WGCNA: an R package for weighted correlation network analysis

Peeling back the evolutionary layers of molecular mechanisms responsive to exercise-stress in the skeletal muscle of the racing horse

The modern horse (Equus caballus) is the product of over 50 million yrs of evolution. The athletic abilities of the horse have been enhanced during the past 6000 yrs under domestication. Therefore, the horse serves as a valuable model to understand the physiology and molecular mechanisms of adaptive responses to exercise. The structure and function of skeletal muscle show remarkable plasticity to the physical and metabolic challenges following exercise. Here, we reveal an evolutionary layer of responsiveness to exercise-stress in the skeletal muscle of the racing horse. We analysed differentially expressed genes and their co-expression networks in a large-scale RNA-sequence dataset comparing expression before and after exercise. By estimating genome-wide dN/dS ratios using six mammalian genomes, and FST and iHS using re-sequencing data derived from 20 horses, we were able to peel back the evolutionary layers of adaptations to exercise-stress in the horse. We found that the oldest and thickest layer (dN/dS) consists of system-wide tissue and organ adaptations. We further find that, during the period of horse domestication, the older layer (FST) is mainly responsible for adaptations to inflammation and energy metabolism, and the most recent layer (iHS) for neurological system process, cell adhesion, and proteolysis.

Microarray analyses of glucocorticoid and vitamin D3 target genes in differentiating cultured human podocytes

Glomerular podocytes are highly differentiated epithelial cells that are key components of the kidney filtration units. Podocyte damage or loss is the hallmark of nephritic diseases characterized by severe proteinuria. Recent studies implicate that hormones including glucocorticoids (ligand for glucocorticoid receptor) and vitamin D3 (ligand for vitamin D receptor) protect or promote repair of podocytes from injury. In order to elucidate the mechanisms underlying hormone-mediated podocyte-protecting activity from injury, we carried out microarray gene expression studies to identify the target genes and corresponding pathways in response to these hormones during podocyte differentiation. We used immortalized human cultured podocytes (HPCs) as a model system and carried out in vitro differentiation assays followed by dexamethasone (Dex) or vitamin D3 (VD3) treatment. Upon the induction of differentiation, multiple functional categories including cell cycle, organelle dynamics, mitochondrion, apoptosis and cytoskeleton organization were among the most significantly affected. Interestingly, while Dex and VD3 are capable of protecting podocytes from injury, they only share limited target genes and affected pathways. Compared to VD3 treatment, Dex had a broader and greater impact on gene expression profiles. In-depth analyses of Dex altered genes indicate that Dex crosstalks with a broad spectrum of signaling pathways, of which inflammatory responses, cell migration, angiogenesis, NF-κB and TGFβ pathways are predominantly altered. Together, our study provides new information and identifies several new avenues for future investigation of hormone signaling in podocytes.

Pig immune response to general stimulus and to porcine reproductive and respiratory syndrome virus infection: a meta-analysis approach

BACKGROUND

The availability of gene expression data that corresponds to pig immune response challenges provides compelling material for the understanding of the host immune system. Meta-analysis offers the opportunity to confirm and expand our knowledge by combining and studying at one time a vast set of independent studies creating large datasets with increased statistical power. In this study, we performed two meta-analyses of porcine transcriptomic data: i) scrutinized the global immune response to different challenges, and ii) determined the specific response to Porcine Reproductive and Respiratory Syndrome Virus (PRRSV) infection. To gain an in-depth knowledge of the pig response to PRRSV infection, we used an original approach comparing and eliminating the common genes from both meta-analyses in order to identify genes and pathways specifically involved in the PRRSV immune response. The software Pointillist was used to cope with the highly disparate data, circumventing the biases generated by the specific responses linked to single studies. Next, we used the Ingenuity Pathways Analysis (IPA) software to survey the canonical pathways, biological functions and transcription factors found to be significantly involved in the pig immune response. We used 779 chips corresponding to 29 datasets for the pig global immune response and 279 chips obtained from 6 datasets for the pig response to PRRSV infection, respectively.

RESULTS

The pig global immune response analysis showed interconnected canonical pathways involved in the regulation of translation and mitochondrial energy metabolism. Biological functions revealed in this meta-analysis were centred around translation regulation, which included protein synthesis, RNA-post transcriptional gene expression and cellular growth and proliferation. Furthermore, the oxidative phosphorylation and mitochondria dysfunctions, associated with stress signalling, were highly regulated. Transcription factors such as MYCN, MYC and NFE2L2 were found in this analysis to be potentially involved in the regulation of the immune response. The host specific response to PRRSV infection engendered the activation of well-defined canonical pathways in response to pathogen challenge such as TREM1, toll-like receptor and hyper-cytokinemia/ hyper-chemokinemia signalling. Furthermore, this analysis brought forth the central role of the crosstalk between innate and adaptive immune response and the regulation of anti-inflammatory response. The most significant transcription factor potentially involved in this analysis was HMGB1, which is required for the innate recognition of viral nucleic acids. Other transcription factors like interferon regulatory factors IRF1, IRF3, IRF5 and IRF8 were also involved in the pig specific response to PRRSV infection.

CONCLUSIONS

This work reveals key genes, canonical pathways and biological functions involved in the pig global immune response to diverse challenges, including PRRSV infection. The powerful statistical approach led us to consolidate previous findings as well as to gain new insights into the pig immune response either to common stimuli or specifically to PRRSV infection.

curatedOvarianData: clinically annotated data for the ovarian cancer transcriptome

This article introduces a manually curated data collection for gene expression meta-analysis of patients with ovarian cancer and software for reproducible preparation of similar databases. This resource provides uniformly prepared microarray data for 2970 patients from 23 studies with curated and documented clinical metadata. It allows users to efficiently identify studies and patient subgroups of interest for analysis and to perform meta-analysis immediately without the challenges posed by harmonizing heterogeneous microarray technologies, study designs, expression data processing methods and clinical data formats. We confirm that the recently proposed biomarker CXCL12 is associated with patient survival, independently of stage and optimal surgical debulking, which was possible only through meta-analysis owing to insufficient sample sizes of the individual studies. The database is implemented as the curatedOvarianData Bioconductor package for the R statistical computing language, providing a comprehensive and flexible resource for clinically oriented investigation of the ovarian cancer transcriptome. The package and pipeline for producing it are available from http://bcb.dfci.harvard.edu/ovariancancer.

Database URL:http://bcb.dfci.harvard.edu/ovariancancer

Transcriptome responses to combinations of stresses in Arabidopsis

Biotic and abiotic stresses limit agricultural yields, and plants are often simultaneously exposed to multiple stresses. Combinations of stresses such as heat and drought or cold and high light intensity have profound effects on crop performance and yields. Thus, delineation of the regulatory networks and metabolic pathways responding to single and multiple concurrent stresses is required for breeding and engineering crop stress tolerance. Many studies have described transcriptome changes in response to single stresses. However, exposure of plants to a combination of stress factors may require agonistic or antagonistic responses or responses potentially unrelated to responses to the corresponding single stresses. To analyze such responses, we initially compared transcriptome changes in 10 Arabidopsis (Arabidopsis thaliana) ecotypes using cold, heat, high-light, salt, and flagellin treatments as single stress factors as well as their double combinations. This revealed that some 61% of the transcriptome changes in response to double stresses were not predic from the responses to single stress treatments. It also showed that plants prioritized between potentially antagonistic responses for only 5% to 10% of the responding transcripts. This indicates that plants have evolved to cope with combinations of stresses and, therefore, may be bred to endure them. In addition, using a subset of this data from the Columbia and Landsberg erecta ecotypes, we have delineated coexpression network modules responding to single and combined stresses.

A developmental transcriptional network for maize defines coexpression modules

Here, we present a genome-wide overview of transcriptional circuits in the agriculturally significant crop species maize (Zea mays). We examined transcript abundance data at 50 developmental stages, from embryogenesis to senescence, for 34,876 gene models and classified genes into 24 robust coexpression modules. Modules were strongly associated with tissue types and related biological processes. Sixteen of the 24 modules (67%) have preferential transcript abundance within specific tissues. One-third of modules had an absence of gene expression in specific tissues. Genes within a number of modules also correlated with the developmental age of tissues. Coexpression of genes is likely due to transcriptional control. For a number of modules, key genes involved in transcriptional control have expression profiles that mimic the expression profiles of module genes, although the expression of transcriptional control genes is not unusually representative of module gene expression. Known regulatory motifs are enriched in several modules. Finally, of the 13 network modules with more than 200 genes, three contain genes that are notably clustered (P < 0.05) within the genome. This work, based on a carefully selected set of major tissues representing diverse stages of maize development, demonstrates the remarkable power of transcript-level coexpression networks to identify underlying biological processes and their molecular components.

