More Recent Comments

Wednesday, December 12, 2007

Is Evolution Linked to Environmental Change?

 
As a general rule, adaptationists are mostly interested in the results of natural selection and not very interested in evolution by random genetic drift. They tend to use the word "evolution" when what they really mean is "natural selection" or adaptation.

Most adaptationists see evolution as positive natural selection. They focus mostly on changes where the population is becoming more fit with respect to the environment. Some of them think that this eventually results in populations that stop evolving because they have become optimized to a particular niche. In such cases, "evolution" (i.e., positive natural selection) will only start up again if the environment changes.

Whenever I mention this I'm usually confronted with a storm of denials. Apparently none of the adaptationists who comment on Sandwalk are guilty of such fuzzy thinking.

I'm so happy for them. The idea that species have exhausted all possible adaptations and reached the very tip top of their adaptive peak seems incredibly naive. The idea that "evolution" will have halted—as opposed to adaptation—seems even more naive.

Now that I've got that off my chest, let's turn to the Hawks et al. (2007) paper that's getting so much press [Accelerated Human Evolution]. Remember, this is a paper about human evolution.

I've read the paper and I can't really comment because there's no data in the paper. What I mean is that there are no examples of the 11,439 "selective events" that they found. It would have been nice to see a few examples of their data just to get some feel for it's quality.

The paper is complicated because it consists mostly of a discussion of the data, which we can't see. The first author, John Hawks, has made an attempt to simplify the work by posting an explanation on his blog john hawks weblog.

Here's an excerpt from the article titled Why human evolution accelerated [my emphasis-LAM]. The article explains why he expected to see an big increase in evolution adaptation following the development of agriculture about 10-40,000 years ago.
Still, a very small fraction of the mutations in any given population will be advantageous. And the longer a population has existed, the more likely it will be close to its adaptive optimum -- the point at which positively selected mutations don't happen because there is no possible improvement. This is the most likely explanation for why very large species in nature don't always evolve rapidly.

Instead, it is when a new environment is imposed that natural populations respond. And when the environment changes, larger populations have an intrinsic advantage, as Fisher showed, because they have a faster potential response by new mutations.

From that standpoint, the ecological changes documented in human history and the archaeological record create an exceptional situation. Humans faced new selective pressures during the last 40,000 years, related to disease, agricultural diets, sedentism, city life, greater lifespan, and many other ecological changes. This created a need for selection.

Larger population sizes allowed the rapid response to selection -- more new adaptive mutations. Together, the the two patterns of historical change have placed humans far from an equilibrium. In that case, we expect that the pace of genetic change due to positive selection should recently have been radically higher than at other times in human evolution.
Now, if I understand this correctly, here's the scenario. About 40,000 years ago humans had pretty much stopped accumulating adaptations because they were becoming optimized to their environment. This is reflected in the data, which shows a slow rate of adaptation at that time.

Then humans started to live in larger communities as they abandoned the hunter-gatherer mode of existence for one based on farming. This created a new environment that was less fit than the previous one. The human population responded to this less fit environment by expanding rapidly in numbers. This created more opportunity for beneficial mutations that were required under the new environmental conditions. The result was a huge increase in the rate of adaptive evolution.


28 comments :

TheBrummell said...

John Hawks: This created a need for selection.

Argh! There's so much wrong with that sentence, I don't know where to begin.

Larry Moran: This created a new environment that was less fit than the previous one. The human population responded to this less fit environment by expanding rapidly in numbers.

Yeah, that part. Does Hawks read his own words?

Anonymous said...

Gene expression has a good discussion (John links to it in one of his articles):

http://www.gnxp.com/blog/2007/12/notes-on-evidence-for-acceleration.php

This is one of the key points being made and it seems quite apt:

If you find the theoretical argument convincing (as I do), then it's easy to accept their major conclusion. However, if you don't find the theoretical argument convincing, the evidence presented in this paper should not convince you

Also:

That is, the entire argument is predicated on perfectly identifying selection in the regions of the parameter space they search. This is a major assumption, and not one I'm willing to make without strong evidence. They provide none.

