Monday, June 19, 2017

Austin Hughes and Neutral Theory

Austin Hughes (1949 - 2015) died a few years ago. He was one of my favorite evolutionary biologists.

Chase Nelson has written a nice summary of Hughes' work at: Austin L. Hughes: The Neutral Theory of Evolution. It's worth reading the first few pages if you aren't clear on the concept. Here's an excerpt ...
When the technology enabling the study of molecular polymorphisms—variations in the sequences of genes and proteins—first arose, a great deal more variability was discovered in natural populations than most evolutionary biologists had expected under natural selection. The neutral theory made the bold claim that these polymorphisms become prevalent through chance alone. It sees polymorphism and long-term evolutionary change as two aspects of the same phenomenon: random changes in the frequencies of alleles. While the neutral theory does not deny that natural selection may be important in adaptive evolutionary change, it does claim that natural selection accounts for a very small fraction of genetic evolution.

A dramatic consequence now follows. Most evolutionary change at the genetic level is not adaptive.

It is difficult to imagine random changes accomplishing so much. But random genetic drift is now widely recognized as one of the most important mechanisms of evolution.
I don't think there's any doubt that this claim is correct as long as you stick to the proper definition of evolution. The vast majority of fixations of alleles are likely due to random genetic drift and not natural selection.

If you don't understand this then you don't understand evolution.

The only quibble I have with the essay is the reference to "Neutral Theory of Evolution" as the antithesis of "Darwinian Evolution" or evolution by natural selection. I think "Neutral Theory" should be restricted to the idea that many alleles are neutral or nearly neutral. These alleles can change in frequency in a population by random genetic drift. The key idea that's anti-Darwinian includes that fact plus two other important facts:
  1. New beneficial alleles can be lost by drift before they ever become fixed. In fact, this is the fate of most new beneficial alleles. It's part of the drift-barrier hypothesis.
  2. Detrimental alleles can occasionally become fixed in a population due to drift.
In both cases, the alleles are not neutral. The key to understanding the overall process is random genetic drift not the idea of neutral alleles—although that's also important.
Originally proposed by Motoo Kimura, Jack King, and Thomas Jukes, the neutral theory of molecular evolution is inherently non-Darwinian. Darwinism asserts that natural selection is the driving force of evolutionary change. It is the claim of the neutral theory, on the other hand, that the majority of evolutionary change is due to chance.
I would just add that it's Neutral Theory PLUS the other effects of random genetic drift that make evolution much more random than most people believe.

Austin Hughes was a skeptic and a creative thinker who often disagreed with the prevailing dogma in the field of evolutionary biology. He was also very religious, a fact I find very puzzling.

His scientific views were often correct, in my opinion.
In 2013, the ENCODE (Encyclopedia of DNA Elements) Project published results suggesting that eighty per cent of the human genome serves some function. This was considered a rebuttal to the widely held view that a large part of the genome was junk, debris collected over the course of evolution. Hughes sided with his friend Dan Graur in rejecting this point of view. Their argument was simple. Only ten per cent of the human genome shows signs of purifying selection, as opposed to neutrality.


  1. The vast majority of changes of gene frequency are due to genetic drift. However a lot of those changes are in sites in the genome that have no phenotypic consequences. In sites where there are effects on phenotype (regulatory or coding sequences) it is still true that most changes of gene frequency are due to drift. Most of that is jiggling of gene frequencies back and forth due to drift. In asking about "most changes" we are adding up the abolute values of the changes, not considering that many of them cancel each other out.

    At those sites, does that establish that most net change in gene frequencies is due to drift? No!

    At those sites, does that establish that most net phenotypic change is due to drift? No!

    Note the difference once one focuses on net change.

    In a beaker with many tiny particles slightly heavier than the liquid, most change is due to Brownian Motion. But the particles do show a tendency to settle.

    1. Let's just consider fixation. If you compare the human and chimp genomes, what fraction of the difference is due to fixation of nearly neutral alleles by random genetic drift and what fraction is due to selection?

      The situation wrt alleles that affect phenotype is controversial, as you well know. We just don't know what percentage of those alleles are effectively neutral, what percentage are detrimental, and what percentage are associated with significant selection coefficients.

      The common assumption is that almost all alleles affecting phenotype are not neutral. But this is the adaptationist position that assumes, as the null hypothesis, that all phenotypes are subject to selection. That's not a valid assumption.

    2. So, are we at least agreed that the mere fact that most changes of gene frequency are the result of genetic drift does not settle these issues?

    3. Joe, I don't know what issues you want to be settled.

      The main "issue" as far as I'm concerned is that most people don't know about Neutral Theory and random genetic drift. Unfortunately, that includes a great many scientists. I met some of them at a Royal Society Meeting in London last November.

      I'd be happy if all scientists and most of the general public realized that there's more to evolution than natural selection.

      We still have a long way to go ....

    4. Sorry, I was unclear. I guess I was reacting to your earlier statements, not Hughes's views. You like to say that most gene frequency change is genetic drift. True in one sense, but to what extent does that imply that drift overwhelms selection?

