Saturday, March 19, 2005

Final rotation

As of about a week and a half ago, my rotation with Tom Knight [or TK, as he’s generally known around here] is over and I’ve moved on to my next rotation. Executive summary of my rotation with TK: he rocks ! The man seems to have detailed knowledge about everything related to prokaryotes [ie bacteria and such] and, I suspect, quite a bit about higher organisms too. What makes this all the more impressive is that he was a “pure” computer scientist for 20+ years and then basically retrained himself to become a biologist. Now that’s what you call mental flexibility ...

I originally had grandiose plans to do some work on an idea I had about programmed cell-cell communication, but after actually working through the details on paper, I realized that a) it was going to require a huge amount of lab work b) I’d be lucky if it worked at all and c) the system I was thinking about wasn’t really as powerful as I’d originally hoped [but if anybody has ideas about how to get in vivo non-templated synthesis of DNA, let me know :-)]. In other words, it wasn’t something that made sense to try during a 2-month rotation while also taking classes. So I went to TK and offered to be his monkey if he needed any experimental work done – I figured whatever I ended up doing for him would be educational since I still barely know the difference between the blunt and the sharp ends of a pipette, so to speak.

And was it ever educational: since Tom only has 3 students and isn’t teaching any classes, he’s around a lot of the time and still does all his own lab work. Couple that with the aforementioned prodigious knowledge and willingness to impart it [and answer dumb questions] and you get what I had: one-on-one instruction from somebody who really knows what he’s doing, which is priceless. Unfortunately I didn’t get a chance to work with him as much as I would have liked, but, let me repeat: TK rocks !

My next rotation is with Prof. David Gifford, who is one of the instructors in my “Computational Functional Genomics” class. His group comes up with all kinds of smart algorithms to answer the “You’ve got a mountain of experimental data. What is it telling you ?” question that I alluded to in an earlier post – lots of machine learning-type stuff. Conveniently enough, I can reuse the work I do in this rotation for my class project in his class and the TA for that class [in which I’ve “understood trainstation” a few times] is also in his lab, so I’ll be able to pester him more easily [Yea, verily, he shall come to dread the sound of my rough feet slouching towards his office door ... or something like that].

My rotation/class project is to try to [re]discover protein-DNA binding motifs based on ChIP-chip binding data. Running this through the degobbledygookifier produces this explanation: in order for a gene to be expressed [ie to make a protein], a bunch of other molecules have to bind to [ie “land on and stick to”] the DNA. The places on the DNA where these other molecules bind are something like “landing strips” where these molecules can land and stick firmly. These landing strips tend to have common features, just like most landing strips have common features like lights, asphalt etc, and the common features are called “motifs”. Based on this idea of common features, people have looked for motifs by basically looking for [relatively short] areas of DNA that seem to have things in common eg they all have the sequence “GGGTGCA”. However, one of the problem with finding these motifs just by looking at the DNA is that the fact that something looks like a motif [ie the molecules could land there] doesn’t mean it’s actually a place where the molecules do land. So, biologists being the smart people they are, they came up with a way to actually trap the molecules after they land, kind of a molecular flypaper. So now you can look for motifs by trapping the molecules after they’ve landed and then seeing whether the places they’ve landed have anything in common. If they do, well, son, you’ve just found yourself a motif ! [Sort of]

So the plan for my rotation is to find some of these motifs using some data Dr.Gifford’s group has for yeast DNA and then, depending on how well things go, maybe for human or mouse DNA. Along the way, it looks like I’ll get a chance to dig into the code behind a popular motif-finding tool called MEME and hopefully finally really understand how an Expectation Maximization [EM] algorithm works. We’ve talked a lot about EM algorithms in class but I currently feel a bit like Justice Potter Stewart felt about pornography: “I can’t define it, but I know it when I see it”. Hopefully at the end of this rotation, I’ll not only be able to define it but also create my own ;-)


Blogger Son1 said...

Tread softly, because you tread on my office-hours. :-)

(I suppose you could also title this, "Aedh Wishes for the chIP-chip Data of Heaven," heh...)

5:49 PM  

Post a Comment

<< Home