Tuesday, June 27, 2006

Paved with good intentions

I sincerely hope I'm not actually on the road to Hell (although the streets around the hospital where I work part of the time can be pretty unpleasant!), but my good intentions of posting regularly here certainly didn't pan out. Such is life...let's see how it goes.

I've been pretty busy for the last few months. Several paper deadlines, finishing off old projects from my last job, lending a hand to the experimentalists (of which more anon) and a new project. The latter is, of course, the most exciting, so I'll describe it briefly.

I do network analysis, and to date I've been mostly working on structural networks - that is, networks based upon evidence of physical interactions. Protein-protein interactions within the cell, genetic regulatory interactions, and (one of my minor triumphs) social networks on the internet. That was a triumph because it was perfectly valid research-I even published a paper on it-but it also was the perfect excuse for spending work hours on IRC. Not quite as good as that project on the ecology of spiders living on tropical beaches that I have in mind, but entertaining! It was a great data set, too.

However, structural networks have their limitations. Biological interactions come in all flavours, even if you restrict yourself to interactions occurring within a cell, and a single type of network can only capture a subset of the important interactions. It makes a lot of sense to combine different types of interaction data into a single network of functional interactions. Doing this raises lots of issues, however: what data do you include; how do you deal with missing data; how do you estimate the reliability of data sets; how do you combine different types of evidence in a proncipled way? And, of course, once you've built your network, what do you use it for; what sort of insights can it give that, say, a protein-protein interaction network won't? Needless to say, I have at least the beginnings of ideas about all of these questions, and I'll go into the issues in more detail later. For now, I'm starting on database design and data collection, using the impossibly charismatic Bacillus subtilis as a test organism. Anyone who can point me to any sort of data on intracellular interactions in B. subtilis feel free to let me know!

Back to flogging the cluster, before the IT guys realize I'm hogging it and come to tell me off...