    5. I think you are having an unreasonable discussion here. If I had a biased coin with a probability of .55 of getting heads and tossed it a million times getting 550,515 heads, it doesn't make sense to me to say that some of these head results are due to coin tossing, while others are due to bias. And thus it wouldn't make sense to me to ask whether more of them are due to coin tossing or the bias either. Each of them is obviously produced by the same process and thus making a difference between one head result and another does not make sense.

    6. @Joe

      Let's consider the case where a new beneficial allele appears in a population. Doesn't drift overwhelm selection most of the time?

    7. Each of them is obviously produced by the same process and thus making a difference between one head result and another does not make sense.

      Does it make sense to you to tease out flips in which the environment may have played a role? Say you were flipping an object shaped in such a way that the wind might affect which side came up, and you were in a location where the wind was blowing during some flips and not others. Would it make sense to separate those where the wind was blowing from the others?

    8. @Judmarc: Not in the model I gave. And there is a simple model that works like this:
      There is a population of N haploid organisms, some of which have allele A.
      We are considering pairs consisting of the next organism to reproduce and the next organism to die. In each case this leads to an increase of the allele frequency of A by 1/N, a decrease of 1/N or no change.
      We now ignore the instances where no change occured.
      THen we obtain the probability for an increase in each of the remaining steps as 1/(1+e^-s) and for a decrease of 1/(1+e^s)=1-1/(1+e^-s). I.e. we do get a biased coin toss model for evolution in that population (the simplifications we made make it bad at some things, you can't work out the mean time to fixation for instance, but it is accurate enough to figure out things like the probability of fixation for mutant alleles for instance). It's worth noting that s is the log of the mean fitness of A-type individuals divided by the mean fitness of not-A-type individuals. There might well be differences between the fitness of organisms and in the case of haploid clonal organisms these are simply twice the probability of reproduction before death. But what we are interested in is how allele frequency changes and that only depends on these mean values (for instance if there is some other locus affecting fitness and there is no LD between that locus and the one we are interested in, this would result in differences between the fitness of individuals, but we can ignore it when we look at the locus we are interested in, because these differences have nothing to do with allele A).
      So, if you want to go that route you end up looking at entire genomes rather than particular genes for which you want to track allele frequencies and you look at individuals, rather than populations. I would argue that at that level you can not really talk about things like selection anymore, because these are explicitly about populations. Individuals reproduce and die. And if a lot of individuals reproduce and die we can rewrite the aggregate effect of these births and deaths as selection and drift (or, as I keep on argueing, just treat them as a single resampling process).
      I think a while ago I had a similar discussion on a forum, where the poster argued that there were instances of selection, like a cheetah catching the slower gazelle and I pointed out that our model asks: Which of the Gazelles is the next to die. That of course also depends on none of the other Gazelles getting struck by lightning first.

    9. In 1973, “A Random Walk Down Wall Street” was published.
      It makes the case that the movements in stock prices could be well modeled as a ‘random walk’ and therefore trying to use mathematics to ‘time the market’ won’t work.
      I believe the hypothesis has proved to be correct.

      But what moves the markets are the ‘buy’ and ‘sell’ decisions of the individual investors. The person doing the buying has a reason for wanting to buy at that price, and the person doing the selling has a reason to accept the price as well.
      The individual decisions are not done ‘at random’, yet the conglomeration can be modeled that way.

      I bring this as a possible analogy for evolution where the changes in allele frequencies can be modeled as ‘at random’, but the individual acts of reproduction and death might not fit that description so well.

    10. @Simon Gunkel: In your coin toss example, over a very large number of trials the incidence of heads would presumably get quite close to 55%. If it stubbornly did not, I should think that might be of interest. Or to make it even more simple, if you are tossing a coin and not approaching 50% incidence of heads over a very large number of trials, you might be interested in what was causing it to come out very close to 55% instead.

    11. True. But in my example the coin tosses were IID, which means they would converge on some value and it is easy to note that the MLE for the probability of getting heads is simply the number of heads divided by the number of tosses and that we can fit confidence intervals around these rather easily as well. If in the simplified moran model we had some number of increases vs. some number of decreases we could perform statistical tests to check whether we could rule out neutrality (s=0) with any confidence. We could compare this to some other allele and check things like whether the respective selection coefficients are significantly different from one another.

  2. What a creationist can draw from these corrective ideas in the subjecty of evolutionary biology is as follows.
    First a wrong idea would of prevailed until superior investigative tactics were made.(study/molecular polymorphisms. A original error in evolutionism is always a option.
    Second. The correction overthrows natural selection at the genetic level. this because natural selection does not work at that level they now decided.
    It means as better investigation takes place, at a more atomic level, in real time, Darwins evolution ideas oncve again come up short. so new ideas are needed to save the day. Like stephen goulds attempt to save the fossil record for evidence of evolutionary change. He gave up a little to save a lot. it failed but he knew there was a serious problem.

    So from these things a creationist can draw a deeper philosophy about scientific investigation and so conclusions..
    Without having a competence to deal with the subject at a high or even entry level status.

  3. "Only ten per cent of the human genome shows signs of purifying selection, as opposed to neutrality."

    Serious question: does that mean that fixation by drift is mainly affecting the non-functional areas, while the functional areas are mostly subject to selection?

  4. Ordered a used copy of Evolution and Human Kinship from Amazon. The description tells me it was a library copy from Institute of Creation Research. Does this mean they were good for something?