Today Sean O’Donoghue talked at the 14th International Conference on Systems Biology (ICSB) in Copenhagen. O’Donoghue is affiliated with the Australia’s Commonwealth Scientific and Industrial Research Organization (CSIRO) and the Garvan Institute of Medical Research both located in Sydney. His talk was officially titled “Visual Analytics: A New Approach for Systems Biology” but he immediately after the start admitted it could better be named “Untangling the Hairball“. Using 6 guidelines he quickly showed the basic principles of data visualization for scientists. Since his talk contained quite some references to journal articles, webservers and online tools I thought it would be useful to put everything together in a post.
We have problem…
…we are very good at making hairballs
Luckily there are some tools available to help us:
VizBi – An young annual conference series that ‘brings together scientists, illustrators, and designers actively using or developing computational visualization to study a diverse range of biological data’ the next event is scheduled for March 2014 in Heidelberg, Germany. (They also host an interesting blog).
VizBi Videos – The recent talks of the VizBi conference are recorded and can be watched back . Currently there are 80 videos available including keynotes from Manuel Lima (Visual Complexity), John Westbrook (Protein DataBase) and Ben Fry (Processing). O’Donoghue really recommended the talk of Tamara Munzner on Visualization Principles.
Points of View in Nature Methods – In the course of 3 years Bang Wong, Martin Krzywinski and several co-authors wrote 35 columns in Nature Methods, they deal with common issues a scientist faces when transforming data into figures. The whole series can be downloaded for free.
The six guidelines that helps you to communicate your message more clearly
1. Use color wisely
Don’t use the rainbow color scale. With a reference to a blog post which sums up multiple reasons (Colorblind people cannot use them, The spectral order of hues has no inherent meaning. etc.)
“It’s not necessary wrong to put a lot of data on a graph, but there are better ways”
2. Use edge bundling
Use the hierarchical edge builder from d3.js as an example (it is interactive so you can play around with the tension).
3. Use space wisely
How we organize data along the X-Y directions is the most important decision we make. Using spring-embedding essentially leaves the overall organization of the graph up to chance. It is much better to impose structure, by using constraint-based layout methods. Especially recommended are methods that use X and Y to communication subcellular location and causality. A nice example of this is provided by the Cerebral plugin to Cytoscape.
4. Use visual analytics
Aquaria – a tool to visualize protein structures and analyze them. This allows you to interact with the data. More also on the wiki page.
Also use the Kinect and Leap Motion to manipulate the view see the movie below
5. Dealing with data
Big data = big screens
6.Add more reality
Famous quote from Tufte “To clarify ,add detail”
Currently working on an animation titled “The hungry microbiome”. The sneak preview looked really amazing, the only thing I could find on the web is a high-res snapshot by Christopher Hammang, Sean O’Donoghue, Christian Stolte, Drew Berry, Trevor Lockett, Julie Clarke, Leah Cosgrove, David Topping .
In conclusion a very elegant and practically applicable talk!