Friday, 9 September 2011

Occam's Rug?

Looking through the latest volume of the Annual Review of Biophysics I came across an interesting piece by Ido Golding: Decision making in living cells: lessons from a simple system. The article looks in great detail at the E. Coli and lambda bacteriophage system, showing that experimental results fit very nicely with theoretical and simulation based predictions. Instead of describing the paper in any detail (a job done very well by the paper itself) I thought I'd highlight some remarks made by Golding which caught my attention.

One of the papers core arguments (as per the title) is that much can be learned about decision making at the cellular level from a simple model system. However, the closing paragraph reads as follows:

...before applying the principles learned here to higher organisms, we must beware: As physicists studying living systems, we are always in jeopardy of over assuming universality. Against that ever-present temptation, one has to keep in mind the distinction between Occam’s razor and Occam’s rug: The former, of course, is the guiding rule for physicists, who will always choose the simplest, most universal explanation for an observed phenomenon. The principle of Occam’s rug, on the other hand, states the following: When studying a living system, a simple elegant narrative often implies that too much of the data was swept under the rug. In other words, always be wary of claims of simplicity and universality in biology.

I suspect the tension between Occam's razor and Occam's rug is at the heart of all biophysical models.

5 comments:

  1. What is meant by 'decision making' in the article? Does it refer to checkpoints and signalling for processes like cell division or chemotaxis? I suppose I could just read the article myself.

    ReplyDelete
  2. I think that this goes right back to our previous discussion about the difference between a model and a theory... the work presented in this paper appears to be an excellent model, but not necessarily an excellent theory in and of its own right.

    I claim that models in physics don't have the property of universality, but (as I have argued before) theories do. After all, if I have an energy level model for a single atom that doesn't allow me to extrapolate to the energies of any other atom, whereas I can apply quantum mechanics to any energy level structure.

    Here we have the same thing: a great model of the system in hand, without universality, but with definite insights into how some greater overarching theory could be applied to similar systems.

    What that overarching theory is remains ambiguous... perhaps a smattering of kinetics, information theory and chemistry?

    ReplyDelete
  3. Re: Wendy’s comment
    From what I can tell, decision making is taken to be the process whereby one of multiple possible options is selected. One example (from memory) in this paper is that the virus, after infecting an E. Coli cell, will either lie dormant or attempt to create as many new copies of itself to infect other cells. In effect “making a choice” of one of two possible actions.

    Re: James’s comment
    I suspect that when discussing any set of ideas terminology is unavoidable, I see it as the personal baggage that everyone brings to a conversation and it often gets in the way of the underlying ideas. I agree that there are fundamental differences between a set of equations that someone comes up with to predict decision making processes in simple biological systems and the Newtonian equations of motion. And when it comes to understanding the why as well as the how of things, mechanistic, first-principals approaches are always more useful.

    The original point I was clumsily trying to make (by calling all “theories” essentially nothing more than models of observed phenomenon) was that in the absence of a “grand unifying theory” there are always areas where even our most general theories fall apart; in essence, that all theories only differ in their degree of generality. As far an I’m aware there are no truly universally applicable physical theories.

    ReplyDelete
  4. Hello everyone,
    Decision making in bacteriophages is definitely an interesting subject; regardless of the universality of the model used, it is an interesting system to test if bacterial 'decisions' are due to molecular interactions. Hopefully, I can read the paper soon.... Did the paper say anything about whether the interaction was due to solely molecular interactions of the virus and the cell?

    One must offer the question, also: what if Occam's razor is used on his rug? Does it result in a blunt razor? Or a sharp rug? (Well, I tried the pun...)

    Regarding James' statement, too: I think that, despite the beasty baggage involved in the discussion (another pun on terminology), science is in a unique position to further the understanding of the universe in such a way that some elements of the world are generally described. Of course, I believe that science cannot explain everything because not everything is scientifically testable (e.g. faith in religion), so there will never be a truly 'GUT'.

    So the effective dissonance of science is the wish for general, low level models that can be extended to higher levels (e.g. quantum) and gain some (hopefully rigorous) understanding of the world (and of maths in the process) that is more general than a 'specific theory', against the wish for models that are accurate in their regime, simple and efficient, allowing us to further more models later on. I think that almost all models are designed to move towards smaller things in order to become more general. I suppose that they are moved towards larger things later...

    This is a fun pony to ride on. It makes for good blog posts, but you should be careful not to fall off, Martin, onto the 'Ground of Sense' via your clumsy application of sensible ideas :).

    ReplyDelete
  5. Sorry, I thought of this just after posting the comment above: I wonder how much data could have been 'swept' by Occam's Broom? Perhaps data manipulation would be to blame, but in the case of living organisms, their complexity at least helps to protect them from Millikan's Oil drop dilemmae.

    I do think, though, that a careful definition of 'universality' of application is required; I would be suspicious of any paper whose subtitle is 'Lessons from a Simple System' (NB: does anyone know an alternative for 'whose' for inanimate objects? Old English is acceptable) simply because that tends to imply universality in areas where it is not. I would much rather, 'Hey look, this mini-system is actually really similar to what we think big-systems work!', or 'Lo, bacteriophages don't act like humans when working towards cell destruction'. In these cases, the reader can see the potential for a universal trend without having the assumption that principles universal in the bacterial domain are obviously true in the macrobiological domain.

    With this statement, it becomes obvious to me that if some kind of universality is present in the mini-domain, we expect it to manifest in the macro-domain. But, if it doesn't, then this probably indicates we've left something out, which means more theory needs to be made. To me, this seems like a different thing to assuming/looking for universality between different systems.

    ReplyDelete