Gene expression-based modeling of human cortical synaptic density

Postnatal cortical synaptic development is characterized by stages of exuberant growth, pruning, and stabilization during adulthood. How gene expression orchestrates these stages of synaptic development is poorly understood. Here we report that synaptic growth-related gene expression alone does not determine cortical synaptic density changes across the human lifespan, but instead, the dynamics of cortical synaptic density can be accurately simulated by a first-order kinetic model of synaptic growth and elimination that incorporates two separate gene expression patterns. Surprisingly, modeling of cortical synaptic density is optimized when genes related to oligodendrocytes are used to determine synaptic elimination rates. Expression of synaptic growth and oligodendrocyte genes varies regionally, resulting in different predictions of synaptic density among cortical regions that concur with previous regional data in humans. Our analysis suggests that modest rates of synaptic growth persist in adulthood, but that this is counterbalanced by increasing rates of synaptic elimination, resulting in stable synaptic number and ongoing synaptic turnover in the human adult cortex. Our approach provides a promising avenue for exploring how complex interactions among genes may contribute to neurobiological phenomena across the human lifespan.

Selection for drinking in the dark alters brain gene coexpression networks

BACKGROUND

Heterogeneous stock (HS/NPT) mice have been used to create lines selectively bred in replicate for elevated drinking in the dark (DID). Both selected lines routinely reach a blood ethanol (EtOH) concentration (BEC) of 1.00 mg/ml or greater at the end of the 4-hour period of access in Day 2. The mechanisms through which genetic differences influence DID are currently unclear. Therefore, the current study examines the transcriptome, the first stage at which genetic variability affects neurobiology. Rather than focusing solely on differential expression (DE), we also examine changes in the ways that gene transcripts collectively interact with each other, as revealed by changes in coexpression patterns.

METHODS

Naïve mice (N = 48/group) were genotyped using the Mouse Universal Genotyping Array, which provided 3,683 informative markers. Quantitative trait locus (QTL) analysis used a marker-by-marker strategy with the threshold for a significant logarithm of odds (LOD) set at 10.6. Gene expression in the ventral striatum was measured using the Illumina Mouse 8.2 array. Differential gene expression and the weighted gene coexpression network analysis (WGCNA) were implemented largely as described elsewhere.

RESULTS

Significant QTLs for elevated BECs after DID were detected on chromosomes 4, 14, and 16; the latter 2 were associated with gene-poor regions. None of the QTLs overlapped with known QTLs for EtOH preference drinking. Ninety-four transcripts were detected as being differentially expressed in both selected lines versus HS controls; there was no overlap with known preference genes. The WGCNA revealed 2 modules as showing significant effects of both selections on intramodular connectivity. A number of genes known to be associated with EtOH phenotypes (e.g., Gabrg1, Glra2, Grik1, Npy2r, and Nts) showed significant changes in connectivity.

CONCLUSIONS

We found marked and consistent effects of selection on coexpression patterns; DE changes were more modest and less concordant. The QTLs and differentially expressed genes detected here are distinct from the preference phenotype. This is consistent with behavioral data and suggests that the DID and preference phenotypes are markedly different genetically.

Gene coexpression networks reveal key drivers of phenotypic divergence in lake whitefish

A functional understanding of processes involved in adaptive divergence is one of the awaiting opportunities afforded by high-throughput transcriptomic technologies. Functional analysis of coexpressed genes has succeeded in the biomedical field in identifying key drivers of disease pathways. However, in ecology and evolutionary biology, functional interpretation of transcriptomic data is still limited. Here, we used Weighted Gene Co-Expression Network Analysis (WGCNA) to identify modules of coexpressed genes in muscle and brain tissue of a lake whitefish backcross progeny. Modules were connected to gradients of known adaptive traits involved in the ecological speciation process between benthic and limnetic ecotypes. Key drivers, that is, hub genes of functional modules related to reproduction, growth, and behavior were identified, and module preservation was assessed in natural populations. Using this approach, we identified modules of coexpressed genes involved in phenotypic divergence and their key drivers, and further identified a module part specifically rewired in the backcross progeny. Functional analysis of transcriptomic data can significantly contribute to the understanding of the mechanisms underlying ecological speciation. Our findings point to bone morphogenetic protein and calcium signaling as common pathways involved in coordinated evolution of trophic behavior, trophic morphology (gill rakers), and reproduction. Results also point to pathways implicating hemoglobins and constitutive stress response (HSP70) governing growth in lake whitefish.

Differential network analysis reveals genetic effects on catalepsy modules

We performed short-term bi-directional selective breeding for haloperidol-induced catalepsy, starting from three mouse populations of increasingly complex genetic structure: an F₂ intercross, a heterogeneous stock (HS) formed by crossing four inbred strains (HS4) and a heterogeneous stock (HS-CC) formed from the inbred strain founders of the Collaborative Cross (CC). All three selections were successful, with large differences in haloperidol response emerging within three generations. Using a custom differential network analysis procedure, we found that gene coexpression patterns changed significantly; importantly, a number of these changes were concordant across genetic backgrounds. In contrast, absolute gene-expression changes were modest and not concordant across genetic backgrounds, in spite of the large and similar phenotypic differences. By inferring strain contributions from the parental lines, we are able to identify significant differences in allelic content between the selected lines concurrent with large changes in transcript connectivity. Importantly, this observation implies that genetic polymorphisms can affect transcript and module connectivity without large changes in absolute expression levels. We conclude that, in this case, selective breeding acts at the subnetwork level, with the same modules but not the same transcripts affected across the three selections.

Blood-informative transcripts define nine common axes of peripheral blood gene expression

We describe a novel approach to capturing the covariance structure of peripheral blood gene expression that relies on the identification of highly conserved Axes of variation. Starting with a comparison of microarray transcriptome profiles for a new dataset of 189 healthy adult participants in the Emory-Georgia Tech Center for Health Discovery and Well-Being (CHDWB) cohort, with a previously published study of 208 adult Moroccans, we identify nine Axes each with between 99 and 1,028 strongly co-regulated transcripts in common. Each axis is enriched for gene ontology categories related to sub-classes of blood and immune function, including T-cell and B-cell physiology and innate, adaptive, and anti-viral responses. Conservation of the Axes is demonstrated in each of five additional population-based gene expression profiling studies, one of which is robustly associated with Body Mass Index in the CHDWB as well as Finnish and Australian cohorts. Furthermore, ten tightly co-regulated genes can be used to define each Axis as "Blood Informative Transcripts" (BITs), generating scores that define an individual with respect to the represented immune activity and blood physiology. We show that environmental factors, including lifestyle differences in Morocco and infection leading to active or latent tuberculosis, significantly impact specific axes, but that there is also significant heritability for the Axis scores. In the context of personalized medicine, reanalysis of the longitudinal profile of one individual during and after infection with two respiratory viruses demonstrates that specific axes also characterize clinical incidents. This mode of analysis suggests the view that, rather than unique subsets of genes marking each class of disease, differential expression reflects movement along the major normal Axes in response to environmental and genetic stimuli.

An inferential framework for biological network hypothesis tests

BACKGROUND

Networks are ubiquitous in modern cell biology and physiology. A large literature exists for inferring/proposing biological pathways/networks using statistical or machine learning algorithms. Despite these advances a formal testing procedure for analyzing network-level observations is in need of further development. Comparing the behaviour of a pharmacologically altered pathway to its canonical form is an example of a salient one-sample comparison. Locating which pathways differentiate disease from no-disease phenotype may be recast as a two-sample network inference problem.

RESULTS

We outline an inferential method for performing one- and two-sample hypothesis tests where the sampling unit is a network and the hypotheses are stated via network model(s). We propose a dissimilarity measure that incorporates nearby neighbour information to contrast one or more networks in a statistical test. We demonstrate and explore the utility of our approach with both simulated and microarray data; random graphs and weighted (partial) correlation networks are used to form network models. Using both a well-known diabetes dataset and an ovarian cancer dataset, the methods outlined here could better elucidate co-regulation changes for one or more pathways between two clinically relevant phenotypes.

