The CoryneRegNetLoader is able to extract transcriptional regulatory networks stored in CoryneRegNet and to import them into Cytoscape. A directed graph is then created, with genes as nodes and regulatory interactions as directed edges. For each node, CoryneRegNetLoader imports the up-to-date gene and protein annotation data, which is stored as node attributes. For each edge, the plugin retrieves the evidence of the underlying protein-DNA interaction and whether the target genes are up- or downregulated. This data is stored as edge attributes. In 16 we described the SOAP-based Web Service server of CoryneRegNet and presented an application example that shows how the community profits from an integrated visualization of corynebacterial network data. With this plugin, we take a step forward and provide an interface to a more powerful environment for integrated biological network analyses.

The COMA plugin implements a method to facilitate consistency checks for gene expression studies given a gene regulatory network. It checks whether the expression levels of all genes are as expected from the network topology.

We discretize the expression levels of genes to overexpressed ('+') and underexpressed ('-') respectively. Furthermore, we do the same for the activation or repression of a target gene by a regulator. Let g ∈ {+, -} be the gene expression of a gene G. Let t ∈ {+, -} be the gene expression of a transcription factor T, which regulates G. Finally, let r ∈ {+, -} be the known gene regulatory interaction of gene G by transcription factor T. Subsequently consider the algebraic signs in the following equation:

t·g = r

If the equation is incorrect, for instance '+' · '-' = '+', this is a putative inconsistency with the given regulatory network; also refer to 5. Based on this, for every gene G of a given microarray experiment with expression g, all transcription factors T that regulate G are considered, and for all transcription factors the expression level t and the type of the regulatory interaction r is determined. Subsequently, COMA applies the above explained inconsistency test. Table 1 summarizes the results for all possible combinations of t, g, and r.

Although COMA was designed to work with the CoryneRegNetLoader, it can be used with any directed biological network as long as there is a Cytoscape edge attribute of type boolean, integer, or float. This attribute is used to distinguish between up- and downregulations and can be defined by the user. Gene expression values can be imported by using the Cytoscape microarray import feature, for instance. Generally, any node attribute of type integer or float can be used for COMA. Each putative inconsistent edge is listed in a result table. If the user selects a certain node-pair, both nodes including the edge between them are highlighted in the Cytoscape graph. Figure 1 show a subset of the complete gene regulatory network of C. glutamicum, extracted from CoryneRegNet by utilizing the CoryneRegNetLoader plugin. The COMA plugin has been applied to the graph by using the results of a gene expression study provided by Brune et. al 17, which is also available in Cytoscape format as Additional file 1. With this plugin, Cytoscape now provides hints for missing regulatory interactions.

Both plugins together facilitate in silico the study and visualization of procaryotic transcriptional gene regulation, particularly in Corynebacteria and E. coli, by combining the manually curated knowledge from the CoryneRegNet database with the power of the graph analysis and visualization capabilities of Cytoscape.

Since various chemical mechanisms take part in the regulation of the genes, in a future version of the COMA plugin we plan to consider also non-transcriptional gene regulatory interactions. Additionally, we plan to extend COMA to check for coherent and incoherent feed-forward loops 18 in a given network. We will also conside further rules to validate microarray data in the context of known regulatory networks (see e.g. 19).