexpression edges in maize. This study provides an early view into maize coexpression space and provides an initial network-based framework for the translation of functional genomic and genetic information between these two vital agricultural species.

Epigenetic predictor of age

From the moment of conception, we begin to age. A decay of cellular structures, gene regulation, and DNA sequence ages cells and organisms. DNA methylation patterns change with increasing age and contribute to age related disease. Here we identify 88 sites in or near 80 genes for which the degree of cytosine methylation is significantly correlated with age in saliva of 34 male identical twin pairs between 21 and 55 years of age. Furthermore, we validated sites in the promoters of three genes and replicated our results in a general population sample of 31 males and 29 females between 18 and 70 years of age. The methylation of three sites—in the promoters of the EDARADD, TOM1L1, and NPTX2 genes—is linear with age over a range of five decades. Using just two cytosines from these loci, we built a regression model that explained 73% of the variance in age, and is able to predict the age of an individual with an average accuracy of 5.2 years. In forensic science, such a model could estimate the age of a person, based on a biological sample alone. Furthermore, a measurement of relevant sites in the genome could be a tool in routine medical screening to predict the risk of age-related diseases and to tailor interventions based on the epigenetic bio-age instead of the chronological age.

Functional modules distinguish human induced pluripotent stem cells from embryonic stem cells

It has been debated whether human induced pluripotent stem cells (iPSCs) and embryonic stem cells (ESCs) express distinctive transcriptomes. By using the method of weighted gene co-expression network analysis, we showed here that iPSCs exhibit altered functional modules compared with ESCs. Notably, iPSCs and ESCs differentially express 17 modules that primarily function in transcription, metabolism, development, and immune response. These module activations (up- and downregulation) are highly conserved in a variety of iPSCs, and genes in each module are coherently co-expressed. Furthermore, the activation levels of these modular genes can be used as quantitative variables to discriminate iPSCs and ESCs with high accuracy (96%). Thus, differential activations of these functional modules are the conserved features distinguishing iPSCs from ESCs. Strikingly, the overall activation level of these modules is inversely correlated with the DNA methylation level, suggesting that DNA methylation may be one mechanism regulating the module differences. Overall, we conclude that human iPSCs and ESCs exhibit distinct gene expression networks, which are likely associated with different epigenetic reprogramming events during the derivation of iPSCs and ESCs.

Gene expression profiling of the brain: pondering facts and fiction

During the last decade brain transcriptome profiling by DNA microarrays has matured, developed sound experimental design standards, reporting practices, analytical procedures, and data sharing resources. It has become a powerful scientific tool in the exploratory research portfolio. Along this journey by trial and error, we encountered a number of intriguing questions and comments--pondering the value of hypothesis-driven research, appropriate sample size, the importance and interpretation of transcripts changes vis-à-vis protein changes, the role of statistical stringency, false discovery and magnitude of expression change, and many other interesting questions. Our field fully acknowledges and tries to address all of these challenges associated with high-throughput, data-driven transcriptomics. As a research field, we strongly advocate implementing the highest standards of our trade, and we deeply believe that transcriptome profiling studies will continue to be essential for deciphering the pathophysiological mechanisms leading to complex brain disorders.

Enhancing biological relevance of a weighted gene co-expression network for functional module identification

Relationships among gene expression levels may be associated with the mechanisms of the disease. While identifying a direct association such as a difference in expression levels between case and control groups links genes to disease mechanisms, uncovering an indirect association in the form of a network structure may help reveal the underlying functional module associated with the disease under scrutiny. This paper presents a method to improve the biological relevance in functional module identification from the gene expression microarray data by enhancing the structure of a weighted gene co-expression network using minimum spanning tree. The enhanced network, which is called a backbone network, contains only the essential structural information to represent the gene co-expression network. The entire backbone network is decoupled into a number of coherent sub-networks, and then the functional modules are reconstructed from these sub-networks to ensure minimum redundancy. The method was tested with a simulated gene expression dataset and case-control expression datasets of autism spectrum disorder and colorectal cancer studies. The results indicate that the proposed method can accurately identify clusters in the simulated dataset, and the functional modules of the backbone network are more biologically relevant than those obtained from the original approach.

