More Recent Comments

Friday, June 29, 2012

Adaptation vs Drift at Evolution Ottawa 2012

I've been looking over the program for the First Joint Congress on Evolutionary Biology to be held in Ottawa, Canada next week [Evolution Ottawa].

The talks are divided into sessions with six short presentations per session. Here's the list of sessions with the words "adaptation" or "selection" in the title.
Adaptation 1
Adaptation 2
Adaptation 3
Adaptation 4
Adaptation 5
Adaptation 6
Adaptation 7
Adaptation 8
Adaptation 9
Adaptation 10
Adaptation and Evolutionary Genetics
Sexual Selection and Mating Systems 1
Sexual Selection and Mating Systems 2
Sexual Selection and Mating Systems 3
Sexual Selection and Mating Systems 4
Sexual Selection and Mating Systems 5
Sexual Selection and Mating Systems 6
Sexual Selection and Mating Systems 7
Sexual Selection and Mating Systems 8
Sexual Selection and Mating Systems 9
Sexual Selection and Mating Systems 10
Sexual Selection and Mating Systems 11
Sexual Selection and Mating Systems 12
Sexual Selection and Mating Systems 13
Sexual Selection and Mating Systems 14
Sexual Selection and Mating Systems 15
Mating Systems and Sexual Selection 1
Mating Systems and Sexual Selection 2
Adaptation and Experimental Evolution 1
Adaptation and Experimental Evolution 2
Sexual Selection and Experimental Evolution
Adaptation and Gene Flow
Adaptation and Genomics
Adaptation and Evolutionary Ecolgy 1
Adaptation and Evolutionary Ecolgy 2
Adaptation and Evolutionary Ecolgy 3
Adaptation and Evolutionary Ecolgy 4
Adaptation and Evolutionary Ecolgy 5
Natural Selection in the Wild: from Genotype to Phenotype
Here's the list of sessions with "Random Genetic Drift" or Neutral" in their titles.
Isn't that strange? There are many more neutral alleles than beneficial alleles and random genetic drift is overwhelmingly the most common mechanism of evolution. I guess most evolutionary biologists have a huge bias toward studying adaptation. One can't help but wonder how many studies were abandoned when the investigators discovered that they couldn't prove natural selection was involved.

Why weren't those studies completed and published as examples of random genetic drift?

There are more than one thousand talks being given at this meeting and only five (5) have the word "drift" in their title—and one of those is about education!


33 comments :

Joe Felsenstein said...

I think there's a bias there. If someone chooses a piece of DNA in the middle of nowhere in the genome, well away from any genes, and then concludes that it doesn't show any sign of its evolution being anything but drift, would we see a paper on that at the meeting?

Would they not be less likely to even start the study. Sure there are most likely lots of places in the genome where nothing much is happening aside from mutation and random genetic drift. But not too many people are intensively studying them.

NickM said...

Oh come now. Probably hundreds and hundreds of those papers are using genetic drift theory -- e.g. anyone detecting signature of selection in a genome is judging against a null hypothesis of background drift. Anyone who makes an estimate of population size, demographic history, phylogeographic history, etc., is mostly assuming mutation and substitution through random drift.

So, knowing about drift helps us find out all kinds of interesting things. It may be, though, that repeating an obvious truth (most genetic mutations are approximately neutral and are spread/lost through drift) is just obvious to everyone in population genetics now, and thus isn't very interesting to say in the titles of research talks, which after all are supposed to be about new findings not old ones.

Alex SL said...

What NickM said.

Every biologist knows that genetic drift exists and accounts for what happens to the vast majority of the genome which is not under selective pressure at any given moment. We systematists usually use non-coding regions and deal daily with these issues (coalescence theory, incomplete lineage sorting, etc.).

But the interesting question that the theory of evolution set out to answer is why there is the illusion of design in nature, and the answer there is selection and not drift. All coding regions can probably be assumed to be at least under stabilizing selection all the time, as one is generally at some disadvantage with a knock-out mutation of any given protein.

Adaptation is simply the more interesting story; you could just as well ask why most of astronomy deals with stars, planets and nebulas when the vast majority of the universe is empty space.

Larry Moran said...

I agree with you that it's not very interesting to study boring bits of junk DNA. However, I believe there are many phenotypic (visible) traits that are not under selection. I find it surprising that almost every graduate student and postdoc who sets out to study one of these traits ends up discovering that it is adaptive.

Don't you think that's an extraordinary coincidence? You'd think that at least some of those traits would turn out to be nearly neutral in the population they study.

Larry Moran said...

You're saying that in the past year or so there were hardly any new finding about the fixation of traits by random genetic drift even though everyone knows that it's the dominant mechanism of evolution?

