Vol. 25 no.

8 2009, pages 1091–1093

BIOINFORMATICS APPLICATIONS NOTE doi:10.1093/bioinformatics/btp101

Systems biology

ClueGO: a Cytoscape plug-in to decipher functionally grouped

gene ontology and pathway annotation networks
Gabriela Bindea1−4,† , Bernhard Mlecnik1−3,† , Hubert Hackl4 , Pornpimol Charoentong4 ,
Marie Tosolini1−3 , Amos Kirilovsky1−3 , Wolf-Herman Fridman1−3,5 , Franck Pagès1−3,5 ,
Zlatko Trajanoski4 and Jérôme Galon1−3,5,∗
1 INSERM, AVENIR Team, Integrative Cancer Immunology, U872, 75006 Paris, 2 Université Paris Descartes,
3 Université Pierre et Marie Curie Paris 6, Cordeliers Research Center, Paris, France, 4 Institute for Genomics
and
Bioinformatics, Graz University of Technology, Graz, Austria and 5 Assistance Publique-Hôpitaux de Paris, HEGP,
Paris, France
Received on November 13, 2008; revised on February 8, 2009; accepted on February 16, 2009

Downloaded from https://academic.oup.com/bioinformatics/article/25/8/1091/324247 by guest on 17 November 2024

Advance Access publication February 23, 2009
Associate Editor: Trey Ideker

ABSTRACT and reconstruct the hierarchical ontology tree, whereas ClueGO

Summary: We have developed ClueGO, an easy to use Cytoscape uses kappa statistics to link the terms in the network. Compared
plug-in that strongly improves biological interpretation of large lists with the approach of Ramos et al. (2008) which creates an
of genes. ClueGO integrates Gene Ontology (GO) terms as well
in silico annotation network based on pathways and protein
interaction data and maps the gene list of interest afterwards,
as KEGG/BioCarta pathways and creates a functionally organized
ClueGO generates a dynamical network structure by already initially
GO/pathway term network. It can analyze one or compare two lists considering the gene lists of interest. ClueGO integrates GO terms
of genes and comprehensively visualizes functionally grouped terms. as well as KEGG/BioCarta pathways and creates a functionally
A one-click update option allows ClueGO to automatically download organized GO/pathway term network. A variety of flexible restriction
the most recent GO/KEGG release at any time. ClueGO provides an criteria allow for visualizations in different levels of specificity.
intuitive representation of the analysis results and can be optionally In addition, ClueGO can compare clusters of genes and visualizes
used in conjunction with the GOlorize plug-in. their functional differences. ClueGO takes advantage of Cytoscape’s
Availability: http://www.ici.upmc.fr/cluego/cluegoDownload.shtml versatile visualization framework and can be used in conjunction
Contact: [email protected] with the GOlorize plug-in (Garcia et al., 2007).
Supplementary information: Supplementary data are available at
Bioinformatics online. 2 METHODS AND IMPLEMENTATION
ClueGO has two major features: it can be either used for the visualization
1 INTRODUCTION of terms corresponding to a list of genes, or the comparison of functional
Since the number of genes that can be analyzed by high-throughput annotations of two clusters.
experiments by far exceeded what can be interpreted by a single
person, different attempts have been initiated in order to capture 2.1 Data import
biological information and systematically organize the wealth Gene identifier sets can be directly uploaded in simple text format or
of data. For example Gene Ontology (GO) (Ashburner et al., interactively derived from gene network graphs visualized in Cytoscape.
2000) annotates genes to biological/cellular/molecular terms in ClueGO supports several gene identifiers and organisms by default and is
a hierarchically structured way, whereas Kyoto encyclopedia of easy extendable for additional ones in a plug-in like manner (Supplementary
Material).
genes and genomes (KEGG) (Kanehisa et al., 2002) and BioCarta
assigns genes to functional pathways. Several functional enrichment
analysis tools (e.g. Boyle et al., 2004; Huang et al., 2007; Maere 2.2 Annotation sources
et al., 2005; Ramos et al., 2008; Zeeberg et al., 2003) and algorithms To allow a fast analysis, ClueGO uses precompiled annotation files including
(e.g. Li et al., 2008) were developed to enhance data interpretation. GO, KEGG and BioCarta for a wide range of organisms. A one-click update
As most of these tools mainly present their results as long lists feature automatically downloads the latest ontology and annotation sources
or complex hierarchical trees, we aimed to develop ClueGO a and creates new precompiled files that are added to the existing ones. This
Cytoscape (Shannon et al., 2003) plug-in to facilitate the biological ensures an up-to-date functional analysis. Additionally ClueGO can easily
interpretation and to visualize functionally grouped terms in the form integrate new annotation sources in a plug-in like way (Supplementary
of networks and charts. Other tools like BiNGO (Maere et al., 2005) Material).
or PIPE (Ramos et al., 2008) assess overrepresented GO terms
2.3 Enrichment tests
∗ Towhom correspondence should be addressed. ClueGO offers the possibility to calculate enrichment/depletion tests for
† The authors wish it to be known that, in their opinion, the first two authors terms and groups as left-sided (Enrichment), right-sided (Depletion) or two-
should be regarded as Joint First Authors. sided (Enrichment/Depletion) tests based on the hypergeometric distribution.

