Interview with Thomas Kufer, Postdoc
- Melissa Cline, Cytoscape postdoc Olivier Garcia, Cytoscape graduate student Catherine Letondal, usability specialist
Background:
This interview describes how a first-time user goes about exploring Cytoscape.
TK is working on characterizing signaling pathways related to two specific molecules, in an area where existing pathway data is sparse. The lab is elucidating signaling pathways through TAP-tag, Y2H, and siRNA experiments. TK would like to use Cytoscape to view the lab's data, along with published interaction and expression datasets. He learned about Cytoscape by chance, through a conversation with Benno. TK seems quite computer-savvy, does not hesitate to experiment, and drops a line of experimentation quickly if the early results do not seem interesting.
TK had previously downloaded and tried Cytoscape, but came across the familiar 1-click install problem (i.e. no plugins were getting loaded). Additionally, he was experimenting with Cytoscape using a sample network that is very large (something pointed to on the tutorial web pages for reference), and when his computer did not have sufficient memory to generate a network view, he lost interest. We got him past the one-click install problem by instructing him to ignore the cytoscape.exe link and to click on cytoscape.bat instead, and started him off with a small sample network.
Starting out:
Immediately, he expressed interest in downloading relevant interaction datasets from external databases, and asked what IDs would work best.
Editing
[pointing to the left panel of Cytoscape] "I could not find an editor here". Expressed interest in a graphical interface for editing network data.
[MC indicated that an editor exists. Immediately, TK looked for it under the Edit menu. After he found it in the File menu, he pulled up the basic editor. Then, he had trouble specifying the node name he was interested in. Tried BioPax editor, and expressed pleasure on seeing molecule types indicated by node size. Expressed confusion on how/where to edit the node name. When the canonical name was edited, looked to see how to enter an interaction. Looked at the BioPax arrow menu, and expressed confusion about not seeing something called "interaction"] "You could say it's an activator" "You have to click the two guys?" [gesturing to the two nodes on the screen] [After MC indicated dragging over the arrow, created edges comfortably]
Setting visual styles
[TK expressed interest in using the canonical name as the display name. MC indicated that this is possible, pointed towards visual style menu, left him to explore. Immediately, he set about switching between visual styles in the pull-down menu, and when none of them worked, expressed a question. CL indicated that maybe he should explore the buttons. Then immediately he clicked on the Define button. He found his way quickly to the node label tab, and then the pace slowed down again. First, he tried the button to set the default node label. When that didn't work, he went to the mapping pull-down. The Pass-through mapper (which is correct) was the only option. When he tried it, and it brought up the pull-down menu, he finally saw the canonical name menu item, and then had little further trouble using the canonical name in the visual style.]
TK: Okay, I'm really happy
Fetching external data
TK: The next step is now that I would like to see if I could integrate data known of interaction partners of this protein
CL: how would you try this if you were alone?
[TK immediately tried right-click menu. Explored web links such as Amigo, realized they were not for import of interaction data, and moved on.]
[TK searched pull-down menus, found "Search cPATH" under Plugins menu, and tried it. cPATH didn't work, perhaps as a consequence of the network proxy he was running.]
[MC, remembering that the data file for cPATH needed updating recently, instructed on how to download and install a new copy of the cPATH plugin from the Cytoscape Plugins page. This required quitting Cytoscape, as Windows apparently locked the existing jar files while Cytoscape was running. In saving files, had trouble seeing how to save both network and attribute files together. MC indicated that this was addressed in the Save State functionality in 2.3. Eventually, Cytoscape was restarted and the old network and attribute data reloaded. Still had no success with cPATH]
[TK: asked what else could be done to download known data. MC suggested links from the tutorial on Fetching External Data]
TK: BIND would be interesting. But if I fetch data from BIND, then can I integrate it with my own data?
[MC indicated that there is a new plugin to do this]
[TK went to the BIND site, but gave up when the site asked for login and registration]
TK: What about other interaction databases?
[MC suggested DIP, which proved cumbersome to query by ID]
[MC then suggested Agilent Literature Search plugin, and guided TK through the download process. The installation process ran perfectly smoothly. TK restarted Cytoscape]
[TK immediately looked for literature search plugin under File menu]
[After TK was directed to the literature search plugin under the Plugins menu, a query window came up on the network proxy. CL was able to give him a proxy that worked, which allowed him to keep on moving forward]
[Literature Search plugin menu came up. TK looked at it briefly, and then entered a gene name in the Terms window. After another brief look, he clicked on the Forward button to execute the search.]
[When the literature search network appeared on the Cytoscape desktop, TK focused on it immediately, and the Literature Search menu immediately ended up buried].
[TK expressed satisfaction that the gene he had searched on appeared in the network. MC explained that this will not always happen, because the articles are selected by the search terms, but then the interactions are taken from sentences in the article. There will not always be an article in which each search term appears with the right sort of sentence]
TK [pointing to nodes in the network]: But here it does not discriminate between activation and interaction, so it does not put that into the network here.
[MC indicated that this information is available, let TK explore to look for it.]
[The first thing TK tried was the interaction lexicon pull-down menu]
[Meanwhile, TK found his way to the pull-down menu to specify number of articles, and immediately increased the number of articles from 10 to 100. This generated a dense network. TK located the node corresponding to his query term, and without prompting, moved the node around the screen to watch how the edges move, and to assess connectivity].
[At that point, the tape stopped]
[TK returned to the network, found the right-click menu, and on seeing the Gather Evidence option, tried that. MC explained to him that if he was already looking at a literature search network, this was not necessary - that Show Sentences was sufficient].
In conclusion, as we wrapped up, TK indicated that he had not read the manual, and had glanced just quickly at the tutorials. Yet as we were leaving, he suggested a Cytoscape training class for potential users at the Pasteur Institute. It is interesting that a user who works very directly, without taking time to read the manuals, still indicates that a training class would be a good use of time