You don't find that surprising?

Do you really believe that the only "new" findings in evolutionary biology are just more examples of adaptation?

Larry Moran said...

But the interesting question that the theory of evolution set out to answer is why there is the illusion of design in nature, and the answer there is selection and not drift.

To me, much of evolution doesn't appear very well designed. Isn't anyone interested in that? The problem with the adaptationist program, as you well know, is that you start with the ASSUMPTION that a trait is "designed" and therefore must be an adaptation.

You might expect that from time to time a researcher would discover that his/her favorite trait wasn't under selection and would publish the results. That doesn't seem to happen very often. Gould & Lewontin have an explanation for that phenomenon.

All coding regions can probably be assumed to be at least under stabilizing selection all the time, as one is generally at some disadvantage with a knock-out mutation of any given protein.

That's correct.

Adaptation is simply the more interesting story; ...

Why? Aren't you interested in knowing whether some traits are non-adaptive?

anthrosciguy said...

I have in the past been an adaptationist and of course adaptation is fascinating; but I now also find the fact of non-adaptive traits extremely interesting. Then there's the puzzle of how to attempt to unentangle non-adaptative from adaptive in any given trait, and how a non-adaptive trait might mimic adaptive to the extent we want to make a story, a just-so, to explain it. There is grandeur in this view of life. Too.

Atheistoclast said...

Finally, Larry Moran understands that the Darwinists have hijacked the science of genetics and are pushing an agenda at these conferences. Their message to the Toronto professor: "Adapt or die."

Alex SL said...

Well, as a botanist I kind of have a hunch there anyway. If you look at leaves, the selective pressure is obviously (under mesic conditions) to make them flat and either small enough so that they do not tear or else make them larger but then compound for the same reason (fern and palm fronds). But I can only explain the astonishing diversity of leaf margin dentations, leaf venations and shape outlines found happily next to each other in the same forest by assuming that these characters are undergoing some mixture of developmental constraints and random walk.

Maybe you are right for some other areas of biology, but I would just say that at least systematists are way more aware of the importance of drift than you appear to assume in your posts.

Anonymous said...

Your view aside, Larry Moran has understood what he wrote in this post for ages. No "finally" about it.

Joe Felsenstein said...

... I believe there are many phenotypic (visible) traits that are not under selection.

How small does a difference in fitness have to be for the trait to not be under selection? In a population of size (say) 1,000,000 the fitness differences have to be down near 1/1,000,000 to have the effect of selection negligible. There is also the issue of how big a change in the character is. Perhaps making my forearm 1 centimeter longer would have little fitness effect (perhaps, though I doubt it). But making it 1 meter longer would have a very noticeable effect.

In any case, even when fitness differences are evolutionarily substantial, they are hard to detect. I hope to see you at some of those talks. They will be looking for evidence of fitness differences. They won't always find them. But proving that the fitness differences are small enough to be neutral is incredibly hard, so that the matter will still be in doubt even if no fitness differences are detected.

Short summary: evolution can do an experiment much bigger and much longer than we can.

Atheistoclast said...

Nick Matzke is wrong. There is no "illusion" of design in Nature. We have all been formed in our mother's wombs through a process which is controlled by a highly intelligent "agency" - for want of a better word. The degree of organization, coordination and synchronization involved during 9 months of gestation is truly astounding. Our body may not be 100% ideal or perfect, but that may because we are designed to be mortal. It is time to abandon the false belief that we are just the product of the accumulation of lots of random mutations in DNA.

Kele said...

There were two "adaptive vs. non-adaptive evolution" sessions at the SMBE 2012 meeting last week. I wish I knew what they had talked about there!

Speakers are listed in the column on the far right http://imgpublic.mci-group.com/ie/PCO/SMBE2012_Timetable_Sun24th.pdf

The Thought Criminal said...

Yeah, ideological influence in pure, pristine science. Partisan manipulation of the idealistic and surpassingly transcendent search for objective knowledge. But, don't tell the kids on the blogs, they can't handle it.

Richard Edwards said...

I think that non-adaptive changes are really interesting too. I think there is something going on beyond just ascertainment bias here. I think part of the problem is the difficulty in convincingly demonstrating that something evolved neutrally, especially if it happened in the past. The null hypothesis is usually neutrality and so you look for deviations away from neutrality. However, there are lots of reasons why deviations might not be significant, even if adaptive evolution was really happening. Failing to disprove the null hypothesis is not the same as proving the null hypothesis. Similarly, you might test a particular possible adaptive explanation experimentally (i.e. does it give an advantage) and find it lacking but that only rules out that particular adaptive explanation, not all possible adaptive explanations.

