WGCNA: an R package for weighted correlation network analysis

Coronary artery endothelial transcriptome in vivo: identification of endoplasmic reticulum stress and enhanced reactive oxygen species by gene connectivity network analysis

BACKGROUND

Endothelial function is central to the localization of atherosclerosis. The in vivo endothelial phenotypic footprints of arterial bed identity and site-specific atherosusceptibility are addressed.

METHODS AND RESULTS

Ninety-eight endothelial cell samples from 13 discrete coronary and noncoronary arterial regions of varying susceptibilities to atherosclerosis were isolated from 76 normal swine. Transcript profiles were analyzed to determine the steady-state in vivo endothelial phenotypes. An unsupervised systems biology approach using weighted gene coexpression networks showed highly correlated endothelial genes. Connectivity network analysis identified 19 gene modules, 12 of which showed significant association with circulatory bed classification. Differential expression of 1300 genes between coronary and noncoronary artery endothelium suggested distinct coronary endothelial phenotypes, with highest significance expressed in gene modules enriched for biological functions related to endoplasmic reticulum (ER) stress and unfolded protein binding, regulation of transcription and translation, and redox homeostasis. Furthermore, within coronary arteries, comparison of endothelial transcript profiles of susceptible proximal regions to protected distal regions suggested the presence of ER stress conditions in susceptible sites. Accumulation of reactive oxygen species throughout coronary endothelium was greater than in noncoronary endothelium consistent with coronary artery ER stress and lower endothelial expression of antioxidant genes in coronary arteries.

CONCLUSIONS

Gene connectivity analyses discriminated between coronary and noncoronary endothelial transcript profiles and identified differential transcript levels associated with increased ER and oxidative stress in coronary arteries consistent with enhanced susceptibility to atherosclerosis.

Mouse genome-wide association and systems genetics identify Asxl2 as a regulator of bone mineral density and osteoclastogenesis

Significant advances have been made in the discovery of genes affecting bone mineral density (BMD); however, our understanding of its genetic basis remains incomplete. In the current study, genome-wide association (GWA) and co-expression network analysis were used in the recently described Hybrid Mouse Diversity Panel (HMDP) to identify and functionally characterize novel BMD genes. In the HMDP, a GWA of total body, spinal, and femoral BMD revealed four significant associations (-log10P>5.39) affecting at least one BMD trait on chromosomes (Chrs.) 7, 11, 12, and 17. The associations implicated a total of 163 genes with each association harboring between 14 and 112 genes. This list was reduced to 26 functional candidates by identifying those genes that were regulated by local eQTL in bone or harbored potentially functional non-synonymous (NS) SNPs. This analysis revealed that the most significant BMD SNP on Chr. 12 was a NS SNP in the additional sex combs like-2 (Asxl2) gene that was predicted to be functional. The involvement of Asxl2 in the regulation of bone mass was confirmed by the observation that Asxl2 knockout mice had reduced BMD. To begin to unravel the mechanism through which Asxl2 influenced BMD, a gene co-expression network was created using cortical bone gene expression microarray data from the HMDP strains. Asxl2 was identified as a member of a co-expression module enriched for genes involved in the differentiation of myeloid cells. In bone, osteoclasts are bone-resorbing cells of myeloid origin, suggesting that Asxl2 may play a role in osteoclast differentiation. In agreement, the knockdown of Asxl2 in bone marrow macrophages impaired their ability to form osteoclasts. This study identifies a new regulator of BMD and osteoclastogenesis and highlights the power of GWA and systems genetics in the mouse for dissecting complex genetic traits.

Metabonomic, transcriptomic, and genomic variation of a population cohort

Comprehensive characterization of human tissues promises novel insights into the biological architecture of human diseases and traits. We assessed metabonomic, transcriptomic, and genomic variation for a large population-based cohort from the capital region of Finland. Network analyses identified a set of highly correlated genes, the lipid-leukocyte (LL) module, as having a prominent role in over 80 serum metabolites (of 134 measures quantified), including lipoprotein subclasses, lipids, and amino acids. Concurrent association with immune response markers suggested the LL module as a possible link between inflammation, metabolism, and adiposity. Further, genomic variation was used to generate a directed network and infer LL module's largely reactive nature to metabolites. Finally, gene co-expression in circulating leukocytes was shown to be dependent on serum metabolite concentrations, providing evidence for the hypothesis that the coherence of molecular networks themselves is conditional on environmental factors. These findings show the importance and opportunity of systematic molecular investigation of human population samples. To facilitate and encourage this investigation, the metabonomic, transcriptomic, and genomic data used in this study have been made available as a resource for the research community.

Stochastic variation of transcript abundance in C57BL/6J mice

Background

Transcripts can exhibit significant variation in tissue samples from inbred laboratory mice. We have designed and carried out a microarray experiment to examine transcript variation across samples from adipose, heart, kidney, and liver tissues of C57BL/6J mice and to partition variation into within-mouse and between-mouse components. Within-mouse variance captures variation due to heterogeneity of gene expression within tissues, RNA-extraction, and array processing. Between-mouse variance reflects differences in transcript abundance between genetically identical mice.

Results

The nature and extent of transcript variation differs across tissues. Adipose has the largest total variance and the largest within-mouse variance. Liver has the smallest total variance, but it has the most between-mouse variance. Genes with high variability can be classified into groups with correlated patterns of expression that are enriched for specific biological functions. Variation between mice is associated with circadian rhythm, growth hormone signaling, immune response, androgen regulation, lipid metabolism, and the extracellular matrix. Genes showing correlated patterns of within-mouse variation are also associated with biological functions that largely reflect heterogeneity of cell types within tissues.