A couple of the authors have popped up in the comments.

John Hawks said...

Does Hawks read his own words?

Well, I do try, but I make mistakes every so often. Clearly, "This induced new selection pressures" would be a better choice of words. On the other hand, the more precise sentence is jargon to anyone not already familiar with the concepts. This creates a need for selection of better words.

I've read the paper and I can't really comment because there's no data in the paper. What I mean is that there are no examples of the 11,439 "selective events" that they found. It would have been nice to see a few examples of their data just to get some feel for it's quality.

I am actually very surprised that you aren't more familiar with the literature on natural selection in humans. I read Sandwalk all the time, and I had the impression that you had a good notion of how selection and drift both contributed to our evolution.

But now, looking back, I realize that you've only ever written about drift. I don't want to chide you for ignoring selection -- to be sure, drift needs more advocates. But I should remind your readers about the well-known examples of selection in recent human populations. Well-characterized genes under positive selection include lactase (adult milk-drinking), G6PD (falciparum malaria), Fy (vivax malaria), CCR5 (HIV now, plausibly smallpox in the past), D4 dopamine receptor (ADHD), SLC24A5 (light pigmentation), OCA2 (pigmentation, eye color), ATP-binding cassette C11 (dry earwax), microcephalin (brain development), CYP3A5 (salt retention), NKX2-2 (insulin regulation), ADH (alcohol metabolism), and PSEN1 (Alzheimer's susceptibility).

Our assessment of selection has found many additional regions -- in fact thousands of loci -- that show patterns of positive selection equivalent to these well-characterized examples.

There are a number of free resources where our list of genes, and others, are available for open access examination. The UCSC genome browser includes a track with our LDD statistic. Additionally, the Pritchard lab's Haplotter tool is available, giving genome-wide results from a different statistic.

Personally, I see very little point to this show-and-tell: we tested a hypothesis of evolutionary change, and the essential data include the age distribution of positively selected alleles, which we present. Our method is very sensitive but reasonably conservative, and its results are consistent with those from weaker methods, which also have identified a thousand or more recently selected alleles. We have a better test for positive selection, I think, but we cannot take the credit for showing that it is widespread in the human genome.

Now, you may say -- "sure, that's positive selection, but what about other mechanisms of evolution?" Or for that matter, what about other measures of evolution -- for instance, substitution rate versus per generation variance in allele frequencies?

I understand completely the strong desire to give genetic drift its due. We can make very clear predictions about drift in recent human populations. The substitution rate of neutral alleles should have remained precisely the same (population size being irrelevant to substitution rate). That seems a less important measure on the time scale of the last 10,000 years, since the mean fixation time of a new neutral mutation in the Neolithic would be on the order of 10 million generations. Let's be clear: increasing population size has little effect on neutral changes, except for vastly increasing the transit time for the small fraction that reach fixation. This increases the neutral diversity in large populations. We may have seen some of that in recent people, but I would hesitate to call it an "acceleration" of neutral diversity.

So, it would seem that you have a valid complaint against me -- I'm saying that "evolution" accelerated, when I should be saying that "this itty-bitty, teeny-weeny part of human evolution accelerated."

Except, well, if half of our genes are streaking from zero to fixation, that's a whole lot bigger impact on our "evolution" than the very slight changes in frequency of neutral alleles. And those neutral alleles aren't safe: when half of human genes are sweeping toward fixation, they are taking a whole lot of neutral hitchhikers along for the ride. It seems pretty obvious that the pace of neutral evolution has accelerated to almost the same degree as positive selection, merely through the hitchhiking effect.

Which leaves phenotypic evolution. Did it accelerate? Well, that's a harder question to quantify. For one thing, the "rate problem" is an old topic in paleontology, and no good resolution of it was ever found. Different time spans have different apparent rates, merely because directional selection cannot continue indefinitely on any character.

Still, there are many human phenotypes that clearly must have accelerated, because they weren't changing before the Neolithic, and they underwent rapid increase of a new state afterward. For instance, skin pigmentation evolution clearly accelerated in Europeans and East Asians during the last 20,000 years, because several genes that have strong effects lightening skin color have originated and increased under positive selection only after that time.

