Monday, July 30, 2007

Biology is noisy

We've known that for a long time. I have friends who choose not to study biology simply because it is so noisy.

I also have friends who assume that the noisiness of biological systems is an artifact of the less precise means we use to measure and control them. I like this article because it shows that if we were to remove the noise from biological systems, we would be removing some of their characteristics -- that noise has a purpose in biological systems. In fact, it's entirely possible that Bacillus cells amplify the noise inherent in a purely chemical or physical model of gene control through secondary measures. Basically, these bacteria want to be noisy, because that noise is the variation that genetics works with. In the overpopulated conditions at which Bacillus is competent, it is trying to take up new pieces of DNA in the hopes that some of them contain useful genes for dealing with the environment (similarly, heat-shocking some E. Coli strains will make them competent). But it's a chance - it's not certain that there's any DNA out there or even that any DNA that is out there won't kill the cell more surely than the overcrowding will. So in order for the species to survive, there has to be a mechanism by which some, but not all (and maybe not even most) cells turn competent. Hence the use of noise.

The one theoretical problem I have with the paper is that their method for reducing noise in the system seems flawed. They "remove" the noise by knocking down another gene - which reduces noise in gene expression but this might be because of a reduction in random noise or it might be because of a reduction in sensitivity to minute changes in the local environment. In the case of the former, then the researchers have indeed done what they set out to do. But in the case of the latter, then the researchers have simply said that when cells are less sensitive to their environments, they are less likely to react to their environments by becoming competent. Which is a bit obvious when put that way. The problem being, of course, that it's not easy to think of a way to remove that confounding variable. You can't test or control the minute variations around a single cell, since the system they are working with needs a large population of bacteria, and beyond stirring the media constantly (which they do) to get uniformity, well, there's not much of a choice. One would have to examine the effect of the mutation they use to knock down noise. There are other markers for response to crowding in Bacillus Subtilis than simply competency, or at least I imagine there are. Perhaps a slower cell cycle. One could perhaps test those secondary markers to see if they mirror the competency -- if they too are less likely, which would imply that the effect seen is a reduced sensitivity -- or if they oppose the competency -- if they do not change, indicating that the effect seen was in fact specific to the competency pathway.

And then the question is, is it worth it? Because in either case, noise and/or the amplification thereof is a necessary part of control of biological pathways. Which is perhaps a significant finding. Since it rather removes us from a strict deterministic paradigm and places us in a statistical one, I think (this is at first glance -- I might have to come back to the idea later). We're all fuzzy, noisy, there are no clear signals - and that fuzz is free will? Who knows.


Duff said...

To be sure I only skimmed the abstract. If you are interested in noise in systems, there is an interesting very fundamental result derived in statistical physics in the 50's I think called the fluctuation-dissipation theorem. Essentially the idea is that in a thermodynamic system, the return to equilibrium from a soft perturbation is achieved through the noise fluctuations at equilibrium. This is interesting, because it means that by measuring the noise of a system at equilibrium, one can statistically predict what the system will do when shoved into a non-equilibrium state, and how quickly it will take to respond, so long as the shove is not too big.

I would check out what Phillipe Cluzel does to understand more about this. He is one that likes noise in biological systems.

Elizabeth said...

I heard a lecture given by one of the people in Cluzel's lab. It was really fun. That talk was on gene networks and evolution, so very computer intensive. In that system, noise worked differently. In a random network, when you had a selecting force, the changes were within noise and such like I think you talked about, but in the "more biological" scale-free network, the changes seemed to be fairly continuous, with little noise until they reached an equilibrium (at which point there was still less noise than in a random network). It was interesting. Some of his other stuff is more relevant to the paper I pulled.