Saturday, January 27, 2007
Evolution vs ID vs Descartes: Part I, SPF
Part I: Single points of failure in networks of protein interaction, cancer as a result.
Evolution vs "intelligent design" is a public debate that is raging in the US. See cover of Time. Out of irritation, I have decided to go and take a look for myself, just to make my own opinion.
One of the tenets of intelligent design is that nature is so well designed, so "perfect" that, well, very simply there had to be a designer. SOMETHING interfered during evolution. The blind statistics could never account for the complexity we see in Nature. The chances are too small.
Coming from a software background, the idea of "system design" is quite natural. It is something we do everyday. But also consider this: we, humans, as a specie, are ALREADY doing "intelligent design". We routinely manipulate the genome of plants every day. Think of transgenic crops, corn is a great example. These have been totally engineered, designed, at a genomic level.
Imagine someone looking at the genome of these transgenic plants, without knowing it was designed. Can one find telltale sign of engineering? Or can all of it be attributed to luck?
Unfortunately luck is always an irreducible argument. One can talk about "abnormal" statistic (Say that an event happens MORE than its odds would indicate) and a lot of science is based on these statistical approaches. But it is not because something is rare that it will not happen. In other words, you can always explain something by statistics by saying "it is possible", its odds are non-null.
A similar kind of argument has puzzled physicists for some time. It is now understood that if some of the fundamental constants of nature were not what they are to a dizzying degree of precision, then our universe simply wouldn't exist!. What a miracle then, that those constants are what they are, just what they are. What invisible hand sets and holds them at those particular values? Are we just really lucky?
It may be, again that is the trouble with statistics. That something is unlikely doesn't mean it doesn't happen! It does! at a rate of its odds. There is ONE lottery winner for 10M players! And yet, there is new winner EVERY DAY on earth. EVERY DAY. Do you believe there is divine intervention for each and every one of them? They might believe it, you know better, it just called luck.
One out of a billion sperms reaches the egg and fertilizes it. Hey! feel proud! you, yes you, already won an amazing race of one in a billion. The odds? staked against YOU in there, the reality? you won! See? you are yourself already very lucky!
Some theories of theoretical physics, in fact predict that all these multi universes "exist" in parallel, some go to hell immediately, some are stable, some a so hot nothing lives in them, some harbor life, just like our universe. We live in one such universe where life exists, let's count our blessings! We are lucky to be alive!
I am mostly convinced that statistics are self-sufficient as an explanation and that small odds do not require divine intervention, just luck, by definition of what odds are. I am willing to accept that randomness, luck and evolution have produced nature.
Unfortunately, I would draw the same conclusion on transgenic corn if I didn't know better. Namely that even if the odds of corn evolving JUST THE RIGHT properties as if engineered could have evolved out of luck.
So I wanted to look for myself at some of these "designs" of nature. I have been attending lectures at Georgia Tech and reading up on "network biology" a discipline of systems biology.
See, most of the protein-protein interactions constitute "pathways", way to get stuff done, proteins produced in the right amount. The pathways are well understood in humans they produce protein that fullfil a particular function. Many pathways remain unknown and yet to be discovered but by and large there is a wealth of information already.
Guess what? it all looks just like any big network. A tangled mess of connections. My friends in the software business and in IT, are already familiar with the notion of "single point of failure". Think about the internet: one of the points of the internet, by design, was to sustain a nuclear attack: there was to be no single point of failure. Single points of failures in a corporate network, for example, are a big no-no.
Take away a large chunk of the network and you can still find a way to go from A to B. IT may take longer but it is feasible. THAT is redundancy of pathways. A sys-admin will get fired if he/she has mis-designed his networks with "single points of failure", and the stuff actually fails. That is what keeps the data of corporate america humming.
Same thing is going on in nature. There is great redundancy in nature's pathways. Highly conserved or ancient pathways usually show a great degree of inter-connectedness and redundancy, thereby lending them robustness and more longevity. A critical biological function usually comes with a highly redundant pathway. It is tough to nuke it, even if some links fail there is always more than enough to fullfil the function.
But here is the part that that I find interesting as you observe these pathways: THERE ARE MANY SINGLE POINTS OF FAILURE IN NATURE'S PATHWAYS.
Cancer is in many ways a result of this. See, cancer is not a "thing", it is not a virus or a bacteria, it is not an 'organism' that invades you. It is a genetic disease, a failure from the inside, a failure of our own genetic code. A bunch of otherwise normal cells will lose some genetic material, or gain some, therefore loosing key pathways and gaining abnormal pathways. Their program goes bonkers.
Specifically, cancer seems to result when cells LOSE the "programmed death" pathways and they also GAIN the capacity to abnormally up-regulate their growth resulting in hyper-growth. Basically they evade death and grow like weeds, all because their pathways are out of whack. The scientific names used to describe this are "apoptosis" for cell death and "onco-genesis" for cell growth.
BTW it is hypothesized today, that in the early stages of cancer, lateral gene-transfer takes place. This is a mechanism akin to what we have come to know in software as AOP (Aspect Oriented Programming) where a particular function is injected in code execution in a cross cutting fashion. AOP is needed to maintain large scale system in a efficient way. OO systems (inheritance based, Mendelian) are just to brittle.
In cells this is achieved by retro-viruses. Retro-viruses will carry an alien sequence and will inject it in the cell. This sequence will persist in DNA. In software this is called "weaving". AIDS is probably the better known retro-virus. Not all "retro-viruses" are bad, in fact it is hypothesized that most are benign and many beneficial. We tend to focus on pathologies but the mechanism is present in nature for good purposes: efficient maintenance.
For example cross cutting transfer of genes has been observed in the wild with transgenic corn. Corn apparently has been using viruses to propagate the features. Since the features are beneficial, the fitness goes up and the feature persists in a larger part of the population, at which point it is transmitted by reproduction and normal mendelian inheritance. Not all viruses carry death. They just carry DNA and RNA, most of it garbage, some of it deadly, some of it beneficial.
This cancer as pahtway degeneracy is but an example. Human pathways are FULL OF SINGLE POINTS OF FAILURE Since we never design this in software, that puzzled me for a while. Why would a designer, if there is one, would introduce points of failure everywhere and willignly?
See failure induces death, which is selection mechanism that allows more fit cells to survive. So network topology is linked to fitness and therefore evolution. Single points of failure favor rapid evolution by death. SPF are evolution at a network level.
Could SPF be designed for evolution?
to be continued...