© 2009 The Author(s)

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/
by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

[18:02 30/3/2009 Bioinformatics-btp101.tex] Page: 1091 1091–1093

 G.Bindea et al.

Furthermore it provides options to calculate mid-P-values and doubling

for two-sided tests to deal with discreetness and conservatism effects as
suggested by (Rivals et al., 2007). To correct the P-values for multiple testing
several standard correction methods are proposed (Bonferroni, Bonferroni
step-down and Benjamini-Hochberg).

2.4 Network generation and visualization

To create the annotations network ClueGO provides predefined functional
analysis settings ranging from general to very specific ones. Furthermore,
the user can adjust the analysis parameters to focus on terms, e.g. in certain
GO level intervals, with particular evidence codes or with a certain number
and percentage of associated genes. An optional redundancy reduction
feature (Fusion) assesses GO terms in a parent–child relation sharing similar
associated genes and preserves the more representative parent or child
term. The relationship between the selected terms is defined based on
their shared genes in a similar way as described by Huang et al. (2007).

Downloaded from https://academic.oup.com/bioinformatics/article/25/8/1091/324247 by guest on 17 November 2024

ClueGO creates first a binary gene-term matrix with the selected terms
and their associated genes. Based on this matrix, a term–term similarity
matrix is calculated using chance corrected kappa statistics to determine the
association strength between the terms. Since the term–term matrix is of
categorical origin, kappa statistic was found to be the most suitable method.
Finally, the created network represents the terms as nodes which are linked
based on a predefined kappa score level. The kappa score level threshold
can initially be adjusted on a positive scale from 0 to 1 to restrict the
network connectivity in a customized way. The size of the nodes reflects the
enrichment significance of the terms. The network is automatically laid out
using the Organic layout algorithm supported by Cytoscape. The functional
groups are created by iterative merging of initially defined groups based on
the predefined kappa score threshold. The final groups are fixed or randomly
colored and overlaid with the network. Functional groups represented by
their most significant (leading) term are visualized in the network providing
an insightful view of their interrelations. Also other ways of selecting the
group leading term, e.g. based on the number or percentage of genes per
term are provided. As an alternative to the kappa score grouping the GO
hierarchy using parent–child relationships can be used to create functional
groups.
When comparing two gene clusters, another original feature of ClueGO
allows to switch the visualization of the groups on the network to the cluster
distribution over the terms. Besides the network, ClueGO provides overview
charts showing the groups and their leading term as well as detailed term
histograms for both, cluster specific and common terms.
Like BiNGO, ClueGO can be used in conjuntion with GOlorize for
functional analysis of a Cytoscape gene network. The created networks,
charts and analysis results can be saved as project in a specified folder and
used for further analysis.