Conclusions

Genetically identical mice can experience different individual outcomes for medically important traits. Variation in gene expression observed between genetically identical mice can identify functional classes of genes that are likely to vary in the absence of experimental perturbations, can inform experimental design decisions, and provides a baseline for the interpretation of gene expression data in interventional studies. The extent of transcript variation among genetically identical mice underscores the importance of stochastic and micro-environmental factors and their phenotypic consequences.

COSINE: COndition-SpecIfic sub-NEtwork identification using a global optimization method

MOTIVATION

The identification of condition specific sub-networks from gene expression profiles has important biological applications, ranging from the selection of disease-related biomarkers to the discovery of pathway alterations across different phenotypes. Although many methods exist for extracting these sub-networks, very few existing approaches simultaneously consider both the differential expression of individual genes and the differential correlation of gene pairs, losing potentially valuable information in the data.

RESULTS

In this article, we propose a new method, COSINE (COndition SpecIfic sub-NEtwork), which employs a scoring function that jointly measures the condition-specific changes of both 'nodes' (individual genes) and 'edges' (gene-gene co-expression). It uses the genetic algorithm to search for the single optimal sub-network which maximizes the scoring function. We applied COSINE to both simulated datasets with various differential expression patterns, and three real datasets, one prostate cancer dataset, a second one from the across-tissue comparison of morbidly obese patients and the other from the across-population comparison of the HapMap samples. Compared with previous methods, COSINE is more powerful in identifying truly significant sub-networks of appropriate size and meaningful biological relevance.

AVAILABILITY

The R code is available as the COSINE package on CRAN: http://cran.r-project.org/web/packages/COSINE/index.html.

Gene networks associated with conditional fear in mice identified using a systems genetics approach

Background

Our understanding of the genetic basis of learning and memory remains shrouded in mystery. To explore the genetic networks governing the biology of conditional fear, we used a systems genetics approach to analyze a hybrid mouse diversity panel (HMDP) with high mapping resolution.

Results

A total of 27 behavioral quantitative trait loci were mapped with a false discovery rate of 5%. By integrating fear phenotypes, transcript profiling data from hippocampus and striatum and also genotype information, two gene co-expression networks correlated with context-dependent immobility were identified. We prioritized the key markers and genes in these pathways using intramodular connectivity measures and structural equation modeling. Highly connected genes in the context fear modules included Psmd6, Ube2a and Usp33, suggesting an important role for ubiquitination in learning and memory. In addition, we surveyed the architecture of brain transcript regulation and demonstrated preservation of gene co-expression modules in hippocampus and striatum, while also highlighting important differences. Rps15a, Kif3a, Stard7, 6330503K22RIK, and Plvap were among the individual genes whose transcript abundance were strongly associated with fear phenotypes.

Conclusion

Application of our multi-faceted mapping strategy permits an increasingly detailed characterization of the genetic networks underlying behavior.

Global developmental gene expression and pathway analysis of normal brain development and mouse models of human neuronal migration defects

Heterozygous LIS1 mutations are the most common cause of human lissencephaly, a human neuronal migration defect, and DCX mutations are the most common cause of X-linked lissencephaly. LIS1 is part of a protein complex including NDEL1 and 14-3-3ε that regulates dynein motor function and microtubule dynamics, while DCX stabilizes microtubules and cooperates with LIS1 during neuronal migration and neurogenesis. Targeted gene mutations of Lis1, Dcx, Ywhae (coding for 14-3-3ε), and Ndel1 lead to neuronal migration defects in mouse and provide models of human lissencephaly, as well as aid the study of related neuro-developmental diseases. Here we investigated the developing brain of these four mutants and wild-type mice using expression microarrays, bioinformatic analyses, and in vivo/in vitro experiments to address whether mutations in different members of the LIS1 neuronal migration complex lead to similar and/or distinct global gene expression alterations. Consistent with the overall successful development of the mutant brains, unsupervised clustering and co-expression analysis suggested that cell cycle and synaptogenesis genes are similarly expressed and co-regulated in WT and mutant brains in a time-dependent fashion. By contrast, focused co-expression analysis in the Lis1 and Ndel1 mutants uncovered substantial differences in the correlation among pathways. Differential expression analysis revealed that cell cycle, cell adhesion, and cytoskeleton organization pathways are commonly altered in all mutants, while synaptogenesis, cell morphology, and inflammation/immune response are specifically altered in one or more mutants. We found several commonly dysregulated genes located within pathogenic deletion/duplication regions, which represent novel candidates of human mental retardation and neurocognitive disabilities. Our analysis suggests that gene expression and pathway analysis in mouse models of a similar disorder or within a common pathway can be used to define novel candidates for related human diseases.

Is my network module preserved and reproducible?

In many applications, one is interested in determining which of the properties of a network module change across conditions. For example, to validate the existence of a module, it is desirable to show that it is reproducible (or preserved) in an independent test network. Here we study several types of network preservation statistics that do not require a module assignment in the test network. We distinguish network preservation statistics by the type of the underlying network. Some preservation statistics are defined for a general network (defined by an adjacency matrix) while others are only defined for a correlation network (constructed on the basis of pairwise correlations between numeric variables). Our applications show that the correlation structure facilitates the definition of particularly powerful module preservation statistics. We illustrate that evaluating module preservation is in general different from evaluating cluster preservation. We find that it is advantageous to aggregate multiple preservation statistics into summary preservation statistics. We illustrate the use of these methods in six gene co-expression network applications including 1) preservation of cholesterol biosynthesis pathway in mouse tissues, 2) comparison of human and chimpanzee brain networks, 3) preservation of selected KEGG pathways between human and chimpanzee brain networks, 4) sex differences in human cortical networks, 5) sex differences in mouse liver networks. While we find no evidence for sex specific modules in human cortical networks, we find that several human cortical modules are less preserved in chimpanzees. In particular, apoptosis genes are differentially co-expressed between humans and chimpanzees. Our simulation studies and applications show that module preservation statistics are useful for studying differences between the modular structure of networks. Data, R software and accompanying tutorials can be downloaded from the following webpage: http://www.genetics.ucla.edu/labs/horvath/CoexpressionNetwork/ModulePreservation.

