Page Edited and Updated by: Christopher Fucile
Protein-Protein Interaction Network VisualizationEdit
Protein interaction network visualization is one of the most user-friendly features developed for the simulation process of protein protein interactions. Drawing of a protein may seem to be easy but generating the same protein with all the types of conformation that it can attain for a simulation process is quite difficult. In this context one of the greatest manually produced molecular structures of its time was done by Kurt Kohn's 1999 map of cell cycle control. In this paper around 1,548 proteins were linked together as determined by yeast two-hybrid assay. Protein-protein interaction network visualization deals with territory that is very similar to that of protein-protein interaction prediction, but differs in several key ways. Foremost, visualization tools were developed simply for illustrating the big picture represented by protein-protein interaction data or expression data, for qualitative assessment, not necessarily for quantitative analysis or prediction. Expression and interaction experiments tend to be on such a large scale that it is difficult to analyze them, or indeed grasp the meaning of the results of any analysis. Visual representation of such large and scattered quantities of data allows trends that are difficult to pinpoint numerically to stand out and provide insight into specific avenues of protein functions and interactions that may be worth exploring first out of the bunch, either through confirmation or rejection and then later of significance or insignificance to the research problem at hand. With a few recent exceptions, visualization tools were not designed with the intent of being used for analysis so much as to show the workings of a molecular system more clearly.
Visualization tools also do not actually predict protein-protein interactions themselves or their characteristics. On the contrary, visualization tools only create a graphical representation of what is already "known" in literature and in molecular interaction repositories such as the Gene Ontology(GO). A side-effect of displaying interaction networks by treating some proteins as members of more general family groupings or going by interactions in different tissues or as analogues in other tissues or organisms is the apparent display of a protein-protein or other molecular interaction that is inferred but may or may not have actually been observed and documented, something that could be misconstrued as a predicted interaction. There is no statistical analysis involved; this phenomenon is merely a fluke of the annotation scheme used as a reference (albeit a useful fluke).
Perhaps the most significant difference between protein-protein interaction network visualization and protein-protein interaction prediction is the nature of the information it provides. Protein-protein interaction prediction is characterized by its concern with how proteins will interact, where they will interact, under what conditions they were interact, and what parts are necessary for their interaction. These characteristics are governed by physical and chemical properties of the proteins or other molecules involved, which may be actual molecules that have been described through extensive proteomics experiments or hypothetical, in silico generated species that are being investigated for pharmaceutical and other applications. Interaction network visualization tools are given no knowledge of physical or chemical properties of the proteins they are displaying other than their presence in the system and sometimes a normalized level of expression. As a result, the information they inadvertently impart concerns only whether or not certain proteins putatively interact with certain other proteins, not how they interact, when they interact, or why they interact.
Due to the ever increasing employment of interaction of network visualization to direct research, scientists decided to automate the process. Some of the automated visualization systems are as follows.
Osprey is a visualization tool that represents genes as nodes and interactions between gene products as edges. It uses the GRID database to determine interactions by default, but can be configured in a number of ways to meet the user's needs. Osprey was developed in answer to the need to text-search the network of interactions and can output networks to a number of graphical formats.
A limited version is available at the BioGRID website: http://www.thebiogrid.org/
A full version for personal or non-profit use is downloadable from the Osprey site: http://biodata.mshri.on.ca/osprey/servlet/Index
VisANT is a tool that creates a visual interface for mining gene/protein interaction data with a focus on being adaptable to new types of data, both in terms of interactions and interacting members. It taps into a number of databases including GenBank, KEGG, and SwissProt.
VisANT is available via web interface at the VisANT site: http://visant.bu.edu/
NAViGaTOR is an interaction visualization tool that combines Java with OpenGL to deliver multi-platform 2D and 3D representations of physical protein-protein interaction networks. It includes options for color-coding network maps according to the GO and information in the literature. NAViGaTOR supports standard import and export formats such as GO and Proteomics Standards Initiative (PSI).
NAViGaTOR is available for download online at the NAViGaTOR site: http://ophid.utoronto.ca/navigator
Cytoscape is primarily a visualization tool written in Java for molecular interaction networks. User-defined interactions are supported, as is the ability to record attributes of nodes, edges, and networks. Annotations can be imported from GO and KEGG. Cytoscape allows visualization of expression data on a network not just by user-defined color-coding, but also through border thickness and color and the option to display expression levels and p-values on the network. Methods for analysis can be added through plugins that allow filtering by expression level, identifying differentially expressed subnetworks, and identifying clusters of highly interconnected nodes (molecules).
Cytoscape is available for download at the Cytoscape site: http://www.cytoscape.org
yEd is a graph editor that allows users to edit existing graphs as well as generate new ones but is not limited to working with biological data and offers no analysis tools. It seems as though yEd might be of use chiefly as a tool for cleaning up and tweaking graphs generated by other, more biologically-specialized visualization tools to be used for posters or other publications. yEd does support widely-used import formats like GML and is available for a number of platforms.
yEd is available for download at the yEd site: http://www.yworks.com/en/products_yed_about.htm
NetPro is a commercially developed and manually-curated protein-protein interactions database. The emphasis in NetPro is on facilitating the literature search and text-mining steps of investigating putative protein-protein interactions and their ramifications. Molecular Connections, the company that maintains and distributes NetPro, is also the producer of a handful of other visualization tools.
NetPro is not free, but more information is provided at the Molecular Connections site: http://www.gene-regulation.com/pub/databases/transpath/6.0/doc/netpro.html
1.^ Kohn, K. W. Molecular Interaction Map of the Mammalian Cell Cycle Control and DNA Repair Systems. Mol Biol Cell. 1999 August; 10(8): 2703–2734. <http://www.pubmedcentral.nih.gov/articlerender.fcgi?tool=pubmed&pubmedid=10436023>
8.^ "Biomedical text mining" Wikipedia, The Free Encyclopedia. Wikimedia Foundation, Inc. 30 March 2008. <http://en.wikipedia.org/wiki/Biomedical_text_mining>
9.^ "NetPro Documentation" NetPro. 30 March 2008. <http://www.gene-regulation.com/pub/databases/transpath/6.0/doc/netpro.html>