Notes from Interview with researcher at Unilever

22nd February, 2006

Interviewers: Chris Workman, Mike Smoot

What kind of work do you do and how do you use Cytoscape?

Identify specific components of biological systems for risk assessment assays. Integrate gene expression with protein-protein interaction networks in humans. Currently using mRNA expression data but plan to be using more proteomics data and reconstructions of signaling pathways.

A question biologists have been and will continue to ask: How do you know state data (gene expression) and the network interactions are correct?

Will help to link different levels of biological data. Node attributes are *critical* to the analysis they're doing. Applying HUGO ids (protein to entrez gene id) is an example of this.

Node attrs are important for communicating to biologists (visualization, etc.) while edge attrs are important for analysis.

Given confidence statistics on interactions, would be nice to use this information in jActiveModules.

What parts of the system do you use most?

Uses a continuum of features: editing networks, setting visual attributes, editing attribute data, working with plugins, etc vizMapper is important for seeing state. Filtering is invaluable. mCODE plugin. jActiveModules.

What aspects of Cytoscape do you find annoying?

Power of filter plugin is good, but the usability is bad. It's not alway clear which filter to use to view certain things. Cant read full name of filters.

In many cases either attributes or other filters need to exist before specific filters can be created and it isn't clear that something is necessary, let alone what is necessary. Examples: Topology filter relies on node/edge filters. Boolean requires other filters.

Would like to see filters better linked to vizMapper. Would be nice to have button to create a new network after a filter is applied. This could accompany a feature that helps users create a new binary attribute after a filter is applied (and make sure filter attribute has a good name).

Concepts could be clearer: For example, be more explicit that filters are a selection tool.

Saving the visual style is hard. Never quite works the way you want or expect. Spends a lot of time having to remap expression data after a network that has been closed has been re-opened. Getting the visual style right again is also difficult.

Saving the visual style, then reloading without the same attributes creates an ambiguous situation - it isn't clear how Cytoscape should react in this situation. (This should be alleviated by new save state features)

Are you expert on the vizMapper?

Intermediate. Takes time to remember how to use it. Panel could be better organized.

Is application of calculators clear?

Yes, fine, but the names are not good. Once you understand calculators and data, they're pretty straightforward.

Colored nodes according to biological process is very nice, however, how colors are assigned is not always clear. There are several boxes in the dialog that aren't easy to understand which makes it easy to make a mistake. It also isn't clear whether the thresholds behind the scenes are the same as traditional gene expression analysis. You need to be very careful!

Recommends a cookbook type of manual. If you're trying to do X, then follow these steps, etc.

How do you run (start) Cytoscape?

Using Windows.

Installation experience?

Installation is straightforward.

User interface and usability. Are the GUI features intuitive?

A general comment that was made with respect to many features was that things are initially difficult to understand, but once you do, they're ok.

Zooms/pans using the mouse buttons and not the menu buttons.

Menu change would be slightly disruptive - just make sure you keep the icons.

Has never used the disable vizmapper menu option and rarely uses the visualization menu. Tends to use the icon bar instead.

Do you like cytoPanels (i.e. frames within frames, or separate windows)?

Very, very good. Really useful. Would like to be able to rename networks.

Is the relationship of the graph model to attribute data to visualization clear?

SIF is key. Easy to explain and what they use for their data. Looks aren't as important as the ability to analyze.

Performance & responsiveness: Operations on large networks too slow?

A small diff between 2.1 and 2.2 - large graph handling is inadequate

Features: which are useful? useless?

Preferences are hard to use - once you know it's OK. The amount of memory should be a preference - to change this you currently have to edit the .bat or .sh file. Uses 1024M instead of 512M.

Plugins: Are they easy to use? Do plugins work for you? Are they easy to use and integrate with Cytoscape?

Getting plugins into the right place so that they run is a bit tricky.

What one feature would make Cytoscape a better tool for your work?

Would like the ability to load data directly from bind/dip/whatever from within Cytoscape over the network. It would be nice to avoid having to download the file, format it, etc.

InterviewNumber3 (last edited 2009-02-12 01:03:26 by localhost)

Funding for Cytoscape is provided by a federal grant from the U.S. National Institute of General Medical Sciences (NIGMS) of the Na tional Institutes of Health (NIH) under award number GM070743-01. Corporate funding is provided through a contract from Unilever PLC.

MoinMoin Appliance - Powered by TurnKey Linux