In network applications, one is often interested in studying whether modules are preserved across multiple networks. For example, to determine whether a pathway of genes is perturbed in a certain condition, one can study whether its connectivity pattern is no longer preserved. Non-preserved modules can either be biologically uninteresting (e.g., reflecting data outliers) or interesting (e.g., reflecting sex specific modules). An intuitive approach for studying module preservation is to cross-tabulate module membership. But this approach often cannot address questions about the preservation of connectivity patterns between nodes. Thus, cross-tabulation based approaches often fail to recognize that important aspects of a network module are preserved. Cross-tabulation methods make it difficult to argue that a module is not preserved. The weak statement (“the reference module does not overlap with any of the identified test set modules”) is less relevant in practice than the strong statement (“the module cannot be found in the test network irrespective of the parameter settings of the module detection procedure”). Module preservation statistics have important applications, e.g. we show that the wiring of apoptosis genes in a human cortical network differs from that in chimpanzees.

Large-scale prediction of long non-coding RNA functions in a coding-non-coding gene co-expression network

Although accumulating evidence has provided insight into the various functions of long-non-coding RNAs (lncRNAs), the exact functions of the majority of such transcripts are still unknown. Here, we report the first computational annotation of lncRNA functions based on public microarray expression profiles. A coding–non-coding gene co-expression (CNC) network was constructed from re-annotated Affymetrix Mouse Genome Array data. Probable functions for altogether 340 lncRNAs were predicted based on topological or other network characteristics, such as module sharing, association with network hubs and combinations of co-expression and genomic adjacency. The functions annotated to the lncRNAs mainly involve organ or tissue development (e.g. neuron, eye and muscle development), cellular transport (e.g. neuronal transport and sodium ion, acid or lipid transport) or metabolic processes (e.g. involving macromolecules, phosphocreatine and tyrosine).

Discovery approaches to UPR in athero-susceptible endothelium in vivo

The endothelium is a monolayer of cells that lines the entire inner surface of the cardiovascular and lymphatic circulations where it controls normal physiological functions through both systemic and local regulation. Endothelial phenotypes are heterogeneous, dynamic and malleable, properties that in large- and medium-sized arteries lead to a central role in the development of focal and regional atherosclerosis. The endothelial phenotype in athero-susceptible sites is different from that in nearby athero-resistant regions. Understanding the in vivo gene, protein, and metabolic expression profiles of susceptible endothelium is, therefore, an important spatiotemporal challenge in atherosclerosis research. Recent studies have demonstrated that endoplasmic reticulum (ER) stress and the UPR are characteristics of susceptible endothelium. Here, we outline global genomic profiling, pathway analyses, and gene connectivity approaches to the identification of UPR and associated pathways as discrete markers of athero-susceptibility in arterial endothelium.

Functional genomics of the brain: uncovering networks in the CNS using a systems approach

The central nervous system (CNS) is undoubtedly the most complex human organ system in terms of its diverse functions, cellular composition, and connections. Attempts to capture this diversity experimentally were the foundation on which the field of neurobiology was built. Until now though, techniques were either painstakingly slow or insufficient in capturing this heterogeneity. In addition, the combination of multiple layers of information needed for a complete picture of neuronal diversity from the epigenome to the proteome requires an even more complex compilation of data. In this era of high-throughput genomics though, the ability to isolate and profile neurons and brain tissue has increased tremendously and now requires less effort. Both microarrays and next-generation sequencing have identified neuronal transcriptomes and signaling networks involved in normal brain development, as well as in disease. However, the expertise needed to organize and prioritize the resultant data remains substantial. A combination of supervised organization and unsupervised analyses are needed to fully appreciate the underlying structure in these datasets. When utilized effectively, these analyses have yielded striking insights into a number of fundamental questions in neuroscience on topics ranging from the evolution of the human brain to neuropsychiatric and neurodegenerative disorders. Future studies will incorporate these analyses with behavioral and physiological data from patients to more efficiently move toward personalized therapeutics.

Integrated genomics of susceptibility to alkylator-induced leukemia in mice

Background

Therapy-related acute myeloid leukemia (t-AML) is a secondary, generally incurable, malignancy attributable to chemotherapy exposure. Although there is a genetic component to t-AML susceptibility in mice, the relevant loci and the mechanism(s) by which they contribute to t-AML are largely unknown. An improved understanding of susceptibility factors and the biological processes in which they act may lead to the development of t-AML prevention strategies.

