Monday, June 27, 2005

Wandering in the microRNA wilderness

Pretty much my biggest goal for this summer is to find myself a thesis project that involves working on microRNAs. I feel a bit like Goldilocks in that I want something that doesn't have too much wet bench work, or too much computation, but is just the right mixture of the two. Oh, and I'm picky about what kind of computation I want to do too -- I want to do something "systems-y", something that is related to the dynamic behavior of a system, which means I don't want to do any/very little sequence analysis. And it'd be nice to work with mammalian cell types in which microRNAs have been shown to play a big role, like blood stem cells or embryonic stem cells.

To get some ideas for thesis topics, I've been reading the existing microRNA literature, in a somewhat scattershot fashion. The papers out there seem to fall into a fairly small number of categories [the links are to pseudo-randomly selected papers in each genre]:

1. Computational prediction of microRNA genes and microRNA target genes, with a bit of experimental validation tacked on. Everybody and their mother seems to be working on this, with a couple of papers published [almost] every week, trying to outdo each other in terms of how many microRNAs and targets they predict. The modus operandi seems to be "more is better".
2. Hard-core biochemistry aimed at trying to figure out the actual molecular machinery of the microRNA pathway, like so.
3. Papers of the form "microRNA X affects process Y because if we take away microRNA X, process Y breaks down", like this.
4. Studies along the lines of "This is what it looks like when you track microRNA expression levels during the development of organism Y [but we don't really have an explanation for the changes in expression level]", for model organisms like zebrafish.

... and then there are a bunch of review papers that try to make sense of the flood of data.

Now, these are all perfectly reasonable things to do, but none of them really float my boat -- type 1: sequence analysis, blergh; types 2 & 3: no computation involved; type 4: so far, the only type of computation I've seen in this sort of paper is some pretty mundane clustering.

In other words, systems biology-type work on microRNAs seems to be non-existent at the moment, or [more likely] not published yet. I think that's because the microRNA field is still relatively young -- to build a detailed differential equation-based/stochastic model of a particular genetic pathway that involves microRNAs, you need to know what turns microRNA expression on/off and what the targets of the microRNA are, and the majority that knowledge simply doesn't exist yet. The other systems bio approach, namely taking a bunch of expression data and then trying to reverse-engineer a network out of that [eg via Bayesian networks] relies on having a reasonable amount of decent-quality expression data, and that doesn't exist yet either [partially because there's still lots of debate about what is and isn't a microRNA, and the list of known microRNAs keeps growing]. Throw in the fact that the microRNA pathway only exists in metazoans [ie multi-cellular animals], and those are much more complex and less well-understood than the bacterial/viral/single-celled eukaryotic organisms that most systems bio work has been done on so far, and it's not surprising that there's a dearth of that kind of research.

From one perspective, this is good news -- it means the field is wide open, unlike the pig pile that prediction of microRNA genes and microRNA targets has become. From a different perspective, it's a bit frustrating to not have any obvious "prior art" from which to draw inspiration [or maybe I just don't see it yet].

In the meantime, I'm just going to keep thinking about things I could do and hope that inspiration strikes sometime soon. Oh, and if you have any ideas that you can bear to part with, feel free to send them my way ;-)


Anonymous PhilipJ said...

This is totally my beef with research too. The really cool work that comes out and makes an interesting and novel contribution to science _has_ no real template from which to work (most of the time). Doing me-too science (e.g., variations of a theme) is still fine, and is the fast majority of papers that get into Science, Nature, et al., but coming up with that gee-whiz new science project is definitely not easy! I think coming up with a really novel science project that is tractable is just as hard as the project itself!

PhilipJ |

10:22 PM  
Anonymous Bill Tozier said...

There is another way -- a Kauffmaniac way. That would involve some sort of abstractification of the process into discrete mathematics, some imposition of reasonable-sounding rules that describe arbitrary interactions of the sort hypothesized among the molecules in question, and large-scale Monte Carlo modeling of artificial systems with a wide range of parameter settings to determine universal organizational traits.

Followed by ten years of dismissal by the lab-using mainstream, rediscovery by fringey weirdos, and near-posthumous accolades as being "before one's time".

So call that "5".

12:59 PM  

Post a Comment

<< Home