Comparative analysis of proteome and transcriptome variation in mouse

The relationships between the levels of transcripts and the levels of the proteins they encode have not been examined comprehensively in mammals, although previous work in plants and yeast suggest a surprisingly modest correlation. We have examined this issue using a genetic approach in which natural variations were used to perturb both transcript levels and protein levels among inbred strains of mice. We quantified over 5,000 peptides and over 22,000 transcripts in livers of 97 inbred and recombinant inbred strains and focused on the 7,185 most heritable transcripts and 486 most reliable proteins. The transcript levels were quantified by microarray analysis in three replicates and the proteins were quantified by Liquid Chromatography–Mass Spectrometry using O(18)-reference-based isotope labeling approach. We show that the levels of transcripts and proteins correlate significantly for only about half of the genes tested, with an average correlation of 0.27, and the correlations of transcripts and proteins varied depending on the cellular location and biological function of the gene. We examined technical and biological factors that could contribute to the modest correlation. For example, differential splicing clearly affects the analyses for certain genes; but, based on deep sequencing, this does not substantially contribute to the overall estimate of the correlation. We also employed genome-wide association analyses to map loci controlling both transcript and protein levels. Surprisingly, little overlap was observed between the protein- and transcript-mapped loci. We have typed numerous clinically relevant traits among the strains, including adiposity, lipoprotein levels, and tissue parameters. Using correlation analysis, we found that a low number of clinical trait relationships are preserved between the protein and mRNA gene products and that the majority of such relationships are specific to either the protein levels or transcript levels. Surprisingly, transcript levels were more strongly correlated with clinical traits than protein levels. In light of the widespread use of high-throughput technologies in both clinical and basic research, the results presented have practical as well as basic implications.

An old dogma in biology states that, in every cell, the flow of biological information is from DNA to RNA to proteins and that the latter act as a working force to determine the organism's phenotype. This model predicts that changes in DNA that affect the clinical phenotype should also similarly change the cellular levels of RNA and protein levels. In this report, we test this prediction by looking at the concordance between DNA variation in population of mouse inbred strains, the RNA and protein variation in the liver tissue of these mice, and variation in metabolic phenotypes. We show that the relationship between various biological traits is not simple and that there is relatively little concordance of RNA levels and the corresponding protein levels in response to DNA perturbations. In addition, we also find that, surprisingly, metabolic traits correlate better to RNA levels than to protein levels. In light of current efforts in searching for the molecular bases of disease susceptibility in humans, our findings highlight the complexity of information flow that underlies clinical outcomes.

Protein expression based multimarker analysis of breast cancer samples

Background

Tissue microarray (TMA) data are commonly used to validate the prognostic accuracy of tumor markers. For example, breast cancer TMA data have led to the identification of several promising prognostic markers of survival time. Several studies have shown that TMA data can also be used to cluster patients into clinically distinct groups. Here we use breast cancer TMA data to cluster patients into distinct prognostic groups.

Methods

We apply weighted correlation network analysis (WGCNA) to TMA data consisting of 26 putative tumor biomarkers measured on 82 breast cancer patients. Based on this analysis we identify three groups of patients with low (5.4%), moderate (22%) and high (50%) mortality rates, respectively. We then develop a simple threshold rule using a subset of three markers (p53, Na-KATPase-β1, and TGF β receptor II) that can approximately define these mortality groups. We compare the results of this correlation network analysis with results from a standard Cox regression analysis.

Results

We find that the rule-based grouping variable (referred to as WGCNA*) is an independent predictor of survival time. While WGCNA* is based on protein measurements (TMA data), it validated in two independent Affymetrix microarray gene expression data (which measure mRNA abundance). We find that the WGCNA patient groups differed by 35% from mortality groups defined by a more conventional stepwise Cox regression analysis approach.

Conclusions

We show that correlation network methods, which are primarily used to analyze the relationships between gene products, are also useful for analyzing the relationships between patients and for defining distinct patient groups based on TMA data. We identify a rule based on three tumor markers for predicting breast cancer survival outcomes.

Hippocampal gene expression analysis highlights Ly6a/Sca-1 as candidate gene for previously mapped novelty induced behaviors in mice

In this study, we show that the covariance between behavior and gene expression in the brain can help further unravel the determinants of neurobehavioral traits. Previously, a QTL for novelty induced motor activity levels was identified on murine chromosome 15 using consomic strains. With the goal of narrowing down the linked region and possibly identifying the gene underlying the quantitative trait, gene expression data from this F₂-population was collected and used for expression QTL analysis. While genetic variation in these mice was limited to chromosome 15, eQTL analysis of gene expression showed strong cis-effects as well as trans-effects elsewhere in the genome. Using weighted gene co-expression network analysis, we were able to identify modules of co-expressed genes related to novelty induced motor activity levels. In eQTL analyses, the expression of Ly6a (a.k.a. Sca-1) was found to be cis-regulated by chromosome 15. Ly6a also surfaced in a group of genes resulting from the network analysis that was correlated with behavior. Behavioral analysis of Ly6a knock-out mice revealed reduced novelty induced motor activity levels when compared to wild type controls, confirming functional importance of Ly6a in this behavior, possibly through regulating other genes in a pathway. This study shows that gene expression profiling can be used to narrow down a previously identified behavioral QTL in mice, providing support for Ly6a as a candidate gene for functional involvement in novelty responsiveness.