Results

In this work we applied an integrated genomics strategy in inbred strains of mice to find novel factors that might contribute to susceptibility. We found that the pre-exposure transcriptional state of hematopoietic stem/progenitor cells predicts susceptibility status. More than 900 genes were differentially expressed between susceptible and resistant strains and were highly enriched in the apoptotic program, but it remained unclear which genes, if any, contribute directly to t-AML susceptibility. To address this issue, we integrated gene expression data with genetic information, including single nucleotide polymorphisms (SNPs) and DNA copy number variants (CNVs), to identify genetic networks underlying t-AML susceptibility. The 30 t-AML susceptibility networks we found are robust: they were validated in independent, previously published expression data, and different analytical methods converge on them. Further, the networks are enriched in genes involved in cell cycle and DNA repair (pathways not discovered in traditional differential expression analysis), suggesting that these processes contribute to t-AML susceptibility. Within these networks, the putative regulators (e.g., Parp2, Casp9, Polr1b) are the most likely to have a non-redundant role in the pathogenesis of t-AML. While identifying these networks, we found that current CNVR and SNP-based haplotype maps in mice represented distinct sources of genetic variation contributing to expression variation, implying that mapping studies utilizing either source alone will have reduced sensitivity.

Conclusion

The identification and prioritization of genes and networks not previously implicated in t-AML generates novel hypotheses on the biology and treatment of this disease that will be the focus of future research.

Systems biology of ovine intestinal parasite resistance: disease gene modules and biomarkers

This study reports on the molecular systems biology of gastrointestinal nematode (GIN) infection and potential biomarkers for GIN resistance in sheep. Microarray gene expression data were obtained for 3 different tissues at 4 time points from sheep artificially challenged with two types of nematodes, Haemonchus contortus (HC) and Trichostrongylus colubriformis (TC). We employed an integrated systems biology approach, integrating 3 main methods: standard differential gene expression analyses, weighted gene co-expression network analyses (WGCNA) and quantitative genetic analyses of gene expression traits of key biomarkers. Using standard differential gene expression analyses we identified differentially expressed genes (DE) which responded differently in sheep challenged with HC compared to those challenged with TC. These interaction genes (e.g. MRPL51, SMEK2, CAT, MAPK1IP1 and SLC25A20A) were enriched in Wnt receptor signalling pathway (p = 0.0132) and positive regulation of NFκβ transcription factor activity (p = 0.00208). We report FCER1A, a gene encoding a high-affinity receptor for the Fc region of immunoglobulin E, which is linked to innate immunity to GIN in sheep. Using weighted gene co-expression network analysis (WGCNA) methods, we identified gene modules that were correlated with the length of infection (disease modules). Hub genes (with high intramodular connectivity) were filtered further to identify biomarkers that are related to the length of infection (e.g. CAT, FBX033, COL15A1, IGFBP7, FBLN1 and IgCgamma). The biomarkers we found in HC networks were significantly associated with functions such as T-cell and B-cell regulations, TNF-alpha, interleukin and cytokine production. In TC networks, biomarkers were significantly associated with functions such as protein catabolic process, heat shock protein binding, protein targeting and localization, cytokine receptor binding, TNF receptor binding, apoptosis and IGF binding. These results provide specific gene targets for therapeutic interventions and provide insights into GIN infections in sheep which may be used to infer the same in related host species. This is also the first study to apply the concept of estimating breeding values of animals to expression traits and reveals 11 heritable candidate biomarkers (0.05 to 0.92) that could be used in selection of animals for GIN resistance.

Identification of a gene module associated with BMD through the integration of network analysis and genome-wide association data

Bone mineral density (BMD) is influenced by a complex network of gene interactions; therefore, elucidating the relationships between genes and how those genes, in turn, influence BMD is critical for developing a comprehensive understanding of osteoporosis. To investigate the role of transcriptional networks in the regulation of BMD, we performed a weighted gene coexpression network analysis (WGCNA) using microarray expression data on monocytes from young individuals with low or high BMD. WGCNA groups genes into modules based on patterns of gene coexpression. and our analysis identified 11 gene modules. We observed that the overall expression of one module (referred to as module 9) was significantly higher in the low-BMD group (p = .03). Module 9 was highly enriched for genes belonging to the immune system-related gene ontology (GO) category "response to virus" (p = 7.6 × 10(-11)). Using publically available genome-wide association study data, we independently validated the importance of module 9 by demonstrating that highly connected module 9 hubs were more likely, relative to less highly connected genes, to be genetically associated with BMD. This study highlights the advantages of systems-level analyses to uncover coexpression modules associated with bone mass and suggests that particular monocyte expression patterns may mediate differences in BMD.

Is human blood a good surrogate for brain tissue in transcriptional studies?

BACKGROUND

Since human brain tissue is often unavailable for transcriptional profiling studies, blood expression data is frequently used as a substitute. The underlying hypothesis in such studies is that genes expressed in brain tissue leave a transcriptional footprint in blood. We tested this hypothesis by relating three human brain expression data sets (from cortex, cerebellum and caudate nucleus) to two large human blood expression data sets (comprised of 1463 individuals).

RESULTS

We found mean expression levels were weakly correlated between the brain and blood data (r range: [0.24,0.32]). Further, we tested whether co-expression relationships were preserved between the three brain regions and blood. Only a handful of brain co-expression modules showed strong evidence of preservation and these modules could be combined into a single large blood module. We also identified highly connected intramodular "hub" genes inside preserved modules. These preserved intramodular hub genes had the following properties: first, their expression levels tended to be significantly more heritable than those from non-preserved intramodular hub genes (p < 10⁻⁹⁰); second, they had highly significant positive correlations with the following cluster of differentiation genes: CD58, CD47, CD48, CD53 and CD164; third, a significant number of them were known to be involved in infection mechanisms, post-transcriptional and post-translational modification and other basic processes.

CONCLUSIONS

Overall, we find transcriptome organization is poorly preserved between brain and blood. However, the subset of preserved co-expression relationships characterized here may aid future efforts to identify blood biomarkers for neurological and neuropsychiatric diseases when brain tissue samples are unavailable.

