Turned up to eleven: Fair and Balanced

Wednesday, April 24, 2002

Godless Capitalist has put up a well reasoned response to my post below, and makes many sound points. I disagree with him on some of them, but I am glad that he accepts a more intricate model of the phenotype/genotype relationship than is often used. I have a bone to pick, however, with his characterization of my comments below.

3. Orwin brings up the complexity of the immune system as a case where measuring the phenotype - the ability to fight off disease - is a "non-starter". In many ways the analogy between intelligence and immunity is a natural one. For example, the concept of inoculation is predicated on the immune system's memory of previously seen pathogens. However, I feel that Orwin is inaccurate by claiming that this complexity means that it is impossible to systematically study intelligence, for several reasons.

I did not claim that this complexity means that it is impossible, merely that it is difficult, and "complex" (in the mathematical sense. I also think that "Godless Capitalist" misinterprets my statements about the immune system. What I meant was that inferences about the development of the immune system are not easily or simply drawn by looking at the phenotype. In the simplest case, looking at resistance to a bacterial infection, for example (as in the Pasteurella multocida example, a phenotype inference is made by noticing the acquisition of protection using an attenuated strain. However, this is completely useless with respect to understanding the genetic elements of immunity, which was my point.
I am pleased to see the state vector model of bacterial functions being presented, and I think it is worth examining in detail. One point that my (adversary? partner? opponent?) may have brushed aside too quickly is the notion of asynchronous stimuli He suggests a model of a massive difference equation (a discrete version of a differential equation) using a very large state vector to tell you the state of the organism at time t, and allowing you to find it for time (t+1). This is, in fact a model that is very useful, and I am fairly certain that there is an online E. coli simulator that uses something similar. The use of state vector approaches to assess the state of every integral (non-subdivided) part of the organism is a classic "bottom up" approach. It is also a very difficult proposition. Just for example, there are about 4000 genes in Escherichia coli, probably the best studied organism on the planet. About 40% of these genes have unknown functions (last time I checked, which was a while ago, but I am sure that hasn't changed too much). Add to that the interactions between gene products and the products of enzymatic pathways, membrane dynamics, diffusion and active transport issues, and you are talking about a very difficult problem. I haven't actually used the "E. coli simulator", but, while I expect it is pretty cool, I haven't seen any published literature that used it at all, so it is probably a pretty incomplete model. When you take into account the incredible scale-up problems you would have trying to apply this to an organism with trillions of cells, interacting with one another, you can see the potential problem. As an aside, in my later post I will discuss some potential cutting edge computational tools (neural networks, DNA computers) that may actually be brought to bear, someday, on these issues. However, even with these models, it is clear that genes in bacteria set a range of activities for the organism, and dictate, to some extent, when those activities occur based on environmental stimuli. There is debate in the field over to what extent environment can have a basic effect on genetics in bacteria. (Note; very large aside follows!!)

There is a system in bacteria called the SOS response, which occurs under conditions of stress for the organism (not "my job is too hard" stress, but the real thing), such as nutrient deprivation, irradiation, etc. It is a "global" response, meaning that many of the cells systems are affected. One interesting aspect of this system, however, is that it can change the way the bacterium deals with DNA proofreading, that is, what happens when the enzymes responsible for replicating the DNA make mistakes. It turns out that bacteria use different strategies depending on their lifecycle stage, and can increase their mutability under certain circumstances. They also trigger the activation of latent viruses in their genomes, which can drastically affect their lives (the viruses break open the bacteria), but that is probably a function of the viral genomes' activation in response to the stimulus, rather than the bacterium. This is an example, however, of a case when the environment actually directly causes a change in the genotype of the organism. Somatic hypermutation in immune cells is another, but it isn't germ-line, so it doesn't really count. (end of extended aside!)

Finally, my argument was never about the value of IQ or g as a predictor of success in our society, or to any other. That is a sociological argument, and completely irrelevant to the question of IQ and genetics, which is my interest. As an aside, the idea of Genius will come up in this discussion later, and how it differs, IMHO, from intelligence (I am not sure I am qualified to comment on either!)