CONCLUSIONS

Formal hypothesis tests for gene- or protein-based networks are a logical progression from existing gene-based and gene-set tests for differential expression. Commensurate with the growing appreciation and development of systems biology, the dissimilarity-based testing methods presented here may allow us to improve our understanding of pathways and other complex regulatory systems. The benefit of our method was illustrated under select scenarios.

Cluster and propensity based approximation of a network

Background

The models in this article generalize current models for both correlation networks and multigraph networks. Correlation networks are widely applied in genomics research. In contrast to general networks, it is straightforward to test the statistical significance of an edge in a correlation network. It is also easy to decompose the underlying correlation matrix and generate informative network statistics such as the module eigenvector. However, correlation networks only capture the connections between numeric variables. An open question is whether one can find suitable decompositions of the similarity measures employed in constructing general networks. Multigraph networks are attractive because they support likelihood based inference. Unfortunately, it is unclear how to adjust current statistical methods to detect the clusters inherent in many data sets.

Results

Here we present an intuitive and parsimonious parametrization of a general similarity measure such as a network adjacency matrix. The cluster and propensity based approximation (CPBA) of a network not only generalizes correlation network methods but also multigraph methods. In particular, it gives rise to a novel and more realistic multigraph model that accounts for clustering and provides likelihood based tests for assessing the significance of an edge after controlling for clustering. We present a novel Majorization-Minimization (MM) algorithm for estimating the parameters of the CPBA. To illustrate the practical utility of the CPBA of a network, we apply it to gene expression data and to a bi-partite network model for diseases and disease genes from the Online Mendelian Inheritance in Man (OMIM).

Conclusions

The CPBA of a network is theoretically appealing since a) it generalizes correlation and multigraph network methods, b) it improves likelihood based significance tests for edge counts, c) it directly models higher-order relationships between clusters, and d) it suggests novel clustering algorithms. The CPBA of a network is implemented in Fortran 95 and bundled in the freely available R package PropClust.

Maximal information component analysis: a novel non-linear network analysis method

Background: Network construction and analysis algorithms provide scientists with the ability to sift through high-throughput biological outputs, such as transcription microarrays, for small groups of genes (modules) that are relevant for further research. Most of these algorithms ignore the important role of non-linear interactions in the data, and the ability for genes to operate in multiple functional groups at once, despite clear evidence for both of these phenomena in observed biological systems.

Results: We have created a novel co-expression network analysis algorithm that incorporates both of these principles by combining the information-theoretic association measure of the maximal information coefficient (MIC) with an Interaction Component Model. We evaluate the performance of this approach on two datasets collected from a large panel of mice, one from macrophages and the other from liver by comparing the two measures based on a measure of module entropy, Gene Ontology (GO) enrichment, and scale-free topology (SFT) fit. Our algorithm outperforms a widely used co-expression analysis method, weighted gene co-expression network analysis (WGCNA), in the macrophage data, while returning comparable results in the liver dataset when using these criteria. We demonstrate that the macrophage data has more non-linear interactions than the liver dataset, which may explain the increased performance of our method, termed Maximal Information Component Analysis (MICA) in that case.

Conclusions: In making our network algorithm more accurately reflect known biological principles, we are able to generate modules with improved relevance, particularly in networks with confounding factors such as gene by environment interactions.

Effects on gene expression in rat liver after administration of RXR agonists: UAB30, 4-methyl-UAB30, and Targretin (Bexarotene)

Examination of three retinoid X receptor (RXR) agonists [Targretin (TRG), UAB30, and 4-methyl-UAB30 (4-Me-UAB30)] showed that all inhibited mammary cancer in rodents and two (TRG and 4-Me-UAB30) strikingly increased serum triglyceride levels. Agents were administered in diets to female Sprague-Dawley rats. Liver RNA was isolated and microarrayed on the Affymetrix GeneChip Rat Exon 1.0 ST array. Statistical tests identified genes that exhibited differential expression and fell into groups, or modules, with differential expression among agonists. Genes in specific modules were changed by one, two, or all three agonists. An interactome analysis assessed the effects on genes that heterodimerize with known nuclear receptors. For proliferator-activated receptor α/RXR-activated genes, the strongest response was TRG > 4-Me-UAB30 > UAB30. Many liver X receptor/RXR-related genes (e.g., Scd-1 and Srebf1, which are associated with increased triglycerides) were highly expressed in TRG and 4-Me-UAB30- but not UAB30-treated livers. Minimal expression changes were associated with retinoic acid receptor or vitamin D receptor heterodimers by any of the agonists. UAB30 unexpectedly and uniquely activated genes associated with the aryl hydrocarbon hydroxylase (Ah) receptor (Cyp1a1, Cyp1a2, Cyp1b1, and Nqo1). Based on the Ah receptor activation, UAB30 was tested for its ability to prevent dimethylbenzanthracene (DMBA)-induced mammary cancers, presumably by inhibiting DMBA activation, and was highly effective. Gene expression changes were determined by reverse transcriptase-polymerase chain reaction in rat livers treated with Targretin for 2.3, 7, and 21 days. These showed similar gene expression changes at all three time points, arguing some steady-state effect. Different patterns of gene expression among the agonists provided insight into molecular differences and allowed one to predict certain physiologic consequences of agonist treatment.

Leveraging non-targeted metabolite profiling via statistical genomics

One of the challenges of systems biology is to integrate multiple sources of data in order to build a cohesive view of the system of study. Here we describe the mass spectrometry based profiling of maize kernels, a model system for genomic studies and a cornerstone of the agroeconomy. Using a network analysis, we can include 97.5% of the 8,710 features detected from 210 varieties into a single framework. More conservatively, 47.1% of compounds detected can be organized into a network with 48 distinct modules. Eigenvalues were calculated for each module and then used as inputs for genome-wide association studies. Nineteen modules returned significant results, illustrating the genetic control of biochemical networks within the maize kernel. Our approach leverages the correlations between the genome and metabolome to mutually enhance their annotation and thus enable biological interpretation. This method is applicable to any organism with sufficient bioinformatic resources.

Modeling host genetic regulation of influenza pathogenesis in the collaborative cross

Genetic variation contributes to host responses and outcomes following infection by influenza A virus or other viral infections. Yet narrow windows of disease symptoms and confounding environmental factors have made it difficult to identify polymorphic genes that contribute to differential disease outcomes in human populations. Therefore, to control for these confounding environmental variables in a system that models the levels of genetic diversity found in outbred populations such as humans, we used incipient lines of the highly genetically diverse Collaborative Cross (CC) recombinant inbred (RI) panel (the pre-CC population) to study how genetic variation impacts influenza associated disease across a genetically diverse population. A wide range of variation in influenza disease related phenotypes including virus replication, virus-induced inflammation, and weight loss was observed. Many of the disease associated phenotypes were correlated, with viral replication and virus-induced inflammation being predictors of virus-induced weight loss. Despite these correlations, pre-CC mice with unique and novel disease phenotype combinations were observed. We also identified sets of transcripts (modules) that were correlated with aspects of disease. In order to identify how host genetic polymorphisms contribute to the observed variation in disease, we conducted quantitative trait loci (QTL) mapping. We identified several QTL contributing to specific aspects of the host response including virus-induced weight loss, titer, pulmonary edema, neutrophil recruitment to the airways, and transcriptional expression. Existing whole-genome sequence data was applied to identify high priority candidate genes within QTL regions. A key host response QTL was located at the site of the known anti-influenza Mx1 gene. We sequenced the coding regions of Mx1 in the eight CC founder strains, and identified a novel Mx1 allele that showed reduced ability to inhibit viral replication, while maintaining protection from weight loss.

Host responses to an infectious agent are highly variable across the human population, however, it is not entirely clear how various factors such as pathogen dose, demography, environment and host genetic polymorphisms contribute to variable host responses and infectious outcomes. In this study, a new in vivo experimental model was used that recapitulates many of the genetic characteristics of an outbred population, such as humans. By controlling viral dose, environment and demographic variables, we were able to focus on the role that host genetic variation plays in influenza virus infection. Both the range of disease phenotypes and the combinations of sets of disease phenotypes at 4 days post infection across this population exhibited a large amount of diversity, reminiscent of the variation seen across the human population. Multiple host genome regions were identified that contributed to different aspects of the host response to influenza infection. Taken together, these results emphasize the critical role of host genetics in the response to infectious diseases. Given the breadth of host responses seen within this population, several new models for unique host responses to infection were identified.