Genetic diversity and striatal gene networks: focus on the heterogeneous stock-collaborative cross (HS-CC) mouse

BACKGROUND

The current study focused on the extent genetic diversity within a species (Mus musculus) affects gene co-expression network structure. To examine this issue, we have created a new mouse resource, a heterogeneous stock (HS) formed from the same eight inbred strains that have been used to create the collaborative cross (CC). The eight inbred strains capture > 90% of the genetic diversity available within the species. For contrast with the HS-CC, a C57BL/6J (B6) × DBA/2J (D2) F2 intercross and the HS4, derived from crossing the B6, D2, BALB/cJ and LP/J strains, were used. Brain (striatum) gene expression data were obtained using the Illumina Mouse WG 6.1 array, and the data sets were interrogated using a weighted gene co-expression network analysis (WGCNA).

RESULTS

Genes reliably detected as expressed were similar in all three data sets as was the variability of expression. As measured by the WGCNA, the modular structure of the transcriptome networks was also preserved both on the basis of module assignment and from the perspective of the topological overlap maps. Details of the HS-CC gene modules are provided; essentially identical results were obtained for the HS4 and F2 modules. Gene ontology annotation of the modules revealed a significant overrepresentation in some modules for neuronal processes, e.g., central nervous system development. Integration with known protein-protein interactions data indicated significant enrichment among co-expressed genes. We also noted significant overlap with markers of central nervous system cell types (neurons, oligodendrocytes and astrocytes). Using the Allen Brain Atlas, we found evidence of spatial co-localization within the striatum for several modules. Finally, for some modules it was possible to detect an enrichment of transcription binding sites. The binding site for Wt1, which is associated with neurodegeneration, was the most significantly overrepresented.

CONCLUSIONS

Despite the marked differences in genetic diversity, the transcriptome structure was remarkably similar for the F2, HS4 and HS-CC. These data suggest that it should be possible to integrate network data from simple and complex crosses. A careful examination of the HS-CC transcriptome revealed the expected structure for striatal gene expression. Importantly, we demonstrate the integration of anatomical and network expression data.

DiffCoEx: a simple and sensitive method to find differentially coexpressed gene modules

BACKGROUND

Large microarray datasets have enabled gene regulation to be studied through coexpression analysis. While numerous methods have been developed for identifying differentially expressed genes between two conditions, the field of differential coexpression analysis is still relatively new. More specifically, there is so far no sensitive and untargeted method to identify gene modules (also known as gene sets or clusters) that are differentially coexpressed between two conditions. Here, sensitive and untargeted means that the method should be able to construct de novo modules by grouping genes based on shared, but subtle, differential correlation patterns.

RESULTS

We present DiffCoEx, a novel method for identifying correlation pattern changes, which builds on the commonly used Weighted Gene Coexpression Network Analysis (WGCNA) framework for coexpression analysis. We demonstrate its usefulness by identifying biologically relevant, differentially coexpressed modules in a rat cancer dataset.

CONCLUSIONS

DiffCoEx is a simple and sensitive method to identify gene coexpression differences between multiple conditions.

Information technology solutions for integration of biomolecular and clinical data in the identification of new cancer biomarkers and targets for therapy

The quest for new cancer biomarkers and targets for therapy requires not only the aggregation and analysis of heterogeneous biomolecular data but also integration of clinical data. In this review we highlight information technology solutions for the integration of biomolecular and clinical data and focus on a solution at the departmental level, i.e., decentralized and medium-scale solution for groups of labs working on a specific topic. Both, hardware and software requirements are described as well as bioinformatics methods and tools for the data analysis. The highlighted IT solutions include storage architecture, high-performance computing, and application servers. Additionally, following computational approaches for data integration are reviewed: data aggregation, integrative data analysis including methodological aspects as well as examples, biomolecular pathways and network reconstruction, and mathematical modelling. Finally, a case study in cancer immunology including the used computational methods is shown, demonstrating how IT solutions for integrating biomolecular and clinical data can help to identify new cancer biomarkers for improving diagnosis and predicting clinical outcome.

An immune response network associated with blood lipid levels

While recent scans for genetic variation associated with human disease have been immensely successful in uncovering large numbers of loci, far fewer studies have focused on the underlying pathways of disease pathogenesis. Many loci which are associated with disease and complex phenotypes map to non-coding, regulatory regions of the genome, indicating that modulation of gene transcription plays a key role. Thus, this study generated genome-wide profiles of both genetic and transcriptional variation from the total blood extracts of over 500 randomly-selected, unrelated individuals. Using measurements of blood lipids, key players in the progression of atherosclerosis, three levels of biological information are integrated in order to investigate the interactions between circulating leukocytes and proximal lipid compounds. Pair-wise correlations between gene expression and lipid concentration indicate a prominent role for basophil granulocytes and mast cells, cell types central to powerful allergic and inflammatory responses. Network analysis of gene co-expression showed that the top associations function as part of a single, previously unknown gene module, the Lipid Leukocyte (LL) module. This module replicated in T cells from an independent cohort while also displaying potential tissue specificity. Further, genetic variation driving LL module expression included the single nucleotide polymorphism (SNP) most strongly associated with serum immunoglobulin E (IgE) levels, a key antibody in allergy. Structural Equation Modeling (SEM) indicated that LL module is at least partially reactive to blood lipid levels. Taken together, this study uncovers a gene network linking blood lipids and circulating cell types and offers insight into the hypothesis that the inflammatory response plays a prominent role in metabolism and the potential control of atherogenesis.

