|Home | About | Journals | Submit | Contact Us | Français|
Increasing complexity of cell signaling network maps requires sophisticated visualization technologies. Simple web-based visualization tools can allow for improved data presentation and collaboration. Researchers studying cell signaling would benefit from having the ability to embed dynamic cell signaling maps in web pages.
Gradual increased intricacy of cell signaling pathways requires sophisticated representations of cell signaling maps. As such, several web-based tools such PATIKAWeb (Dogrusoz et al., 2006), BioPP (Viswanathan et al., 2007) and VisAnt (Hu et al., 2005) have been developed to aid in cell signaling networks visualization and analysis. These visualization tools commonly use JAVA Web Start or JAVA Applet technologies that allow networks to be dynamically and interactively visualized in web-browsers. However, these JAVA-based tools do not allow signaling maps to be embedded inline within other webpage content and easily communicate with the browser, as would be desired when presenting data for collaboration, or for dynamically displaying pathway diagrams from a database such as the connection maps database CMADES at Science Magazine’s Signal Transduction Knowledge Environment (STKE) (Gough, 2002). There is some functionality for linking visualized network maps with other content on web pages and communication with the browser in VisAnt, but this requires the user to have JAVA Web Start properly configured (Hu et al., 2005). Hence, there is a niche for a powerful yet simplified light-weight viewer. Requirements include compromise between details and overall viewing, minimal crossing of links, internet-readiness and compatibility with several protein–protein interactions data formats (Rojdestvenski, 2003). Most interaction maps used for visualization of intracellular signaling networks on the web are presented as either HTML embedded static images (Kitano et al., 2005) or using Scalable Vector Graphics (SVG http://www.w3.org/Graphics/SVG/) (Gough, 2002; Li et al., 2002). Both approaches can be made interactive; e.g. scaling smoothly across multiple zoom levels, panning by dragging the images up and down, as well as providing hyperlinks embedded in the map images as mouse-over and mouse-click events. Currently, most popular web browsers, such as Internet Explorer or Safari lack native support for SVG while other browsers like Firefox only recently added some basic SVG support. As such, viewing SVG interactive images requires third party plug-in software, such as Adobe’s SVG viewer (http://www.adobe.com/svg/). Raster-based images, such a Joint Photographic Experts Group (JPEG), Graphics Interchange Format (GIF) or Portable Network Graphics (PNG) formats, have more universal native support across platforms, but are static; e.g. lack the ability to zoom in and out on the image. A high-resolution image may allow clear viewing of individual nodes in a rendered signaling network at a high level of zooming, but when attempting to zoom out, the raster image is often distorted, where edges/links may disappear. Alternatively, a low-resolution image can give a clear picture of the overall network, but does not allow clear visualization of individual nodes when zooming in. Additionally, panning raster-based images is often implemented with scroll bars and frames but this design is limited and rough (Viswanathan et al., 2007). Both visualization technologies: SVG and static raster images, allow for hyper-linking nodes to other web pages by clicking on the displayed nodes. This is accomplished by either specifying hyperlinks in the SVG source code, or by using an ‘imagemap’ in the HTML code that displays the raster image.
The layout of the nodes in maps is accomplished by the fileloader module. This provides developers with an easy way to add to AVIS additional layout algorithms. The included fileloaders that we implemented can be configured to use all of currently available Graphviz layout algorithms: dot, neato, fdp, circo and twopi. Alternatively, file formats, such as the Pajek.net file format (Batageli and Mryar, 2003) and a native.avis file format we developed (described in the manual) can include the layout in the input text file. As such, other layout programs can be used before importing into AVIS. Alternatively, a simple Perl wrapper can be added to automatically import other layout algorithms.
The graph rendering is accomplished by one of the two included graph drawing modules or by a user-defined graph drawing module. The default rendering functions generate a layout.dot file, calling Graphviz to do the actual rendering. In AVIS we implemented an alternative option: AVISgraphDraw. AVISgraphDraw uses PerlMagick (ImageMagic) image drawing API to draw the network map images. Both of these options are fairly simple graph rendering tools, but can draw a wide variety of networks as long as they can be described in terms of properties such as positions, shapes, size, colors of nodes, and placement and paths of edges. If more sophisticated graph rendering tools with other features are required, it is simple to add a Perl wrapper for such tools. The process of developing such wrappers is described in the user manual.
AVIS supports several standard graph representation formats such as Pajek.net (Batageli and Mryar, 2003) and Systems Biology Markup Language (SBML) (Deckard et al., 2006; Kitano et al., 2005) as well as a simple structured text file format we developed called .sig (Ma’ayan et al., 2005). It also supports a more flexible native.avis format which allows editing of any graph features, giving users more control over the placement of nodes, colors of nodes and arrows, shapes of nodes and types of arrows as well as more control on setting up hyperlinks. AVIS was implemented for visualization of PathwayGenerator, a tool that displays over 4000 automatically generated mammalian cell signaling maps created from a large-scale literature-based neuronal cell signaling network we developed (Ma’ayan et al., 2005); AVIS was also implemented for visualization of NodeNeighborhood: a tool to visualize first and second interacting neighbors of mammalian proteins based on data from the IntAct database (Kerrien et al., 2007) as well as two yeast-2-hybrid studies of mammalian protein interactions; and AVIS was implemented in Genes2Networks, a tool that can be used to create subnetworks from lists of genes and proteins by using 10 background mammalian protein interaction networks to ‘connect’ seed lists of gene symbols.
Other visualization programs are compatible with standards such as BioPAX, and PSI-MI (Stromback and Lambrix, 2005). AVIS can be adapted and upgraded to support any of these formats with simple Perl wrapped plug-ins. We hope that interested developers would find this tool easy to leverage for visualization of their system of interest.
This research is supported by NIH Grant No. GM-072853 and an advanced center grant from NYSTAR to R.I.
Conflict of Interest: none declared.