Cross-species transcriptional network analysis reveals conservation and variation in response to metal stress in cyanobacteria

BACKGROUND

As one of the most dominant bacterial groups on Earth, cyanobacteria play a pivotal role in the global carbon cycling and the Earth atmosphere composition. Understanding their molecular responses to environmental perturbations has important scientific and environmental values. Since important biological processes or networks are often evolutionarily conserved, the cross-species transcriptional network analysis offers a useful strategy to decipher conserved and species-specific transcriptional mechanisms that cells utilize to deal with various biotic and abiotic disturbances, and it will eventually lead to a better understanding of associated adaptation and regulatory networks.

RESULTS

In this study, the Weighted Gene Co-expression Network Analysis (WGCNA) approach was used to establish transcriptional networks for four important cyanobacteria species under metal stress, including iron depletion and high copper conditions. Cross-species network comparison led to discovery of several core response modules and genes possibly essential to metal stress, as well as species-specific hub genes for metal stresses in different cyanobacteria species, shedding light on survival strategies of cyanobacteria responding to different environmental perturbations.

CONCLUSIONS

The WGCNA analysis demonstrated that the application of cross-species transcriptional network analysis will lead to novel insights to molecular response to environmental changes which will otherwise not be achieved by analyzing data from a single species.

canEvolve: a web portal for integrative oncogenomics

Background & Objective

Genome-wide profiles of tumors obtained using functional genomics platforms are being deposited to the public repositories at an astronomical scale, as a result of focused efforts by individual laboratories and large projects such as the Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium. Consequently, there is an urgent need for reliable tools that integrate and interpret these data in light of current knowledge and disseminate results to biomedical researchers in a user-friendly manner. We have built the canEvolve web portal to meet this need.

Results

canEvolve query functionalities are designed to fulfill most frequent analysis needs of cancer researchers with a view to generate novel hypotheses. canEvolve stores gene, microRNA (miRNA) and protein expression profiles, copy number alterations for multiple cancer types, and protein-protein interaction information. canEvolve allows querying of results of primary analysis, integrative analysis and network analysis of oncogenomics data. The querying for primary analysis includes differential gene and miRNA expression as well as changes in gene copy number measured with SNP microarrays. canEvolve provides results of integrative analysis of gene expression profiles with copy number alterations and with miRNA profiles as well as generalized integrative analysis using gene set enrichment analysis. The network analysis capability includes storage and visualization of gene co-expression, inferred gene regulatory networks and protein-protein interaction information. Finally, canEvolve provides correlations between gene expression and clinical outcomes in terms of univariate survival analysis.

Conclusion

At present canEvolve provides different types of information extracted from 90 cancer genomics studies comprising of more than 10,000 patients. The presence of multiple data types, novel integrative analysis for identifying regulators of oncogenesis, network analysis and ability to query gene lists/pathways are distinctive features of canEvolve. canEvolve will facilitate integrative and meta-analysis of oncogenomics datasets.

Availability

The canEvolve web portal is available at http://www.canevolve.org/.

Whole genome gene expression meta-analysis of inflammatory bowel disease colon mucosa demonstrates lack of major differences between Crohn's disease and ulcerative colitis

Background

In inflammatory bowel disease (IBD), genetic susceptibility together with environmental factors disturbs gut homeostasis producing chronic inflammation. The two main IBD subtypes are Ulcerative colitis (UC) and Crohn’s disease (CD). We present the to-date largest microarray gene expression study on IBD encompassing both inflamed and un-inflamed colonic tissue. A meta-analysis including all available, comparable data was used to explore important aspects of IBD inflammation, thereby validating consistent gene expression patterns.

Methods

Colon pinch biopsies from IBD patients were analysed using Illumina whole genome gene expression technology. Differential expression (DE) was identified using LIMMA linear model in the R statistical computing environment. Results were enriched for gene ontology (GO) categories. Sets of genes encoding antimicrobial proteins (AMP) and proteins involved in T helper (Th) cell differentiation were used in the interpretation of the results. All available data sets were analysed using the same methods, and results were compared on a global and focused level as t-scores.

Results

Gene expression in inflamed mucosa from UC and CD are remarkably similar. The meta-analysis confirmed this. The patterns of AMP and Th cell-related gene expression were also very similar, except for IL23A which was consistently higher expressed in UC than in CD. Un-inflamed tissue from patients demonstrated minimal differences from healthy controls.

Conclusions

There is no difference in the Th subgroup involvement between UC and CD. Th1/Th17 related expression, with little Th2 differentiation, dominated both diseases. The different IL23A expression between UC and CD suggests an IBD subtype specific role. AMPs, previously little studied, are strongly overexpressed in IBD. The presented meta-analysis provides a sound background for further research on IBD pathobiology.

Systems analysis of human brain gene expression: mechanisms for HIV-associated neurocognitive impairment and common pathways with Alzheimer's disease

Background

Human Immunodeficiency Virus-1 (HIV) infection frequently results in neurocognitive impairment. While the cause remains unclear, recent gene expression studies have identified genes whose transcription is dysregulated in individuals with HIV-association neurocognitive disorder (HAND). However, the methods for interpretation of such data have lagged behind the technical advances allowing the decoding genetic material. Here, we employ systems biology methods novel to the field of NeuroAIDS to further interrogate extant transcriptome data derived from brains of HIV + patients in order to further elucidate the neuropathogenesis of HAND. Additionally, we compare these data to those derived from brains of individuals with Alzheimer’s disease (AD) in order to identify common pathways of neuropathogenesis.

Methods

In Study 1, using data from three brain regions in 6 HIV-seronegative and 15 HIV + cases, we first employed weighted gene co-expression network analysis (WGCNA) to further explore transcriptome networks specific to HAND with HIV-encephalitis (HIVE) and HAND without HIVE. We then used a symptomatic approach, employing standard expression analysis and WGCNA to identify networks associated with neurocognitive impairment (NCI), regardless of HIVE or HAND diagnosis. Finally, we examined the association between the CNS penetration effectiveness (CPE) of antiretroviral regimens and brain transcriptome. In Study 2, we identified common gene networks associated with NCI in both HIV and AD by correlating gene expression with pre-mortem neurocognitive functioning.

Results

Study 1: WGCNA largely corroborated findings from standard differential gene expression analyses, but also identified possible meta-networks composed of multiple gene ontology categories and oligodendrocyte dysfunction. Differential expression analysis identified hub genes highly correlated with NCI, including genes implicated in gliosis, inflammation, and dopaminergic tone. Enrichment analysis identified gene ontology categories that varied across the three brain regions, the most notable being downregulation of genes involved in mitochondrial functioning. Finally, WGCNA identified dysregulated networks associated with NCI, including oligodendrocyte and mitochondrial functioning. Study 2: Common gene networks dysregulated in relation to NCI in AD and HIV included mitochondrial genes, whereas upregulation of various cancer-related genes was found.

Conclusions

While under-powered, this study identified possible biologically-relevant networks correlated with NCI in HIV, and common networks shared with AD, opening new avenues for inquiry in the investigation of HAND neuropathogenesis. These results suggest that further interrogation of existing transcriptome data using systems biology methods can yield important information.

Massive-scale gene co-expression network construction and robustness testing using random matrix theory

The study of gene relationships and their effect on biological function and phenotype is a focal point in systems biology. Gene co-expression networks built using microarray expression profiles are one technique for discovering and interpreting gene relationships. A knowledge-independent thresholding technique, such as Random Matrix Theory (RMT), is useful for identifying meaningful relationships. Highly connected genes in the thresholded network are then grouped into modules that provide insight into their collective functionality. While it has been shown that co-expression networks are biologically relevant, it has not been determined to what extent any given network is functionally robust given perturbations in the input sample set. For such a test, hundreds of networks are needed and hence a tool to rapidly construct these networks. To examine functional robustness of networks with varying input, we enhanced an existing RMT implementation for improved scalability and tested functional robustness of human (Homo sapiens), rice (Oryza sativa) and budding yeast (Saccharomyces cerevisiae). We demonstrate dramatic decrease in network construction time and computational requirements and show that despite some variation in global properties between networks, functional similarity remains high. Moreover, the biological function captured by co-expression networks thresholded by RMT is highly robust.