Nonlinear gene cluster analysis with labeling for microarray gene expression data in organ development

Background

The gene networks underlying closure of the optic fissure during vertebrate eye development are not well-understood. We use a novel clustering method based on nonlinear dimension reduction with data labeling to analyze microarray data from laser capture microdissected (LCM) cells at the site and developmental stages (days 10.5 to 12.5) of optic fissure closure.

Results

Our nonlinear methods created clusters of genes that mapped onto more specific biological processes and functions related to eye development as defined by Gene Ontology at lower false discovery rates than conventional linear cluster algorithms. Our new methods build on the advantages of LCM to isolate pure phenotypic populations within complex tissues in order to identify systems biology relationships among critical gene products expressed at lower copy number.

Conclusions

The combination of LCM of embryonic organs, gene expression microarrays, and nonlinear dimension reduction with labeling is a potentially useful approach to extract subtle spatial and temporal co-variations within the gene regulatory networks that specify mammalian organogenesis and organ function. Our results motivate further analysis of nonlinear dimension reduction with labeling within other microarray data sets from LCM dissected tissues or other cell specific samples to determine the more general utility of our method for uncovering more specific biological functional relationships.

Transcriptome instability in colorectal cancer identified by exon microarray analyses: Associations with splicing factor expression levels and patient survival

Background

Colorectal cancer (CRC) is a heterogeneous disease that, on the molecular level, can be characterized by inherent genomic instabilities; chromosome instability and microsatellite instability. In the present study we analyze genome-wide disruption of pre-mRNA splicing, and propose transcriptome instability as a characteristic that is analogous to genomic instability on the transcriptome level.

Methods

Exon microarray profiles from two independent series including a total of 160 CRCs were investigated for their relative amounts of exon usage differences. Each exon in each sample was assigned an alternative splicing score calculated by the FIRMA algorithm. Amounts of deviating exon usage per sample were derived from exons with extreme splicing scores.

Results

There was great heterogeneity within both series in terms of sample-wise amounts of deviating exon usage. This was strongly associated with the expression levels of approximately half of 280 splicing factors (54% and 48% of splicing factors were significantly correlated to deviating exon usage amounts in the two series). Samples with high or low amounts of deviating exon usage, associated with overall transcriptome instability, were almost completely separated into their respective groups by hierarchical clustering analysis of splicing factor expression levels in both sample series. Samples showing a preferential tendency towards deviating exon skipping or inclusion were associated with skewed transcriptome instability. There were significant associations between transcriptome instability and reduced patient survival in both sample series. In the test series, patients with skewed transcriptome instability showed the strongest prognostic association (P = 0.001), while a combination of the two characteristics showed the strongest association with poor survival in the validation series (P = 0.03).

Conclusions

We have described transcriptome instability as a characteristic of CRC. This transcriptome instability has associations with splicing factor expression levels and poor patient survival.

Transcriptomic analysis of autistic brain reveals convergent molecular pathology

Autism spectrum disorder (ASD) is a common, highly heritable neuro-developmental condition characterized by marked genetic heterogeneity^1–3. Thus, a fundamental question is whether autism represents an etiologically heterogeneous disorder in which the myriad genetic or environmental risk factors perturb common underlying molecular pathways in the brain⁴. Here, we demonstrate consistent differences in transcriptome organization between autistic and normal brain by gene co-expression network analysis. Remarkably, regional patterns of gene expression that typically distinguish frontal and temporal cortex are significantly attenuated in the ASD brain, suggesting abnormalities in cortical patterning. We further identify discrete modules of co-expressed genes associated with autism: a neuronal module enriched for known autism susceptibility genes, including the neuronal specific splicing factor A2BP1/FOX1, and a module enriched for immune genes and glial markers. Using high-throughput RNA-sequencing we demonstrate dysregulated splicing of A2BP1-dependent alternative exons in ASD brain. Moreover, using a published autism GWAS dataset, we show that the neuronal module is enriched for genetically associated variants, providing independent support for the causal involvement of these genes in autism. In contrast, the immune-glial module showed no enrichment for autism GWAS signals, indicating a non-genetic etiology for this process. Collectively, our results provide strong evidence for convergent molecular abnormalities in ASD, and implicate transcriptional and splicing dysregulation as underlying mechanisms of neuronal dysfunction in this disorder.

