Bioinformatics Advance Access originally published online on September 17, 2008
Bioinformatics 2008 24(22):2636-2638; doi:10.1093/bioinformatics/btn492
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
FunNet: an integrative tool for exploring transcriptional interactions
1INSERM, UMR-S 872, Les Cordeliers, Eq. 7 Nutriomique, Paris F-75006, 2Pierre et Marie Curie-Paris 6 University, Cordeliers Research Center, UMR-S 872, Paris F-75006, 3IRD, UR Géodes, Centre IRD de l'Ile de France, Bondy F-93143 and 4Assistance Publique-Hôpitaux de Paris, Pitié Salpêtrière Hospital, Nutrition and Endocrinology department, Paris F-75013, France
*To whom correspondence should be addressed.
 |
ABSTRACT |
|---|
 TOP ABSTRACT 1 MOTIVATION2 THE ANALYTICAL MODEL3 IMPLEMENTATIONREFERENCES |
|---|
Summary: We describe here an exploratory tool, called FunNet, which implements an original systems biology approach, aiming to improve the biological relevance of the modular interaction patterns identified in transcriptional co-expression networks. A suitable analytical model, involving two abstraction layers, has been devised to relate expression profiles to the knowledge on transcripts’ biological roles, extracted from genomic databases, into a comprehensive exploratory framework. This approach has been implemented into a user-friendly web tool to promote its open use by the community.
Availability: http://www.funnet.info
Contact: edi.prifti{at}crc.jussieu.fr
|
1 MOTIVATION |
|---|
|
TOPABSTRACT1 MOTIVATION2 THE ANALYTICAL MODEL3 IMPLEMENTATIONREFERENCES |
|---|
Interactions analysis has emerged as an increasingly popular framework for exploring the complex system of relations that characterize the functional organization of cellular environments. In this context, network abstractions are used to represent the interdependence between various interacting entities (i.e. molecules, cellular and extracellular structures, processes and regulatory pathways), and support the formal assessment of their contextual relevance. Among many other applications, the exploratory analysis of the complex interactions that condition transcriptional expression patterns has brought significant insights into the robustness of the cells’ functional organization, its evolution and the adaptive mechanisms that are continuously shaping transcriptional regulation. Built by relating together transcripts that share similar expression profiles, the gene co-expression networks display many topological properties commonly observed in other complex interaction networks. Among them, the hierarchical organization of their modular architectures, spread over multiple scales, is thought to carry remarkable biological significance, illustrating the strong relationship which connects transcripts’ regulatory patterns to the functional organization of the cell (Allocco et al., 2004; Barabasi and Oltvai, 2004).
The assessment of gene co-expression relies on various measures, including correlation coefficients between transcriptional profiles, as well as entropic informational criteria. Despite their intuitive character, the biological relevance of these measures is limited by their sensitivity to experimental noise that affects high-throughput gene expression measurements. To increase the robustness to noise several topological overlap criteria have been recently suggested (Yip and Horvath, 2007; Zhang and Horvath, 2005).
Here, we propose a conceptually different strategy, aiming to improve the biological relevance of the modular interaction patterns identified in co-expression networks, in a manner that is complementary to the aforementioned techniques. To this purpose, an original systems biology approach has been devised to extract the available knowledge of transcripts’ roles from genomic databases, and use it to direct the analysis of the transcriptional interactions. A user-friendly web tool, named FunNet, has been built to promote the open use of this integrative approach by the community.
|
2 THE ANALYTICAL MODEL |
|---|
|
TOPABSTRACT1 MOTIVATION2 THE ANALYTICAL MODEL3 IMPLEMENTATIONREFERENCES |
|---|
An original two layer analytical model has been imagined to support the integration of the available information on transcripts’ biological roles into the co-expression network analysis (Fig. 1). A detailed description of the integrative approach implemented in FunNet is presented elsewhere (Henegar et al., 2008, Henegar et al., 2008; Henegar et al., manuscript in preparation). The analytical model is organized around an abstraction of the cellular processes, represented formally by the biological themes that annotate transcripts’ roles in genomic databases [Gene Ontology (GO), KEGG, etc.]. A non-linear dynamical system allows quantifying the contextual proximity of annotating themes based on the similarity of their annotated transcriptional profiles. Relevant themes are then related into a theme proximity network, overlaying the gene co-expression network, to capture and illustrate the interdependence in between the transcriptional regulation of associated cellular processes. A spectral clustering technique is applied to identify co-expression modules by grouping together themes sharing closely related transcriptional patterns. In the end, interactional centrality measures, such as intra-modular connectivity or network betweenness (Yu et al., 2007), are computed to assess the contextual significance of the analyzed elements (i.e. transcripts and biological themes). This approach has been applied to explore the transcriptomic signature of the adipose tissue, to identify the pathological mechanisms underlying the alteration of its metabolic functions in obese subjects (Henegar et al., 2008).
|
|
3 IMPLEMENTATION |
|---|
|
TOPABSTRACT1 MOTIVATION2 THE ANALYTICAL MODEL3 IMPLEMENTATIONREFERENCES |
|---|
FunNet computational routines were implemented in R and are available as a standalone package from worldwide CRAN mirrors (http://cran.r-project.org/mirrors). The graphical interface was written in PHP and is installed on a public web server, which has been referenced in the microarray tools’ section of the GO website (http://www.geneontology.org). The web interface is divided into four main sections. The first one allows the user to select a reference genome and the type of analytical situation to be considered. In the second, data files are uploaded, while the third provides various analytical options. The last section summarizes the uploaded data and selected options before concluding the submission.