Laser microdissection and microarray analysis of Tuber melanosporum ectomycorrhizas reveal functional heterogeneity between mantle and Hartig net compartments

The ectomycorrhizal (ECM) symbiosis, a mutualistic plant-fungus association, plays a fundamental role in forest ecosystems by enhancing plant growth and by providing host protection from root diseases. The cellular complexity of the symbiotic organ, characterized by the differentiation of structurally specialized tissues (i.e. the fungal mantle and the Hartig net), is the major limitation to study fungal gene expression in such specific compartments. We investigated the transcriptional landscape of the ECM fungus Tuber melanosporum during the major stages of its life cycle and we particularly focused on the complex symbiotic stage by combining the use of laser capture microdissection and microarray gene expression analysis. We isolated the fungal/soil (i.e. the mantle) and the fungal/plant (i.e. the Hartig net) interfaces from transverse sections of T. melanosporum/Corylus avellana ectomycorrhizas and identified the distinct genetic programmes associated with each compartment. Particularly, nitrogen and water acquisition from soil, synthesis of secondary metabolites and detoxification mechanisms appear to be important processes in the fungal mantle. In contrast, transport activity is enhanced in the Hartig net and we identified carbohydrate and nitrogen-derived transporters that might play a key role in the reciprocal resources' transfer between the host and the symbiont.

Systems Biology Approaches and Applications in Obesity, Diabetes, and Cardiovascular Diseases

The metabolically connected triad of obesity, diabetes, and cardiovascular diseases is a major public health threat, and is expected to worsen due to the global shift toward energy-rich and sedentary living. Despite decades of intense research, a large part of the molecular pathogenesis behind complex metabolic diseases remains unknown. Recent advances in genetics, epigenomics, transcriptomics, proteomics and metabolomics enable us to obtain large-scale snapshots of the etiological processes in multiple disease-related cells, tissues and organs. These datasets provide us with an opportunity to go beyond conventional reductionist approaches and to pinpoint the specific perturbations in critical biological processes. In this review, we summarize systems biology methodologies such as functional genomics, causality inference, data-driven biological network construction, and higher-level integrative analyses that can produce novel mechanistic insights, identify disease biomarkers, and uncover potential therapeutic targets from a combination of omics datasets. Importantly, we also demonstrate the power of these approaches by application examples in obesity, diabetes, and cardiovascular diseases.

Gene coexpression network analysis identifies genes and biological processes shared among anterior pituitary and brain areas that affect estrous behavior in dairy cows

The expression of estrous (sexually receptive) behavior (EB), a key fertility trait in dairy cows, has been declining over the past few decades both in intensity and duration. Improved knowledge of the genomic factors underlying EB, which is currently lacking, may lead to novel applications to enhance fertility. Our objective was to identify genes and biological processes shared among the bovine anterior pituitary (AP) and four brain areas that act together to regulate EB by investigating networks of coexpressed genes between these tissues. We used a systems biology approach called weighted gene coexpression network analysis for defining gene coexpression networks using gene expression data from the following tissues collected from 14 cows at estrus: AP, dorsal hypothalamus (DH), ventral hypothalamus (VH), amygdala (AM), and hippocampus (HC). Consensus modules of coexpressed genes were identified between the networks for the AM-DH, HC-DH, VH-DH, AP-DH, and AM-HC tissue pairs. The correlation between the module's eigengene (weighted average gene expression profile) and levels of EB exhibited by the experimental cows were tested. Estrous behavior-correlated modules were found enriched for gene ontology terms like glial cell development and regulation of neural projection development as well as for Kyoto Encyclopedia of Genes and Genomes pathway terms related to brain degenerative diseases. General cellular processes like oxidative phosphorylation and ribosome and biosynthetic processes were found enriched in several correlated modules, indicating increased transcription and protein synthesis. Stimulation of ribosomal RNA synthesis is known from rodent studies to be a primary event in the activation of neuronal cells and pathways involved in female reproductive behavior and this precedes the estrogen-driven expansion of dendrites and synapses. Similar processes also operate in cows to affect EB. Hub genes within EB-correlated modules (e.g. NEFL, NDRG2, GAP43, THY1, and TCF7L2, among others) are strong candidates among genes regulating EB expression. The study improved our understanding of the genomic regulation of EB in dairy cows by providing new insights into genes and biological processes shared among the bovine AP and brain areas acting together to regulate EB. The new knowledge could lead to the development of novel management strategies to monitor and improve reproductive performance in dairy cows (for example, biomarkers for estrus detection).

PLAU inferred from a correlation network is critical for suppressor function of regulatory T cells

Human FOXP3(+)CD25(+)CD4(+) regulatory T cells (Tregs) are essential to the maintenance of immune homeostasis. Several genes are known to be important for murine Tregs, but for human Tregs the genes and underlying molecular networks controlling the suppressor function still largely remain unclear. Here, we describe a strategy to identify the key genes directly from an undirected correlation network which we reconstruct from a very high time-resolution (HTR) transcriptome during the activation of human Tregs/CD4(+) T-effector cells. We show that a predicted top-ranked new key gene PLAU (the plasminogen activator urokinase) is important for the suppressor function of both human and murine Tregs. Further analysis unveils that PLAU is particularly important for memory Tregs and that PLAU mediates Treg suppressor function via STAT5 and ERK signaling pathways. Our study demonstrates the potential for identifying novel key genes for complex dynamic biological processes using a network strategy based on HTR data, and reveals a critical role for PLAU in Treg suppressor function.

Transcriptome comparison and gene coexpression network analysis provide a systems view of citrus response to 'Candidatus Liberibacter asiaticus' infection

BACKGROUND

Huanglongbing (HLB) is arguably the most destructive disease for the citrus industry. HLB is caused by infection of the bacterium, Candidatus Liberibacter spp. Several citrus GeneChip studies have revealed thousands of genes that are up- or down-regulated by infection with Ca. Liberibacter asiaticus. However, whether and how these host genes act to protect against HLB remains poorly understood.

RESULTS

As a first step towards a mechanistic view of citrus in response to the HLB bacterial infection, we performed a comparative transcriptome analysis and found that a total of 21 Probesets are commonly up-regulated by the HLB bacterial infection. In addition, a number of genes are likely regulated specifically at early, late or very late stages of the infection. Furthermore, using Pearson correlation coefficient-based gene coexpression analysis, we constructed a citrus HLB response network consisting of 3,507 Probesets and 56,287 interactions. Genes involved in carbohydrate and nitrogen metabolic processes, transport, defense, signaling and hormone response were overrepresented in the HLB response network and the subnetworks for these processes were constructed. Analysis of the defense and hormone response subnetworks indicates that hormone response is interconnected with defense response. In addition, mapping the commonly up-regulated HLB responsive genes into the HLB response network resulted in a core subnetwork where transport plays a key role in the citrus response to the HLB bacterial infection. Moreover, analysis of a phloem protein subnetwork indicates a role for this protein and zinc transporters or zinc-binding proteins in the citrus HLB defense response.

CONCLUSION

Through integrating transcriptome comparison and gene coexpression network analysis, we have provided for the first time a systems view of citrus in response to the Ca. Liberibacter spp. infection causing HLB.

Random generalized linear model: a highly accurate and interpretable ensemble predictor

BACKGROUND

Ensemble predictors such as the random forest are known to have superior accuracy but their black-box predictions are difficult to interpret. In contrast, a generalized linear model (GLM) is very interpretable especially when forward feature selection is used to construct the model. However, forward feature selection tends to overfit the data and leads to low predictive accuracy. Therefore, it remains an important research goal to combine the advantages of ensemble predictors (high accuracy) with the advantages of forward regression modeling (interpretability). To address this goal several articles have explored GLM based ensemble predictors. Since limited evaluations suggested that these ensemble predictors were less accurate than alternative predictors, they have found little attention in the literature.

RESULTS

Comprehensive evaluations involving hundreds of genomic data sets, the UCI machine learning benchmark data, and simulations are used to give GLM based ensemble predictors a new and careful look. A novel bootstrap aggregated (bagged) GLM predictor that incorporates several elements of randomness and instability (random subspace method, optional interaction terms, forward variable selection) often outperforms a host of alternative prediction methods including random forests and penalized regression models (ridge regression, elastic net, lasso). This random generalized linear model (RGLM) predictor provides variable importance measures that can be used to define a "thinned" ensemble predictor (involving few features) that retains excellent predictive accuracy.

CONCLUSION

RGLM is a state of the art predictor that shares the advantages of a random forest (excellent predictive accuracy, feature importance measures, out-of-bag estimates of accuracy) with those of a forward selected generalized linear model (interpretability). These methods are implemented in the freely available R software package randomGLM.

