Molecular Evolution Activities

Changing Terms Transcript

Changing Terms of the Neutralist-Selectionist Controversy, and Prospects for Resolving It


Michael Dietrich: One of the things we were also interested in was having a discussion about where the neutralist-selectionist controversy is moving. Has it gotten somewhere? And if not, what would it take?

Richard Lewontin: We have seen something about that on the website already.

Dietrich: You [Ohta] have written about this, with Marty Kreitman, and that is what got us thinking that maybe people were ready to talk about this—not necessarily from a historical perspective, but from a "Where are we now?" assessment.

Tomoko Ohta: If we look at all cases, there are some neutral mutations, some nearly neutral mutations, and some selective mutations.

Lewontin: That's the way it is. Have you all seen that April [2002] paper of Bustamante and ten other authors on Arabidopsis and Drosophila? He's collected thirty-six different genes. And sure enough, in Drosophila the estimates of selective replacement were on the positive side, but most of them were non-significant. Half of them were non-significant, half were significant. And for Arabidopsis they are all on the negative side. The fixations were against selection. But all small. Now you have forty genes with which you can then make a rough statistical generalization. Their argument was that Arabidopsis is different from Drosophila, because Arabidopsis is selfing. Therefore, it has a smaller effective population size. And therefore the probability of fixing quasi-deleterious genes is greater, and so that's what you see. And so I think that kind of paper, which brings together a lot of stuff, demonstrates Ohta's point, that it's some of this and some of that, and then what we have is a statistical generalization. End of story.

William Provine: But surely we have a more specific situation than that, if we pay attention to the different levels. At first it seemed to me that the neutral theory had bifurcated evolutionary biology. Molecular evolution had its set of drummers, and it was sure as hell not the same drummers that you had for phenotypic evolution. But then I realized, that's not going far enough. Because, if you're talking about a DNA sequence level, which is what I think Kimura had in mind from the beginning, he wanted to get to the genome level, but he only had protein data to work on. And that's a lousy way to figure out what's really going on at the DNA sequence level. That's the data he had. And there were all kinds of problems with it.

So she [Ohta] invents a nearly neutral theory that is more appropriate for the protein data, in her opinion. But Kimura can't quite grasp where to go until the DNA sequence data comes in. In his book, he's right in the middle of it—he doesn't know where. Then by 1985: "Wow! I have the DNA data. I know where I belong. This is mine." So, you've got a DNA sequence level for one. You've got a protein level for another. You've got a membrane level, which Kimura was not particularly interested in, but which is very real. And those three levels are, surely, as Tom Cavalier-Smith argues, all "interacting." So you can't talk about one level by itself. Nevertheless, you can see right off that the DNA sequence level is much more appropriately understood, especially as null hypotheses, by the neutral theory—by the completely neutral theory. But you can't do that at the protein level.


Doolittle: It seems to me that the issue is not the data about specific instances, because some of them are neutral and some of them aren't, and in theory we know how to figure that out. But it seems that the underlying philosophical schism between neutralists and selectionists is always there and comes up often in genomic thinking. People occasionally start looking at new genomes and say "The way we know that the nucleotides really matter because the cell is a vast machine…."

Lewontin: It's ideology.

Ford Doolittle: There's a big ideology about this. And there's the line that it's just junk DNA. So there is that aspect of the selectionist view.

Provine: [addressing Doolittle] But in your field, would you say that that the null hypothesis is always now selective neutrality? You sequence DNA and ….

Doolittle: No.

Lewontin: [in response to Provine] Well, that's partly because if you don't make that that the null hypothesis, then you are swimming in parameter space.

Provine: But before Kimura's theory, selection was always the null hypothesis.

Doolittle: [in response to Provine's question] If you are talking about people doing genome evolution then, no. There are certainly people in my field who would say that all those insertion sequences and all that repetitive DNA has a function.

John Beatty: Well, we maybe just subtly switched questions. As it was originally posed, is the neutral theory correct? Can we even talk in those terms? Or can we only say that it applies to a certain number of loci, and selection applies to other loci, or it applies to one level and the selectionist accounts apply to another level? Or it applies to one functional class of genes and selectionist theory applies to a different functional class? We are changing the terms of the resolution of the debate.

And also Will [Provine], to switch to the issue of whether the neutral theory is the appropriate null model for investigating evolution, isn't that a somewhat different question than "Is the neutral theory true?" It has to do with whether the neutral theory is a good investigating device.

Lewontin: Right.

Provine: Your null assumption is not necessarily that you accept that theory.


James Crow: I think what you [Beatty] are saying is correct. We're in that era where we study it case by case. And we ought to stop talking in these gross generalities that do not have much meaning.

Provine: These generalities do mean something. On the average, it's a heck of a lot more obvious that selection is operating on the protein level than at the DNA sequence level.

Lewontin: But there are a lot of constraints at the DNA level—selective constraints.

Provine: When you measure by various tests whether things are selectively neutral or not, they much more closely approximate neutral at the genomic level than at the protein level.

Lewontin: Look, we have codon usage bias, a major issue that you must understand. That's not neutral.

Provine: I agree.

Lewontin: And it has it has phylogenetic patterns. It differs between proteins. Codon usage bias is universal. You have patterns of conservation in three prime regions of a gene where they don't belong. What is going on there? Unless it is a recent pseudo gene, I think the evidence is clear that selective constraints exist all over the nucleotides.

Doolittle: It's like the c-value paradox.

Lewontin: Of course.

Provine: Why did Marty Kreitman come to your conference [honoring Lewontin on his retirement]? To say that, "You know I've been working on this Adh locus for a very long time, and by gum, I did all these sequences and all these tests, and I hate to tell you folks, but it's all neutral."

Lewontin: No, no. Look. In order to do the protein level, you have to make as a first approximation that the so-called "synonomous" positions don't matter. You make that as a first approximation in order to detect a signal, above that level, of the importance of the amino acid substitutions. But you don't have to take seriously the claim that the level is really zero, because all of the evidence that we have is that it is not zero. Synonymous positions are under selection. Of course they are.

Crow: It's just small compared to the other things.

George Smith: Correct me if I am wrong, but what I just heard is that reasoning in the field has changed in the last thirty years. My question is, to what extent did the initial controversy help that process of reconsidering evidential reasoning and laying it out with care, or would it had happened regardless?

Lewontin: It created a problematic for which methods had to be invented. Most of the statistical methods invented, for example, have been been invented deliberately. The whole of coalescent theory, and the statistical testing that comes out of that, was invented for a particular problematic, which was, "How can we estimate the selection coefficients on amino acids for example, or any character?"

Beatty: Do we agree that it is worth having a module on how the terms of the neutralist-selectionist controversy have changed over time? Or how have the terms of resolution of the controversy changed over time?

Lewontin: That's where the action is.

Beatty: I think John Gillespie would be happy to initiate a discussion like this and maybe we could pit him against Will [Provine].

Lewontin: Where there's a Will, there's a way.

Beatty: Towards the end of the packet that I passed out, next to the last piece, we have John Gillespie's comments on my question about, "How would you ever know if you had settled the neutralist-selectionist debate?" He basically sides with Warren Ewen's comments, in the online discussion [Discussion: "The Early Reception of the Neutral Theory"], that it really does not make sense to talk about the truth, the correctness, of the neutral theory, or the nearly neutral theory, or even his particular selectionist models. It comes down to a matter of proportions. We could just have John clean this up a bit and post it. Maybe Will can post an alternative view of things.

Lewontin: I think his [Gillespie's] comment is right on.

Crow: Me too.


Return to the Dibner Workshop Homepage