Freelancing science

It’s already a while since IBM launched their data visualization framework Many Eyes. While initially resistant (there’s nothing that Gnuplot/R/Graphviz cannot do) I’ve decided to have a closer look. Obviously I didn’t get it before – Many Eyes is not about making the visualization easier (although IBM did quite a lot in that direction). It’s about sharing both data and approach to that data.

Many Eyes encourages to test things. A single perl one-liner and we see the most often occuring domains in proteins of Bacillus anthracis:

Or maybe we want to know which of these domains co-occur (nex to each other) in a single protein (only the biggest cluster shown):

(note that this is the output of quick hacks – I wouldn’t call it a scientific analysis)
Many Eyes is a service for general data. What about making such thing for the biological data analysis? The workflows may be shared on the myExperiment, and the data (input and output, and a visualization of the latter) on a site like Many Eyes? And deposition of the data would be required for certain papers? So far the results of the bioinformatic analysis are (sometimes) attached as a supplementary material in some weird format (pdf or doc). This at least make it accessible for years, but there’s no access to the original data and no way to verify if the analysis was correct other than looking at the results (and usually that’s not enough). Is there anything like that available? If not, do you think it would be valuable to build a service like that?

One short note: I’ve started this blog with a hope that maybe I would write something useful for the next Bio::Blogs edition, which would send me first visitors. To my surprise this site was found almost within hours from the first post by Pedro Beltrao. It looks like science bloggers never sleep :).

Last year I had a chance to make a short course on protein structure prediction. One of the points I made was preparing the publication quality pictures of the models. While the Rasmol (I’m linking to open source version here) has definitely its well deserved place on the scientists computers, it is not the best choice for publication figures. My personal suggestions are listed below:

• VMD by UIUC – my favourite, steep learning curve, writes POVRay files, recent version includes Tachyon renderer and is able to use a neat feature – “ambient occlusion“
• Chimera by UCSF – pretty easy to use, recent version can render biomolecules with POVRay
• Pymol by DeLano Scientific – easy as Rasmol, has internal renderer capable producing very nice images, another favourite for completely different reasons than VMD
• Qutemol by ISTI-CNR – pretty new software and to me still in alpha state, impressive real-time rendering with ambient occlusion, capable of producing images in prof. David Goodsell style (see Molecule of the Month at PDB)
• Molscript by Avatar Software, the oldest and the most difficult to use, however the clarity of the final image is often hard to beat

Of course three first programs can do much more than just visualize the protein structure – they can be used in detailed structural analysis, can do superimpositions of protein structures, analyze trajectories from molecular simulations, display density maps, deal with alignments and many other things.

Below you can find examples of images obtained with the above software. YadA adhesin picture has an “artsy” look, but at least it shows wide range of possibilities.

VMD

Pymol

Molscript