Genetic architecture of gene expression in the chicken

Background

The annotation of many genomes is limited, with a large proportion of identified genes lacking functional assignments. The construction of gene co-expression networks is a powerful approach that presents a way of integrating information from diverse gene expression datasets into a unified analysis which allows inferences to be drawn about the role of previously uncharacterised genes. Using this approach, we generated a condition-free gene co-expression network for the chicken using data from 1,043 publically available Affymetrix GeneChip Chicken Genome Arrays. This data was generated from a diverse range of experiments, including different tissues and experimental conditions. Our aim was to identify gene co-expression modules and generate a tool to facilitate exploration of the functional chicken genome.

Results

Fifteen modules, containing between 24 and 473 genes, were identified in the condition-free network. Most of the modules showed strong functional enrichment for particular Gene Ontology categories. However, a few showed no enrichment. Transcription factor binding site enrichment was also noted.

Conclusions

We have demonstrated that this chicken gene co-expression network is a useful tool in gene function prediction and the identification of putative novel transcription factors and binding sites. This work highlights the relevance of this methodology for functional prediction in poorly annotated genomes such as the chicken.

iBIG: an integrative network tool for supporting human disease mechanism studies

Understanding the mechanism of complex human diseases is a major scientific challenge. Towards this end, we developed a web-based network tool named iBIG (stands for integrative BIoloGy), which incorporates a variety of information on gene interaction and regulation. The generated network can be annotated with various types of information and visualized directly online. In addition to the gene networks based on physical and pathway interactions, networks at a functional level can also be constructed. Furthermore, a supplementary R package is provided to process microarray data and generate a list of important genes to be used as input for iBIG. To demonstrate its usefulness, we collected 54 microarrays on common human diseases including cancer, neurological disorders, infectious diseases and other common diseases. We processed the microarray data with our R package and constructed a network of functional modules perturbed in common human diseases. Networks at the functional level in combination with gene networks may provide new insight into the mechanism of human diseases. iBIG is freely available at http://lei.big.ac.cn/ibig.

Systems-level analysis of genome-wide association data

Genome-wide association studies (GWAS) have emerged as the method of choice for identifying common variants affecting complex disease. In a GWAS, particular attention is placed, for obvious reasons, on single-nucleotide polymorphisms (SNPs) that exceed stringent genome-wide significance thresholds. However, it is expected that many SNPs with only nominal evidence of association (e.g., P < 0.05) truly influence disease. Efforts to extract additional biological information from entire GWAS datasets have primarily focused on pathway-enrichment analyses. However, these methods suffer from a number of limitations and typically fail to lead to testable hypotheses. To evaluate alternative approaches, we performed a systems-level analysis of GWAS data using weighted gene coexpression network analysis. A weighted gene coexpression network was generated for 1918 genes harboring SNPs that displayed nominal evidence of association (P ≤ 0.05) from a GWAS of bone mineral density (BMD) using microarray data on circulating monocytes isolated from individuals with extremely low or high BMD. Thirteen distinct gene modules were identified, each comprising coexpressed and highly interconnected GWAS genes. Through the characterization of module content and topology, we illustrate how network analysis can be used to discover disease-associated subnetworks and characterize novel interactions for genes with a known role in the regulation of BMD. In addition, we provide evidence that network metrics can be used as a prioritizing tool when selecting genes and SNPs for replication studies. Our results highlight the advantages of using systems-level strategies to add value to and inform GWAS.

Connecting signaling pathways underlying communication to ASD vulnerability

Language is a human-specific trait that likely facilitated the rapid increase in higher cognitive function in our species. A consequence of the selective pressures that have permitted language and cognition to flourish in humans is the unique vulnerability of humans to developing cognitive disorders such as autism. Therefore, progress in understanding the genetic and molecular mechanisms of language evolution should provide insight into such disorders. Here, we discuss the few genes that have been identified in both autism-related pathways and language. We also detail the use of animal models to uncover the function of these genes at a mechanistic and circuit level. Finally, we present the use of comparative genomics to identify novel genes and gene networks involved in autism. Together, all of these approaches will allow for a broader and deeper view of the molecular brain mechanisms involved in the evolution of language and the gene disruptions associated with autism.

Approaches to modeling gene regulatory networks: a gentle introduction

This chapter is split into two main sections; first, I will present an introduction to gene networks. Second, I will discuss various approaches to gene network modeling which will include some examples for using different data sources. Computational modeling has been used for many different biological systems and many approaches have been developed addressing the different needs posed by the different application fields. The modeling approaches presented here are not limited to gene regulatory networks and occasionally I will present other examples. The material covered here is an update based on several previous publications by Thomas Schlitt and Alvis Brazma (FEBS Lett 579(8),1859-1866, 2005; Philos Trans R Soc Lond B Biol Sci 361(1467), 483-494, 2006; BMC Bioinformatics 8(suppl 6), S9, 2007) that formed the foundation for a lecture on gene regulatory networks at the In Silico Systems Biology workshop series at the European Bioinformatics Institute in Hinxton.

OPTIMAS-DW: a comprehensive transcriptomics, metabolomics, ionomics, proteomics and phenomics data resource for maize

BACKGROUND

Maize is a major crop plant, grown for human and animal nutrition, as well as a renewable resource for bioenergy. When looking at the problems of limited fossil fuels, the growth of the world's population or the world's climate change, it is important to find ways to increase the yield and biomass of maize and to study how it reacts to specific abiotic and biotic stress situations. Within the OPTIMAS systems biology project maize plants were grown under a large set of controlled stress conditions, phenotypically characterised and plant material was harvested to analyse the effect of specific environmental conditions or developmental stages. Transcriptomic, metabolomic, ionomic and proteomic parameters were measured from the same plant material allowing the comparison of results across different omics domains. A data warehouse was developed to store experimental data as well as analysis results of the performed experiments.

DESCRIPTION

The OPTIMAS Data Warehouse (OPTIMAS-DW) is a comprehensive data collection for maize and integrates data from different data domains such as transcriptomics, metabolomics, ionomics, proteomics and phenomics. Within the OPTIMAS project, a 44K oligo chip was designed and annotated to describe the functions of the selected unigenes. Several treatment- and plant growth stage experiments were performed and measured data were filled into data templates and imported into the data warehouse by a Java based import tool. A web interface allows users to browse through all stored experiment data in OPTIMAS-DW including all data domains. Furthermore, the user can filter the data to extract information of particular interest. All data can be exported into different file formats for further data analysis and visualisation. The data analysis integrates data from different data domains and enables the user to find answers to different systems biology questions. Finally, maize specific pathway information is provided.

CONCLUSIONS

With OPTIMAS-DW a data warehouse for maize was established, which is able to handle different data domains, comprises several analysis results that will support researchers within their work and supports systems biological research in particular. The system is available at http://www.optimas-bioenergy.org/optimas_dw.

Systems genetic analysis of osteoblast-lineage cells

The osteoblast-lineage consists of cells at various stages of maturation that are essential for skeletal development, growth, and maintenance. Over the past decade, many of the signaling cascades that regulate this lineage have been elucidated; however, little is known of the networks that coordinate, modulate, and transmit these signals. Here, we identify a gene network specific to the osteoblast-lineage through the reconstruction of a bone co-expression network using microarray profiles collected on 96 Hybrid Mouse Diversity Panel (HMDP) inbred strains. Of the 21 modules that comprised the bone network, module 9 (M9) contained genes that were highly correlated with prototypical osteoblast maker genes and were more highly expressed in osteoblasts relative to other bone cells. In addition, the M9 contained many of the key genes that define the osteoblast-lineage, which together suggested that it was specific to this lineage. To use the M9 to identify novel osteoblast genes and highlight its biological relevance, we knocked-down the expression of its two most connected “hub” genes, Maged1 and Pard6g. Their perturbation altered both osteoblast proliferation and differentiation. Furthermore, we demonstrated the mice deficient in Maged1 had decreased bone mineral density (BMD). It was also discovered that a local expression quantitative trait locus (eQTL) regulating the Wnt signaling antagonist Sfrp1 was a key driver of the M9. We also show that the M9 is associated with BMD in the HMDP and is enriched for genes implicated in the regulation of human BMD through genome-wide association studies. In conclusion, we have identified a physiologically relevant gene network and used it to discover novel genes and regulatory mechanisms involved in the function of osteoblast-lineage cells. Our results highlight the power of harnessing natural genetic variation to generate co-expression networks that can be used to gain insight into the function of specific cell-types.