The association of multiple interacting genes with specific phenotypes in rice using gene coexpression networks

Discovering gene sets underlying the expression of a given phenotype is of great importance, as many phenotypes are the result of complex gene-gene interactions. Gene coexpression networks, built using a set of microarray samples as input, can help elucidate tightly coexpressed gene sets (modules) that are mixed with genes of known and unknown function. Functional enrichment analysis of modules further subdivides the coexpressed gene set into cofunctional gene clusters that may coexist in the module with other functionally related gene clusters. In this study, 45 coexpressed gene modules and 76 cofunctional gene clusters were discovered for rice (Oryza sativa) using a global, knowledge-independent paradigm and the combination of two network construction methodologies. Some clusters were enriched for previously characterized mutant phenotypes, providing evidence for specific gene sets (and their annotated molecular functions) that underlie specific phenotypes.

Identifying modules of coexpressed transcript units and their organization of Saccharopolyspora erythraea from time series gene expression profiles

Background

The Saccharopolyspora erythraea genome sequence was released in 2007. In order to look at the gene regulations at whole transcriptome level, an expression microarray was specifically designed on the S. erythraea strain NRRL 2338 genome sequence. Based on these data, we set out to investigate the potential transcriptional regulatory networks and their organization.

Methodology/Principal Findings

In view of the hierarchical structure of bacterial transcriptional regulation, we constructed a hierarchical coexpression network at whole transcriptome level. A total of 27 modules were identified from 1255 differentially expressed transcript units (TUs) across time course, which were further classified in to four groups. Functional enrichment analysis indicated the biological significance of our hierarchical network. It was indicated that primary metabolism is activated in the first rapid growth phase (phase A), and secondary metabolism is induced when the growth is slowed down (phase B). Among the 27 modules, two are highly correlated to erythromycin production. One contains all genes in the erythromycin-biosynthetic (ery) gene cluster and the other seems to be associated with erythromycin production by sharing common intermediate metabolites. Non-concomitant correlation between production and expression regulation was observed. Especially, by calculating the partial correlation coefficients and building the network based on Gaussian graphical model, intrinsic associations between modules were found, and the association between those two erythromycin production-correlated modules was included as expected.

Conclusions

This work created a hierarchical model clustering transcriptome data into coordinated modules, and modules into groups across the time course, giving insight into the concerted transcriptional regulations especially the regulation corresponding to erythromycin production of S. erythraea. This strategy may be extendable to studies on other prokaryotic microorganisms.

Divergence of human and mouse brain transcriptome highlights Alzheimer disease pathways

Because mouse models play a crucial role in biomedical research related to the human nervous system, understanding the similarities and differences between mouse and human brain is of fundamental importance. Studies comparing transcription in human and mouse have come to varied conclusions, in part because of their relatively small sample sizes or underpowered methodologies. To better characterize gene expression differences between mouse and human, we took a systems-biology approach by using weighted gene coexpression network analysis on more than 1,000 microarrays from brain. We find that global network properties of the brain transcriptome are highly preserved between species. Furthermore, all modules of highly coexpressed genes identified in mouse were identified in human, with those related to conserved cellular functions showing the strongest between-species preservation. Modules corresponding to glial and neuronal cells were sufficiently preserved between mouse and human to permit identification of cross species cell-class marker genes. We also identify several robust human-specific modules, including one strongly correlated with measures of Alzheimer disease progression across multiple data sets, whose hubs are poorly-characterized genes likely involved in Alzheimer disease. We present multiple lines of evidence suggesting links between neurodegenerative disease and glial cell types in human, including human-specific correlation of presenilin-1 with oligodendrocyte markers, and significant enrichment for known neurodegenerative disease genes in microglial modules. Together, this work identifies convergent and divergent pathways in mouse and human, and provides a systematic framework that will be useful for understanding the applicability of mouse models for human brain disorders.

Interspecific resources: a major tool for quantitative trait locus cloning and speciation research

Positional cloning of the quantitative trait locus (QTL) still encounters numerous difficulties, which explains why thousands of QTL have been mapped, while only a few have been identified at the molecular level. Here, we focus on a specific mapping tool that exists in plant and animal model species: interspecific recombinant congenic strains (IRCSs) or interspecific nearly isogenic lines (NILs). Such panels exhibit a much higher sequence diversity than intraspecific sets, thus enhancing the contrasts between phenotypes. In animals, it allows statistical significance to be reached even when using a limited number of individuals. Therefore, we argue that interspecific resources may constitute a major genetic tool for positional cloning and for understanding some bases of speciation mechanisms.

ToppCluster: a multiple gene list feature analyzer for comparative enrichment clustering and network-based dissection of biological systems

ToppCluster is a web server application that leverages a powerful enrichment analysis and underlying data environment for comparative analyses of multiple gene lists. It generates heatmaps or connectivity networks that reveal functional features shared or specific to multiple gene lists. ToppCluster uses hypergeometric tests to obtain list-specific feature enrichment P-values for currently 17 categories of annotations of human-ortholog genes, and provides user-selectable cutoffs and multiple testing correction methods to control false discovery. Each nameable gene list represents a column input to a resulting matrix whose rows are overrepresented features, and individual cells per-list P-values and corresponding genes per feature. ToppCluster provides users with choices of tabular outputs, hierarchical clustering and heatmap generation, or the ability to interactively select features from the functional enrichment matrix to be transformed into XGMML or GEXF network format documents for use in Cytoscape or Gephi applications, respectively. Here, as example, we demonstrate the ability of ToppCluster to enable identification of list-specific phenotypic and regulatory element features (both cis-elements and 3'UTR microRNA binding sites) among tissue-specific gene lists. ToppCluster's functionalities enable the identification of specialized biological functions and regulatory networks and systems biology-based dissection of biological states. ToppCluster can be accessed freely at http://toppcluster.cchmc.org.