3.1 Input data
FunNet was designed to address two types of situations: the analysis of one set of transcriptional profiles, or the simultaneous analysis of two sets of gene expression profiles distinguished by their overall regulation pattern in the analyzed condition (i.e. up- versus down-regulated transcripts, as those illustrated in Fig. 1). Expression data have to be submitted as specifically formatted text files, in which transcripts are identified through EntrezGene IDs. More details on the format of the data files and the available options are provided online.
3.2 Analytical options
The first set of options relates to the functional profiling of gene expression data, performed to identify a list of significantly overrepresented GO and KEGG categories, among those annotating the biological roles of the analyzed transcripts. Several ways of computing the gene enrichment for ontological themes are provided in relation to their informational specificity (i.e. the degree of precision with which they illustrate the biological information available on transcripts’ roles), or their conceptual granularity (i.e. the degree of generality of their conceptual content, indicated by the terminological level of the GO lattice to which they belong). A decorrelated annotation procedure, inspired by a method proposed by Alexa et al. (2006), is also provided. A false discovery ratio (FDR) correction approach (Storey and Tibshirani, 2003) can be optionally selected to assess the statistical significance of the annotating themes’ overrepresentation.
A second set of options is provided in relation to the analysis of transcriptional interactions. As co-expression measures, FunNet uses either Spearman's or Pearson's correlation coefficients, or the Euclidean distance in between the vectors of gene expression measurements. A co-expression significance threshold can be estimated through a separate procedure, which computes the goodness of fit of the theoretical scale-free topological model for various threshold values applied to the analyzed gene expression data. Optionally, a topological overlap criterion can be combined with the aforementioned co-expression measures to improve their robustness to experimental noise (Zhang and Horvath, 2005).
3.3 Results
After the completion of the analyses, the results can be downloaded in a compressed archive format. Various files are generated, some of which contain details on the functional profile of the analyzed transcripts, while others are specifically formatted to allow for the visualizing and annotating of the identified transcriptional modules in the Cytoscape® software (Shannon et al., 2003). Besides the overall network architecture, various patterns can be thus illustrated (Fig. 1), including the overall regulation of the analyzed transcripts and annotating themes, the strength of their interactions within the network, and their interactional centrality. The last one is thought to illustrate the contextual relevance of the analyzed elements (Barabasi and Oltvai, 2004), and therefore may provide a valuable insight to the choice of target processes and genes to be further explored in subsequent biological experiments.
Funding: European Community 6th framework HEPADIP project; the ‘Association de Langue Française pour l'Etude du Diabàte et des Maladies Métaboliques’ (ALFEDIAM).
Conflict of Interest: none declared.
|
FOOTNOTES |
|---|
Associate Editor: John Quackenbush
Received on June 17, 2008; revised on July 23, 2008; accepted on September 12, 2008
|
REFERENCES |
|---|
|
TOPABSTRACT1 MOTIVATION2 THE ANALYTICAL MODEL3 IMPLEMENTATIONREFERENCES |
|---|
Alexa A, et al. Improved scoring of functional groups from gene expression data by decorrelating GO graph structure. Bioinformatics (2006) 22:1600–1607.
Allocco DJ, et al. Quantifying the relationship between co-expression, co-regulation and gene function. BMC Bioinformatics (2004) 5:18.[CrossRef][Medline]
Barabasi AL, Oltvai ZN. Network biology: understanding the cell's functional organization. Nat. Rev. Genet. (2004) 5:101–113.[CrossRef][Web of Science][Medline]
Henegar C, et al. Adipose tissue transcriptomic signature highlights the pathological relevance of extracellular matrix in human obesity. Genome Biol. (2008) 9:R14.[CrossRef][Medline]
Shannon P, et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. (2003) 13:2498–2504.
Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc. Natl Acad. Sci. USA (2003) 100:9440–9445.
Yip AM, Horvath S. Gene network interconnectedness and the generalized topological overlap measure. BMC Bioinformatics (2007) 8:22.[CrossRef][Medline]
Yu H, et al. The importance of bottlenecks in protein networks: correlation with gene essentiality and expression dynamics. PLoS Comput. Biol. (2007) 3:e59.[CrossRef][Medline]
Zhang B, Horvath S. A general framework for weighted gene co-expression network analysis. Stat. Appl. Genet. Mol. Biol. (2005) 4. Article17.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  C. Poitou, A. Divoux, A. Faty, J. Tordjman, D. Hugol, A. Aissat, M. Keophiphath, C. Henegar, S. Commans, and K. Clement Role of Serum Amyloid A in Adipocyte-Macrophage Cross Talk and Adipocyte Cholesterol Efflux J. Clin. Endocrinol. Metab., May 1, 2009; 94(5): 1810 - 1817. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||