The osteoblast-lineage consists of a range of cells from osteogenic precursors that mature into bone-forming osteoblasts to osteocytes that are entombed in bone. Each cell in the lineage serves a number of distinct and critical roles in the growth and maintenance of the skeleton, as well as many extra-skeletal functions. Over the last decade, many of the major regulatory pathways governing the differentiation and activity of these cells have been discovered. In contrast, little is known regarding the composition or function of gene networks within the lineage. The goal of this study was to increase our understanding of how genes are organized into networks in osteoblasts. Towards this goal, we used microarray gene expression profiles from bone to identify a group of genes that formed a network specific to the osteoblast-lineage. We used the knowledge of this network to identify novel genes that are important for regulating various aspects of osteoblast function. These data improve our understanding of the gene networks operative in cells of the osteoblast-lineage.

The network structure of human personality according to the NEO-PI-R: matching network community structure to factor structure

Introduction

Human personality is described preferentially in terms of factors (dimensions) found using factor analysis. An alternative and highly related method is network analysis, which may have several advantages over factor analytic methods.

Aim

To directly compare the ability of network community detection (NCD) and principal component factor analysis (PCA) to examine modularity in multidimensional datasets such as the neuroticism-extraversion-openness personality inventory revised (NEO-PI-R).

Methods

434 healthy subjects were tested on the NEO-PI-R. PCA was performed to extract factor structures (FS) of the current dataset using both item scores and facet scores. Correlational network graphs were constructed from univariate correlation matrices of interactions between both items and facets. These networks were pruned in a link-by-link fashion while calculating the network community structure (NCS) of each resulting network using the Wakita Tsurumi clustering algorithm. NCSs were matched against FS and networks of best matches were kept for further analysis.

Results

At facet level, NCS showed a best match (96.2%) with a ‘confirmatory’ 5-FS. At item level, NCS showed a best match (80%) with the standard 5-FS and involved a total of 6 network clusters. Lesser matches were found with ‘confirmatory’ 5-FS and ‘exploratory’ 6-FS of the current dataset. Network analysis did not identify facets as a separate level of organization in between items and clusters. A small-world network structure was found in both item- and facet level networks.

Conclusion

We present the first optimized network graph of personality traits according to the NEO-PI-R: a ‘Personality Web’. Such a web may represent the possible routes that subjects can take during personality development. NCD outperforms PCA by producing plausible modularity at item level in non-standard datasets, and can identify the key roles of individual items and clusters in the network.

Comparison and evaluation of pathway-level aggregation methods of gene expression data

Background

Microarray experiments produce expression measurements in genomic scale. A way to derive functional understanding of the data is to focus on functional sets of genes, such as pathways, instead of individual genes. While a common practice for the pathway-level analysis has been functional enrichment analysis such as over-representation analysis and gene set enrichment analysis, an alternative approach has also been explored. In this approach, gene expression data are first aggregated at pathway level to transform the original data into a compact representation in which each row corresponds to a pathway instead of a gene. Thereafter the pathway expression data can be used for differential expression and classification analyses in pathway space, leveraging existing algorithms usually applied to gene expression data. While several studies have proposed the pathway-level aggregation methods, it remains unclear how they compare with one another, since the evaluations were done to a limited extent. Thus this study presents a comprehensive evaluation of six most prominent aggregation methods.

Results

The compared methods include five existing methods--mean of all member genes (Mean all), mean of condition-responsive genes (Mean CORGs), analysis of sample set enrichment scores (ASSESS), principal component analysis (PCA), and partial least squares (PLS)--and a variant of an existing method (Mean top 50%, averaging top half of member genes). Comprehensive and stringent benchmarking was performed by collecting seven pairs of related but independent datasets encompassing various phenotypes. Aggregation was done in the space of KEGG pathways. Performance of the methods was assessed by classification accuracy validated both internally and externally, and by examining the correlative extent of pathway signatures between the dataset pairs. The assessment revealed that (i) the best accuracy and correlation were obtained from ASSESS and Mean top 50%, (ii) Mean all showed the lowest accuracy, and (iii) Mean CORGs and PLS gave rise to the largest extent of discordance in the pathway signature correlation.

Conclusions

The two best performing method (ASSESS and Mean top 50%) are suggested to be preferred. The benchmarking analysis also suggests that there is both room and necessity for developing a novel method for pathway-level aggregation.

A network module-based method for identifying cancer prognostic signatures

Discovering robust prognostic gene signatures as biomarkers using genomics data can be challenging. We have developed a simple but efficient method for discovering prognostic biomarkers in cancer gene expression data sets using modules derived from a highly reliable gene functional interaction network. When applied to breast cancer, we discover a novel 31-gene signature associated with patient survival. The signature replicates across 5 independent gene expression studies, and outperforms 48 published gene signatures. When applied to ovarian cancer, the algorithm identifies a 75-gene signature associated with patient survival. A Cytoscape plugin implementation of the signature discovery method is available at http://wiki.reactome.org/index.php/Reactome_FI_Cytoscape_Plugin.

Comparison of co-expression measures: mutual information, correlation, and model based indices

Background

Co-expression measures are often used to define networks among genes. Mutual information (MI) is often used as a generalized correlation measure. It is not clear how much MI adds beyond standard (robust) correlation measures or regression model based association measures. Further, it is important to assess what transformations of these and other co-expression measures lead to biologically meaningful modules (clusters of genes).

Results

We provide a comprehensive comparison between mutual information and several correlation measures in 8 empirical data sets and in simulations. We also study different approaches for transforming an adjacency matrix, e.g. using the topological overlap measure. Overall, we confirm close relationships between MI and correlation in all data sets which reflects the fact that most gene pairs satisfy linear or monotonic relationships. We discuss rare situations when the two measures disagree. We also compare correlation and MI based approaches when it comes to defining co-expression network modules. We show that a robust measure of correlation (the biweight midcorrelation transformed via the topological overlap transformation) leads to modules that are superior to MI based modules and maximal information coefficient (MIC) based modules in terms of gene ontology enrichment. We present a function that relates correlation to mutual information which can be used to approximate the mutual information from the corresponding correlation coefficient. We propose the use of polynomial or spline regression models as an alternative to MI for capturing non-linear relationships between quantitative variables.

Conclusion

The biweight midcorrelation outperforms MI in terms of elucidating gene pairwise relationships. Coupled with the topological overlap matrix transformation, it often leads to more significantly enriched co-expression modules. Spline and polynomial networks form attractive alternatives to MI in case of non-linear relationships. Our results indicate that MI networks can safely be replaced by correlation networks when it comes to measuring co-expression relationships in stationary data.

Adipose co-expression networks across Finns and Mexicans identify novel triglyceride-associated genes

Background

High serum triglyceride (TG) levels is an established risk factor for coronary heart disease (CHD). Fat is stored in the form of TGs in human adipose tissue. We hypothesized that gene co-expression networks in human adipose tissue may be correlated with serum TG levels and help reveal novel genes involved in TG regulation.

Methods

Gene co-expression networks were constructed from two Finnish and one Mexican study sample using the blockwiseModules R function in Weighted Gene Co-expression Network Analysis (WGCNA). Overlap between TG-associated networks from each of the three study samples were calculated using a Fisher’s Exact test. Gene ontology was used to determine known pathways enriched in each TG-associated network.

Results

We measured gene expression in adipose samples from two Finnish and one Mexican study sample. In each study sample, we observed a gene co-expression network that was significantly associated with serum TG levels. The TG modules observed in Finns and Mexicans significantly overlapped and shared 34 genes. Seven of the 34 genes (ARHGAP30, CCR1, CXCL16, FERMT3, HCST, RNASET2, SELPG) were identified as the key hub genes of all three TG modules. Furthermore, two of the 34 genes (ARHGAP9, LST1) reside in previous TG GWAS regions, suggesting them as the regional candidates underlying the GWAS signals.

Conclusions

This study presents a novel adipose gene co-expression network with 34 genes significantly correlated with serum TG across populations.

NetwoRx: connecting drugs to networks and phenotypes in Saccharomyces cerevisiae

Drug modes of action are complex and still poorly understood. The set of known drug targets is widely acknowledged to be biased and incomplete, and so gives only limited insight into the system-wide effects of drugs. But a high-throughput assay unique to yeast-barcode-based chemogenomic screens-can measure the individual drug response of every yeast deletion mutant in parallel. NetwoRx (http://ophid.utoronto.ca/networx) is the first resource to store data from these extremely valuable yeast chemogenomics experiments. In total, NetwoRx stores data on 5924 genes and 466 drugs. In addition, we applied data-mining approaches to identify yeast pathways, functions and phenotypes that are targeted by particular drugs, compute measures of drug-drug similarity and construct drug-phenotype networks. These data are all available to search or download through NetwoRx; users can search by drug name, gene name or gene set identifier. We also set up automated analysis routines in NetwoRx; users can query new gene sets against the entire collection of drug profiles and retrieve the drugs that target them. We demonstrate with use case examples how NetwoRx can be applied to target specific phenotypes, repurpose drugs using mode of action analysis, investigate bipartite networks and predict new drugs that affect yeast aging.