Gene expression profiling in C57BL/6J and A/J mouse inbred strains reveals gene networks specific for brain regions independent of genetic background

BACKGROUND

We performed gene expression profiling of the amygdala and hippocampus taken from inbred mouse strains C57BL/6J and A/J. The selected brain areas are implicated in neurobehavioral traits while these mouse strains are known to differ widely in behavior. Consequently, we hypothesized that comparing gene expression profiles for specific brain regions in these strains might provide insight into the molecular mechanisms of human neuropsychiatric traits. We performed a whole-genome gene expression experiment and applied a systems biology approach using weighted gene co-expression network analysis.

RESULTS

We were able to identify modules of co-expressed genes that distinguish a strain or brain region. Analysis of the networks that are most informative for hippocampus and amygdala revealed enrichment in neurologically, genetically and psychologically related pathways. Close examination of the strain-specific gene expression profiles, however, revealed no functional relevance but a significant enrichment of single nucleotide polymorphisms in the probe sequences used for array hybridization. This artifact was not observed for the modules of co-expressed genes that distinguish amygdala and hippocampus.

CONCLUSIONS

The brain-region specific modules were found to be independent of genetic background and are therefore likely to represent biologically relevant molecular networks that can be studied to complement our knowledge about pathways in neuropsychiatric disease.

Robust gene network analysis reveals alteration of the STAT5a network as a hallmark of prostate cancer

We develop a general method to identify gene networks from pair-wise correlations between genes in a microarray data set and apply it to a public prostate cancer gene expression data from 69 primary prostate tumors. We define the degree of a node as the number of genes significantly associated with the node and identify hub genes as those with the highest degree. The correlation network was pruned using transcription factor binding information in VisANT (http://visant.bu.edu/) as a biological filter. The reliability of hub genes was determined using a strict permutation test. Separate networks for normal prostate samples, and prostate cancer samples from African Americans (AA) and European Americans (EA) were generated and compared. We found that the same hubs control disease progression in AA and EA networks. Combining AA and EA samples, we generated networks for low low (<7) and high (≥7) Gleason grade tumors. A comparison of their major hubs with those of the network for normal samples identified two types of changes associated with disease: (i) Some hub genes increased their degree in the tumor network compared to their degree in the normal network, suggesting that these genes are associated with gain of regulatory control in cancer (e.g. possible turning on of oncogenes). (ii) Some hubs reduced their degree in the tumor network compared to their degree in the normal network, suggesting that these genes are associated with loss of regulatory control in cancer (e.g. possible loss of tumor suppressor genes). A striking result was that for both AA and EA tumor samples, STAT5a, CEBPB and EGR1 are major hubs that gain neighbors compared to the normal prostate network. Conversely, HIF-lα is a major hub that loses connections in the prostate cancer network compared to the normal prostate network. We also find that the degree of these hubs changes progressively from normal to low grade to high grade disease, suggesting that these hubs are master regulators of prostate cancer and marks disease progression. STAT5a was identified as a central hub, with ~120 neighbors in the prostate cancer network and only 81 neighbors in the normal prostate network. Of the 120 neighbors of STAT5a, 57 are known cancer related genes, known to be involved in functional pathways associated with tumorigenesis. Our method is general and can easily be extended to identify and study networks associated with any two phenotypes.

A systems genetics approach implicates USF1, FADS3, and other causal candidate genes for familial combined hyperlipidemia

We hypothesized that a common SNP in the 3' untranslated region of the upstream transcription factor 1 (USF1), rs3737787, may affect lipid traits by influencing gene expression levels, and we investigated this possibility utilizing the Mexican population, which has a high predisposition to dyslipidemia. We first associated rs3737787 genotypes in Mexican Familial Combined Hyperlipidemia (FCHL) case/control fat biopsies, with global expression patterns. To identify sets of co-expressed genes co-regulated by similar factors such as transcription factors, genetic variants, or environmental effects, we utilized weighted gene co-expression network analysis (WGCNA). Through WGCNA in the Mexican FCHL fat biopsies we identified two significant Triglyceride (TG)-associated co-expression modules. One of these modules was also associated with FCHL, the other FCHL component traits, and rs3737787 genotypes. This USF1-regulated FCHL-associated (URFA) module was enriched for genes involved in lipid metabolic processes. Using systems genetics procedures we identified 18 causal candidate genes in the URFA module. The FCHL causal candidate gene fatty acid desaturase 3 (FADS3) was associated with TGs in a recent Caucasian genome-wide significant association study and we replicated this association in Mexican FCHL families. Based on a USF1-regulated FCHL-associated co-expression module and SNP rs3737787, we identify a set of causal candidate genes for FCHL-related traits. We then provide evidence from two independent datasets supporting FADS3 as a causal gene for FCHL and elevated TGs in Mexicans.

By integrating a genetic polymorphism with genome-wide gene expression levels, we were able to attribute function to a genetic polymorphism in the USF1 gene. The USF1 gene has previously been associated with a common dyslipidemia, FCHL. FCHL is characterized by elevated levels of total cholesterol, triglycerides, or both. We demonstrate that this genetic polymorphism in USF1 contributes to FCHL disease risk by modulating the expression of a group of genes functionally related to lipid metabolism, and that this modulation is mediated by USF1. One of the genes whose expression is modulated by USF1 is FADS3, which was also implicated in a recent genome-wide association study for lipid traits. We demonstrated that a genetic polymorphism from the FADS3 region, which was associated with triglycerides in a GWAS study of Caucasians, was also associated with triglycerides in Mexican FCHL families. Our analysis provides novel insight into the gene expression profile contributing to FCHL disease risk, and identifies FADS3 as a new gene for FCHL in Mexicans.