When it comes to historical changes, are we ever going to be convincingly able to say that a specific trait evolved non-adaptively? Even if there is no obvious benefit now, it may have evolved adaptively under a selective pressure that is no longer active. I'm not sure what the solution to this is. The mathematics and simulations of random genetic drift make it a sure thing. Is that not enough? (Do we have to come up with a specific explanation for every evolutionary change in history?)

john harshman said...

I see this thread is really drawing in the conspiracy theorists and other loonies. Was that your intention, Larry?

I think Joe has a big point: you really can't conduct a test for neutrality. All you can show is that the selection coefficient is above or below a detectable level. Levels below those usually detectable in the lab can easily be high enough to be effective. It may be that levels below those detectable by comparative sequence analysis can also be high enough to be effective, though I'm not sure. So all you can generally say is not that X trait is evolving neutrally but that selection is too low to detect. You can reject a null hypothesis of neutrality, which makes studies of selection possible. But you can't (at least usually) reject a null hypothesis of selection in the lab. In comparative biology, we can assume neutrality if a sequence appears to be evolving at a neutral rate; that's quite a reasonable assumption, but it's hardly a conclusive test.

Now of course I think that most of the average eukaryote genome is indeed evolving neutrally. But having said that, what's interesting about saying some particular piece is evolving neutrally? It's what we expect. What's interesting is finding a piece that isn't evolving neutrally.

Unknown said...

thanks

Richard Edwards said...

I think the point is that "non-adaptive traits" and "neutral loci" are not the same thing. It is, indeed, pretty boring to observe a particular bit of an average genome is evolving neutrally. I think what Larry is getting at is that there might be lots of heritable physical (and behavioural) traits of organisms that evolved non-adaptively. They may not even be entirely neutral but they might be sufficiently close to neutral that random drift (including founder effects etc. after population bottlenecks) fixes some of them independently of any selection pressure that might be on them. A non-adaptive trait may be the product of a number of underlying loci, which may or may not themselves be evolving neutrally. (There is also pleiotropy, of course.)

Joe Felsenstein said...

It may be that levels below those detectable by comparative sequence analysis can also be high enough to be effective, though I'm not sure.

Put that way, I'm not sure either. (So good question). However most protein coding loci will show conservation of many regions because of purifying selection. In that sense they are not neutral. But if we alter the question and ask whether the changes that do occur are neutral, slightly deleterious or favorable, then the tests by comparative sequence analysis must miss detecting many of the favorable changes, A change that is favored only at a single site, only in one particular branch of the tree, is not detected as favorable by present codon models most of the time. They are good at detecting rates of substitution that exceed neutral rates over many sites and/or over the whole tree.

Does that mean that I am saying that we know that most changes in most loci are favorable? No, we don't know that. Just that, whatever fraction of them is favorable, we will miss detecting many of those.

I might add that observing conservation does not reject the neutral theory, Kimura was quite careful about that, way back at the beginning. He included purifying selection in the theory: it asserted only that those changes that did occur were (mostly) neutral. Even in 1968 it was already known from protein sequence work that loci varied a lot in their rates of amino acid substitution -- he agreed that this was owing to different degrees of purifying selection and that was not what he was making assertions about.

Larry Moran said...

How small does a difference in fitness have to be for the trait to not be under selection? In a population of size (say) 1,000,000 the fitness differences have to be down near 1/1,000,000 to have the effect of selection negligible.

It's true that very small differences in fitness can be significant in large populations. I presume that most studies will include estimates of effective population sizes in order to bolster their case for adaptation. In many cases, especially with large animals, the effective population sizes may be less than 1000.

I hope to see you at some of those talks.

I won't be attending too many of those adaptationist talks but I may look in on a few just to see their evidence for adaptation and why they have ruled out random genetic drift. (Surely they will include estimates of effective population size and selection coefficients right?)

Larry Moran said...

Do we have to come up with a specific explanation for every evolutionary change in history?

No, I don't think we do. I think it's quite scientific to admit that we just don't know whether a trait is adaptive or not. I'd like to see a lot more papers where the authors say, "We can't rule out the possibility that this trait may are evolved by non-adaptive mechanisms."

That's why I find it so astonishing to see so many papers where the authors are absolutely convinced that they are dealing with an adaptation.

John Harshman said...

Cabbages: Same problem with phenotypic traits as with direct genetic analysis: selection too weak to detect can still be significant. So how do you decide a trait arose (or is maintained) through drift, when it might also be selection too weak to detect? How do you reject a null hypothesis of selection?