Fig. 1. ClueGO example analysis of up- and down-regulated NK cell genes

3 CASE STUDY in peripheral blood from healthy human donors. (a) GO/pathway terms
specific for upregulated genes. The bars represent the number of genes
To demonstrate how ClueGO assesses and compares biological
associated with the terms. The percentage of genes per term is shown as bar
functions for clusters of genes we selected up- and down-regulated label. (b) Overview chart with functional groups including specific terms for
natural killer (NK) cell genes in healthy donors from an expression upregulated genes. (c) Functionally grouped network with terms as nodes
profile of human peripheral blood lymphocytes (GSE6887, Gene linked based on their kappa score level (≥0.3), where only the label of
Expression Omnibus). For upregulated NK genes ClueGO revealed the most significant term per group is shown. The node size represents the
specific terms like ‘Natural killer cell mediated cytotoxicity’ in term enrichment significance. Functionally related groups partially overlap.
the group ‘Cellular defense response’. Downregulated in NK cells Not grouped terms are shown in white. (d) The distribution of two clusters
compared with the reference (a pool of all immune cell types) were visualized on network (c). Terms with up/downregulated genes are shown
genes involved in the innate immune response (Macrophages), but in red/green, respectively. The color gradient shows the gene proportion of
each cluster associated with the term. Equal proportions of the two clusters
also in the adaptive immune response (T and B cell). The common
are represented in white.
functionality refers to characteristics of leukocytes (chemotaxis),
besides other terms involved in cell division and metabolism (Fig. 1).

1092

[18:02 30/3/2009 Bioinformatics-btp101.tex] Page: 1092 1091–1093

 ClueGO

4 SUMMARY Boyle,E.I. et al. (2004) GO::TermFinder–open source software for accessing Gene
Ontology information and finding significantly enriched gene ontology terms
ClueGO is a user friendly Cytoscape plug-in to analyze interrelations associated with a list of genes. Bioinformatics, 20, 3710–3715.
of terms and functional groups in biological networks. A variety Garcia,O. et al. (2007) GOlorize: a cytoscape plug-in for network visualization
of flexible adjustments allow for a profound exploration of gene with Gene Ontology-based layout and coloring. Bioinformatics, 23,
clusters in annotation networks. Our tool is easily extendable to 394–396.
new organisms and identifier types as well as new annotation Huang,D.W. et al. (2007) The DAVID Gene Functional Classification Tool: a novel
biological module-centric algorithm to functionally analyze large gene lists.
sources which can be included in a transparent, plug-in like manner. Genome Biol., 8, R183–R183.
Furthermore, the one-click update feature of ClueGO ensures an Kanehisa,M. et al. (2002) The KEGG databases at GenomeNet. Nucleic Acids Res., 30,
up-to-date analysis at any time. 42–46.
Li,Y. et al. (2008) A global pathway crosstalk network. Bioinformatics, 24,
1442–1447.
ACKNOWLEDGEMENTS
Maere,S. et al. (2005) BiNGO: a Cytoscape plugin to assess overrepresentation
We thank A Van Cortenbosch for the name of the tool. of Gene Ontology categories in biological networks. Bioinformatics, 21,
3448–3449.
Funding: INSERM; Ville de Paris; INCa; the Austrian Ministry for Ramos,H. et al. (2008) The protein information and property explorer: an easy-to-use,
Science and Research, Project GEN-AU; BINII; the European 7FP rich-client web application for the management and functional analysis of proteomic
Grant Agreement 202230 (GENINCA). data. Bioinformatics, 24, 2110–2111.

Downloaded from https://academic.oup.com/bioinformatics/article/25/8/1091/324247 by guest on 17 November 2024

Rivals,I. et al. (2007) Enrichment or depletion of a GO category within a class of genes:
Conflict of Interest: none declared. which test? Bioinformatics, 23, 401–407.
Shannon,P. et al. (2003) Cytoscape: a software environment for integrated models of
biomolecular interaction networks. Genome Res., 13, 2498–2504.
REFERENCES
Zeeberg,B.R. et al. (2003) GoMiner: a resource for biological interpretation of genomic
Ashburner,M. et al. (2000) Gene ontology: tool for the unification of biology. The Gene and proteomic data. Genome Biol., 4, R28–R28.
Ontology Consortium. Nat. Genet., 25, 25–29.

1093

[18:02 30/3/2009 Bioinformatics-btp101.tex] Page: 1093 1091–1093