Weighted gene co-expression network analysis of the peripheral blood from Amyotrophic Lateral Sclerosis patients

Background

Amyotrophic Lateral Sclerosis (ALS) is a lethal disorder characterized by progressive degeneration of motor neurons in the brain and spinal cord. Diagnosis is mainly based on clinical symptoms, and there is currently no therapy to stop the disease or slow its progression. Since access to spinal cord tissue is not possible at disease onset, we investigated changes in gene expression profiles in whole blood of ALS patients.

Results

Our transcriptional study showed dramatic changes in blood of ALS patients; 2,300 probes (9.4%) showed significant differential expression in a discovery dataset consisting of 30 ALS patients and 30 healthy controls. Weighted gene co-expression network analysis (WGCNA) was used to find disease-related networks (modules) and disease related hub genes. Two large co-expression modules were found to be associated with ALS. Our findings were replicated in a second (30 patients and 30 controls) and third dataset (63 patients and 63 controls), thereby demonstrating a highly significant and consistent association of two large co-expression modules with ALS disease status. Ingenuity Pathway Analysis of the ALS related module genes implicates enrichment of functional categories related to genetic disorders, neurodegeneration of the nervous system and inflammatory disease. The ALS related modules contain a number of candidate genes possibly involved in pathogenesis of ALS.

Conclusion

This first large-scale blood gene expression study in ALS observed distinct patterns between cases and controls which may provide opportunities for biomarker development as well as new insights into the molecular mechanisms of the disease.

The genetics of quantitative traits: challenges and prospects

A major challenge in current biology is to understand the genetic basis of variation for quantitative traits. We review the principles of quantitative trait locus mapping and summarize insights about the genetic architecture of quantitative traits that have been obtained over the past decades. We are currently in the midst of a genomic revolution, which enables us to incorporate genetic variation in transcript abundance and other intermediate molecular phenotypes into a quantitative trait locus mapping framework. This systems genetics approach enables us to understand the biology inside the 'black box' that lies between genotype and phenotype in terms of causal networks of interacting genes.

Signed weighted gene co-expression network analysis of transcriptional regulation in murine embryonic stem cells

Background

Recent work has revealed that a core group of transcription factors (TFs) regulates the key characteristics of embryonic stem (ES) cells: pluripotency and self-renewal. Current efforts focus on identifying genes that play important roles in maintaining pluripotency and self-renewal in ES cells and aim to understand the interactions among these genes. To that end, we investigated the use of unsigned and signed network analysis to identify pluripotency and differentiation related genes.

Results

We show that signed networks provide a better systems level understanding of the regulatory mechanisms of ES cells than unsigned networks, using two independent murine ES cell expression data sets. Specifically, using signed weighted gene co-expression network analysis (WGCNA), we found a pluripotency module and a differentiation module, which are not identified in unsigned networks. We confirmed the importance of these modules by incorporating genome-wide TF binding data for key ES cell regulators. Interestingly, we find that the pluripotency module is enriched with genes related to DNA damage repair and mitochondrial function in addition to transcriptional regulation. Using a connectivity measure of module membership, we not only identify known regulators of ES cells but also show that Mrpl15, Msh6, Nrf1, Nup133, Ppif, Rbpj, Sh3gl2, and Zfp39, among other genes, have important roles in maintaining ES cell pluripotency and self-renewal. We also report highly significant relationships between module membership and epigenetic modifications (histone modifications and promoter CpG methylation status), which are known to play a role in controlling gene expression during ES cell self-renewal and differentiation.

Conclusion

Our systems biologic re-analysis of gene expression, transcription factor binding, epigenetic and gene ontology data provides a novel integrative view of ES cell biology.

Network module detection: Affinity search technique with the multi-node topological overlap measure

Background

Many clustering procedures only allow the user to input a pairwise dissimilarity or distance measure between objects. We propose a clustering method that can input a multi-point dissimilarity measure d(i1, i2, ..., iP) where the number of points P can be larger than 2. The work is motivated by gene network analysis where clusters correspond to modules of highly interconnected nodes. Here, we define modules as clusters of network nodes with high multi-node topological overlap. The topological overlap measure is a robust measure of interconnectedness which is based on shared network neighbors. In previous work, we have shown that the multi-node topological overlap measure yields biologically meaningful results when used as input of network neighborhood analysis.

Findings

We adapt network neighborhood analysis for the use of module detection. We propose the Module Affinity Search Technique (MAST), which is a generalized version of the Cluster Affinity Search Technique (CAST). MAST can accommodate a multi-node dissimilarity measure. Clusters grow around user-defined or automatically chosen seeds (e.g. hub nodes). We propose both local and global cluster growth stopping rules. We use several simulations and a gene co-expression network application to argue that the MAST approach leads to biologically meaningful results. We compare MAST with hierarchical clustering and partitioning around medoid clustering.

Conclusion

Our flexible module detection method is implemented in the MTOM software which can be downloaded from the following webpage:

Enabling a systems biology approach to immunology: focus on innate immunity

Immunity is not simply the product of a series of discrete linear signalling pathways; rather it is comprised of a complex set of integrated responses arising from a dynamic network of thousands of molecules subject to multiple influences. Its behaviour often cannot be explained or predicted solely by examining its components. Here, we review recently developed resources for the systems-level investigation of immunity. Although innate immunity is emphasized here, its considerable overlap with adaptive immunity makes many of these resources relevant to both arms of the immune response. We discuss recent studies implementing these approaches and illustrate the potential of systems biology to generate novel insights into the complexities of innate immunity.