Microarray analysis and scale-free gene networks identify candidate regulators in drought-stressed roots of loblolly pine (P. taeda L.)

BACKGROUND

Global transcriptional analysis of loblolly pine (Pinus taeda L.) is challenging due to limited molecular tools. PtGen2, a 26,496 feature cDNA microarray, was fabricated and used to assess drought-induced gene expression in loblolly pine propagule roots. Statistical analysis of differential expression and weighted gene correlation network analysis were used to identify drought-responsive genes and further characterize the molecular basis of drought tolerance in loblolly pine.

RESULTS

Microarrays were used to interrogate root cDNA populations obtained from 12 genotype × treatment combinations (four genotypes, three watering regimes). Comparison of drought-stressed roots with roots from the control treatment identified 2445 genes displaying at least a 1.5-fold expression difference (false discovery rate = 0.01). Genes commonly associated with drought response in pine and other plant species, as well as a number of abiotic and biotic stress-related genes, were up-regulated in drought-stressed roots. Only 76 genes were identified as differentially expressed in drought-recovered roots, indicating that the transcript population can return to the pre-drought state within 48 hours. Gene correlation analysis predicts a scale-free network topology and identifies eleven co-expression modules that ranged in size from 34 to 938 members. Network topological parameters identified a number of central nodes (hubs) including those with significant homology (E-values ≤ 2 × 10-30) to 9-cis-epoxycarotenoid dioxygenase, zeatin O-glucosyltransferase, and ABA-responsive protein. Identified hubs also include genes that have been associated previously with osmotic stress, phytohormones, enzymes that detoxify reactive oxygen species, and several genes of unknown function.

CONCLUSION

PtGen2 was used to evaluate transcriptome responses in loblolly pine and was leveraged to identify 2445 differentially expressed genes responding to severe drought stress in roots. Many of the genes identified are known to be up-regulated in response to osmotic stress in pine and other plant species and encode proteins involved in both signal transduction and stress tolerance. Gene expression levels returned to control values within a 48-hour recovery period in all but 76 transcripts. Correlation network analysis indicates a scale-free network topology for the pine root transcriptome and identifies central nodes that may serve as drivers of drought-responsive transcriptome dynamics in the roots of loblolly pine.

mirConnX: condition-specific mRNA-microRNA network integrator

mirConnX is a user-friendly web interface for inferring, displaying and parsing mRNA and microRNA (miRNA) gene regulatory networks. mirConnX combines sequence information with gene expression data analysis to create a disease-specific, genome-wide regulatory network. A prior, static network has been constructed for all human and mouse genes. It consists of computationally predicted transcription factor (TF)-gene associations and miRNA target predictions. The prior network is supplemented with known interactions from the literature. Dynamic TF- and miRNA-gene associations are inferred from user-provided expression data using an association measure of choice. The static and dynamic networks are then combined using an integration function with user-specified weights. Visualization of the network and subsequent analysis are provided via a very responsive graphic user interface. Two organisms are currently supported: Homo sapiens and Mus musculus. The intuitive user interface and large database make mirConnX a useful tool for clinical scientists for hypothesis generation and explorations. mirConnX is freely available for academic use at http://www.benoslab.pitt.edu/mirconnx.

Contribution of microarray data to the advancement of knowledge on the Mycobacterium tuberculosis interactome: use of the random partial least squares approach

Following the central dogma of molecular biology, where data flows from gene to protein through transcript, information on gene expression provides information on the functional state of an organism. Microarray technology arose to measure the expression level of thousands of genes simultaneously. These vast amounts of data generated at all levels of biological organization help to identify co-expressed genes, which may reveal proteins interacting in a complex or acting in the same pathway without direct physical contact. Discovering associations of regulatory patterns of characterized proteins with those of hypothetical proteins may identify functional relationships between them and facilitate the characterization of proteins of unknown function. Here we make use of the random partial least squares regression technique (r-PLS) to trace connections between co-expressed genes in Mycobacterium tuberculosis using data downloaded from public microarray databases. We generated the overall topology of a microbial co-expression network with the exact complexity of the model. This approach provides a general method for generating a co-expression network of an organism for the purpose of systems-level analyses.