Background: Cosmin makes an interesting subject because he has used Cytoscape extensively, and in a highly-customized fashion. His lab uses Cytoscape with a plugin that was developed for them, specifically, by a former master's student in Benno's group. They work with a large set of yeast interactions, obtained from numerous published datasets, and compiled into one gigantic SIF file. The custom plugin, called Extend, performs a function similar to a database lookup to limit the network view the subset of the network that the user is most interested in.
Basically, the whole massive network is loaded into Cytoscape as the current network, but the current view consists of only those nodes that the user has requested. The execution starts with the user specifying one more more node to put into the view (and onto the canvas). If there are edges between nodes that are on the canvas, they are displayed. If a node has edges to other nodes not on the canvas, it is marked with a thick green node border. The user can right-click on a node to pull in the node's other interactions, effectively adding its first neighbors to the view.
The complete network uses numerous edge types. Each edge type describes the type of interaction (e.g. yeast two-hybrid), and interaction dataset. Edge attributes associate the edges with the PubMed IDs of the interaction datasets. So, for any edge on the Cytoscape canvas, the user can pull up the publication that supports the interaction. Line types are used to differentiate interaction types and interaction datasets.
This involves a great deal of customization. One of Cosmin's current laments is that moving Cytoscape from one computer to another involves moving a lot of components: the Cytoscape jars, the plugin jar, the data files, and the user configuration files (esp. vizmap.props).
The attached PDF file is a detailed transcript of the interview, including numerous screen shots.