Network analysis of a Pkd1-mouse model of autosomal dominant polycystic kidney disease identifies HNF4α as a disease modifier

Autosomal Dominant Polycystic Kidney Disease (ADPKD; MIM ID's 173900, 601313, 613095) leads to end-stage kidney disease, caused by mutations in PKD1 or PKD2. Inactivation of Pkd1 before or after P13 in mice results in distinct early- or late-onset disease. Using a mouse model of ADPKD carrying floxed Pkd1 alleles and an inducible Cre recombinase, we intensively analyzed the relationship between renal maturation and cyst formation by applying transcriptomics and metabolomics to follow disease progression in a large number of animals induced before P10. Weighted gene co-expression network analysis suggests that Pkd1-cystogenesis does not cause developmental arrest and occurs in the context of gene networks similar to those that regulate/maintain normal kidney morphology/function. Knowledge-based Ingenuity Pathway Analysis (IPA) software identifies HNF4α as a likely network node. These results are further supported by a meta-analysis of 1,114 published gene expression arrays in Pkd1 wild-type tissues. These analyses also predict that metabolic pathways are key elements in postnatal kidney maturation and early steps of cyst formation. Consistent with these findings, urinary metabolomic studies show that Pkd1 cystic mutants have a distinct profile of excreted metabolites, with pathway analysis suggesting altered activity in several metabolic pathways. To evaluate their role in disease, metabolic networks were perturbed by inactivating Hnf4α and Pkd1. The Pkd1/Hnf4α double mutants have significantly more cystic kidneys, thus indicating that metabolic pathways could play a role in Pkd1-cystogenesis.

Autosomal Dominant Polycystic Kidney Disease (ADPKD) is the most common genetic cause of polycystic kidney disease and is responsible for 4.6% of the end-stage renal disease (ESRD) cases in the United States. It is most often caused by mutation in the PKD1 gene. To understand this disease, we made a mouse model in which we could delete the Pkd1 gene and study the animal as its kidney becomes cystic. Using this model, we had previously found that the maturation status of the animal determines whether cysts form within days or within months, and we had narrowed down this switch to a two-day interval. In the current study, we used the rapid cyst-forming model to analyze the expression pattern of thousands of genes in mutant and control kidneys, and metabolites excreted in the urine. Our results identify a number of genes that may be involved in cyst formation and suggest that metabolic changes may play a role in ADPKD and could alter disease progression. These analyses also predict that metabolic pathways are key elements in normal postnatal kidney maturation.

Age-associated changes in gene expression in human brain and isolated neurons

Previous studies have suggested that there are genes whose expression levels are associated with chronological age. However, which genes show consistent age association across studies, and which are specific to a given organism or tissue remains unresolved. Here, we reassessed this question using 2 independently ascertained series of human brain samples from 2 anatomic regions, the frontal lobe of the cerebral cortex and cerebellum. Using microarrays to estimate gene expression, we found 60 associations between expression and chronological age that were statistically significant and were replicated in both series in at least 1 tissue. There were a greater number of significant associations in the frontal cortex compared with the cerebellum. We then repeated the analysis in a subset of samples using laser capture microdissection to isolate Purkinje neurons from the cerebellum. We were able to replicate 5 gene associations from either frontal cortex or cerebellum in the Purkinje cell dataset, suggesting that there is a subset of genes which have robust changes with aging. Of these, the most consistent and strongest association was with expression of RHBDL3, a rhomboid protease family member. We confirmed several hits using an independent technique (quantitative reverse transcriptase polymerase chain reaction) and in an independent published sample series that used a different array platform. We also interrogated larger patterns of age-related gene expression using weighted gene correlation network analysis. We found several modules that showed significant associations with chronological age and, of these, several that showed negative associations were enriched for genes encoding components of mitochondria. Overall, our results show that there is a distinct and reproducible gene signature for aging in the human brain.

Genetic analysis of DNA methylation and gene expression levels in whole blood of healthy human subjects

BACKGROUND

The predominant model for regulation of gene expression through DNA methylation is an inverse association in which increased methylation results in decreased gene expression levels. However, recent studies suggest that the relationship between genetic variation, DNA methylation and expression is more complex.

RESULTS

Systems genetic approaches for examining relationships between gene expression and methylation array data were used to find both negative and positive associations between these levels. A weighted correlation network analysis revealed that i) both transcriptome and methylome are organized in modules, ii) co-expression modules are generally not preserved in the methylation data and vice-versa, and iii) highly significant correlations exist between co-expression and co-methylation modules, suggesting the existence of factors that affect expression and methylation of different modules (i.e., trans effects at the level of modules). We observed that methylation probes associated with expression in cis were more likely to be located outside CpG islands, whereas specificity for CpG island shores was present when methylation, associated with expression, was under local genetic control. A structural equation model based analysis found strong support in particular for a traditional causal model in which gene expression is regulated by genetic variation via DNA methylation instead of gene expression affecting DNA methylation levels.

CONCLUSIONS

Our results provide new insights into the complex mechanisms between genetic markers, epigenetic mechanisms and gene expression. We find strong support for the classical model of genetic variants regulating methylation, which in turn regulates gene expression. Moreover we show that, although the methylation and expression modules differ, they are highly correlated.

Network analysis reveals centrally connected genes and pathways involved in CD8+ T cell exhaustion versus memory

Exhausted CD8(+) T cells function poorly and are negatively regulated by inhibitory receptors. Transcriptional profiling has identified gene expression changes associated with exhaustion. However, the transcriptional pathways critical to the differences between exhausted and functional memory CD8(+) T cells are unclear. We thus defined transcriptional coexpression networks to define pathways centrally involved in exhaustion versus memory. These studies revealed differences between exhausted and memory CD8(+) T cells including the following: lack of coordinated transcriptional modules of quiescence during exhaustion, centrally connected hub genes, pathways such as transcription factors, genes involved in regulation of immune responses, and DNA repair genes, as well as differential connectivity for genes including T-bet, Eomes, and other transcription factors. These data identify pathways involved in CD8(+) T cell exhaustion, and highlight the context-dependent nature of transcription factors in exhaustion versus memory.

Mean Expression of the X-Chromosome is Associated with Neuronal Density

Background: Neurodegenerative diseases are characterized by key features such as loss of neurons, astrocytosis, and microglial activation/proliferation. These changes cause differences in the density of cell types between control and disease subjects, confounding results from gene expression studies. Chromosome X (ChrX) is known to be specifically important in the brain. We hypothesized the existence of a chromosomal signature of gene expression associated with the X-chromosome for neurological conditions not normally associated with that chromosome. The hypothesis was investigated using publicly available microarray datasets from studies on Parkinson’s disease, Alzheimer’s disease, and Huntington’s disease. Data were analyzed using Chromowave, an analytical tool for detecting spatially extended expression changes along chromosomes. To examine associations with neuronal density and astrocytosis, the expression of cell specific reporter genes was extracted. The association between these genes and the expression patterns extracted by Chromowave was then analyzed. Further analyses of the X:Autosome ratios for laser dissected neurons, microglia cultures and whole tissue were performed to detect cell specific differences. Results: We observed an extended pattern of low expression of ChrX consistent in all the neurodegenerative disease brain datasets. There was a strong correlation between mean ChrX expression and the pattern extracted from the autosomal genes representing neurons, but not with mean autosomal expression. No chromosomal patterns associated with the neuron specific genes were found on other chromosomes. The chromosomal expression pattern was not present in datasets from blood cells. The X:Autosome expression ratio was also higher in neuronal cells than in tissues with a mix of cell types. Conclusions: The results suggest that neurological disorders show as a reduction in mean expression of many genes along ChrX. The most likely explanation for this finding relates to the documented general up-regulation of ChrX in brain tissue which, this work suggests, occurs primarily in neurons. If validated, this cell specific ChrX expression warrants further research as understanding the biological reasons and mechanisms for this expression, may help to elucidate a connection with the development of neurodegenerative disorders.