If we're talking about pleiotropy, how do you detach particular characters from the summed effect of all the characters influenced by that locus? Not easy, though I suppose you could try altering the environment experienced by one character while keeping the environment of the other character constant, at least in principle.

Larry Moran said...

Does that mean that I am saying that we know that most changes in most loci are favorable? No, we don't know that. Just that, whatever fraction of them is favorable, we will miss detecting many of those.

If you can't actually demonstrate that an allele confers a selective advantage then saying that's it's adaptive is just speculation (i.e. adaptationism). Right?

Based on everything we know today, it looks like the vast majority of substitutions in protein coding genes are nearly neutral alleles fixed by random genetic drift. Do you agree?

John Harshman said...

Oh, and the definition of "neutral" we're using here is that selection is too weak to overcome drift, i.e. Ohta's "nearly neutral theory". That's why anyone is bothering to talk about population size.

Richard Edwards said...

I agree. I guess the "we don't know" stories are not as interesting or high impact as the "we have evidence of..." stories, so they tend to get dropped faster. I am surprised that there aren't more papers on pleiotropic changes - most examples here seem to come from anatomy and the "hard-wired mistakes" of evolutionary history. Again, though, maybe it is not so much a case of people not looking as it being really hard to draw a convincing conclusion. There is a horrible ascertainment bias towards positive results across all of science, not just in the adaptionist/neutrality issue. Until scientists are no longer judged on citation metrics, I cannot see this going away (sadly).

Joe Felsenstein said...

If you can't actually demonstrate that an allele confers a selective advantage then saying that's it's adaptive is just speculation (i.e. adaptationism). Right?

Wrong! The situation is that when selection is not detected. sometimes it will actually really be there. But when it is detected, it is detected. Not infallibly, as no detection of anything in science is infallible. But detected.

Richard Edwards said...

@John. I agree that it is not at all easy to conclusively demonstrate that evolution of a particular trait is non-adaptive. This is why, I think, such studies are rarely published. My main point was just about the comment that we expect a particular locus to be evolving neutrally and therefore it is boring. I don't think this is the real issue. We don't expect phenotypes to be evolving neutrally (do we?) and therefore it would be interesting if they were. (The exception may be certain classes of molecular phenotypes - fixed amino acid substitutions are often expected to be neutral although neutrality here is often still interesting. Changes in gene expression might be neutral but I would definitely find that interesting.)

@Larry. If you can't actually demonstrate that an allele confers (conferred?) a selective advantage then saying that it is adaptive would indeed be speculation. That said, if you can't actually demonstrate that an allele conferred no selective advantage then saying that's it not adaptive is also speculation. As you said above, the scientific thing is to admit that we just don't know whether that trait is adaptive or not.

Sia said...

I’m no Biologist and am completely out of my depth here so please ignore this if it sounds too stupid but I don’t see any conflict between random genetic drift and adaptation. One provides the building blocks and the latter figures out which of the available building blocks to use and how to utilise/adapt [to] them in order to maximise their benefit. So if we are talking about evolutionary benefits, of which survival and ability to successfully reproduce are fairly high up on a grand scheme of things, then surely the main emphasis in the organisms’ reverse engineering would fall on the study of adaptation.

That is not to say that every single trait in particular behavioural ones would be adaptive but I guess it would be a lot easier to isolate these using the simple method of elimination.

john harshman said...

Nobody says there is a conflict between drift and adaptation, so you don't need to worry about that question. The argument is about the relative numbers of session titles we should devote to each of them at a meeting of evolutionary biologists. Since that's such a silly argument, it's devolved into a less silly one: the relative importance of each in explaining phenotypic evolution, and the relative ease of rejecting each if used as a null model for the other.

But I wouldn't say that drift supplies the raw materials for selection, since you bring that up. If anything, mutation supplies the raw materials. It's true that if the environment changes (including genetic background under the umbrella of "environment"), alleles that were once neutral can be adaptive. And that can be interesting. But again, the interest seems to be with adaptation.

Tam Hunt said...

Larry, I'm reading Conway Morris's interesting book, Life's Solution, and he makes the case repeatedly that the ubiquity of convergence (as he tries to demonstrate in the book) is very strong evidence for selection and adaptationism. I'm not convinced, but I'm curious what your response is to this argument?

Larry Moran said...

I agree with that part of what Conway Morris says.

Michael M said...

I don't think Larry denies that adaptation occurs or that natural selection is the primary mechanism of adaptation. He just questions the assertion that the majority of interesting features of living organisms are adaptations.

Tam Hunt said...

From what I've read so far I think Conway Morris falls firmly in the Adaptationist camp, and also firmly in the anti-contingency camp, both contra Gould. CM and Gould seem to have had a bit of a feud over the years on these key issues.