I can understand that many people are unaware of the anthropological literature on recent phenotypic evolution. Body size reduced markedly after the appearance of agriculture, followed by even more pronounced reductions in brain size and tooth size. Head shape changed markedly, with skulls becoming relatively broader. All these changes occurred in a global context -- essentially in every population that adopted agricultural and sedentary lifestyles. And they happened at a much faster rate than at any time in the previous million years. It is clear that many skeletal phenotypes have accelerated in the magnitude of change per generation, and we may hypothesize that this was caused by the underlying acceleration in positive selection across the genome.

I really want to thank you for taking the time to read the paper and think about what it means. Just as I've been writing this comment (which I'll cross-post to my blog), I've reminded myself that neutral evolution must in fact be accelerating right along with new positively selected alleles. But folks reading this should themselves a favor and read a bit more about human history!

John Hawks said...

Now, if I understand this correctly, here's the scenario. About 40,000 years ago humans had pretty much stopped accumulating adaptations because they were becoming optimized to their environment. This is reflected in the data, which shows a slow rate of adaptation at that time.

I should point out that the "slow" rate of new positively selected mutations before 40,000 years ago was still more than 10 times the Haldane limit -- the human-chimpanzee comparison looks like one adaptive substitution per 20 generations, compared to one per 300 generations under the Haldane limit. There's nothing slow about human evolution.

Then humans started to live in larger communities as they abandoned the hunter-gatherer mode of existence for one based on farming. This created a new environment that was less fit than the previous one. The human population responded to this less fit environment by expanding rapidly in numbers.

It looks like you may be pointing to an apparent paradox -- if the environment was worse, then how come the population grew?

But positive selection occurs because of differences in relative fitness, not absolute fitness. In demographic terms the average number of offspring per individual went up. But the variance in the number of offspring also went up. Assuming this, the next part of your paragraph follows directly:

This created more opportunity for beneficial mutations that were required under the new environmental conditions. The result was a huge increase in the rate of adaptive evolution.

Amanda said...

I'm really new to all this stuff, and I find this post very interesting. However, I don't really understand it. I do understand what Hawks is saying(I think): in a nutshell, the rate of evolution in humans was slowed as they became optimized for their environment. But then the environment changed (switch to agricultural lifestyle)and human "evolution" sped up again.

What I don't understand is the concept of genetic drift. If someone could explain it to me and how it could account for the differences in rates of evolution, in layman's terms, that'd be great. Of course, if not, I'll look it up, but it'd be helpful to hear about it in the context of this post.

Thanks so much!

Larry Moran said...

John Hawks says,

But now, looking back, I realize that you've only ever written about drift. I don't want to chide you for ignoring selection -- to be sure, drift needs more advocates. But I should remind your readers about the well-known examples of selection in recent human populations. Well-characterized genes under positive selection include lactase ...

I written often about adaptation and natural selection but, you're right, I tend to write more about scientists who mis-interpret data because they are biased toward an adaptationist explanation.

Scientists with that bias tend to interpret every positive bit of evidence as showing "well-characterized" support for their preferred worldview. I haven't had a chance to research every one of your examples to see if all scientists agree with you or if there is some dissent. However, I am aware of dissenting views about some of your examples.

More importantly, I'm aware of the problems with your techniques. In fact I blogged about some of these problems a few months ago [What Is the Cause of Genetic Differences in Domesticated Rice Varieties?].

As you know, it is sometimes very difficult to tell the differences between selective sweeps and effects due to random genetic drift, especially the bottleneck effect. The human species is subdivided into many complex populations, as you noted in your paper, and making an accurate model of evolution in these populations is not straightforward.

No doubt you have high confidence in your assumptions. You believe, for example, that your data reveals selective sweeps in the Yuroba group and not random genetic drift. You probably believe that all of your "selective events" in the Japanese group (2783) are valid as are all of the 2367 mostly different "selective events" in the Chinese group.

You may be correct but it does seem strange that 11,000 different genes are being selected in the four groups. That's half of all the genes in our genome. Not only that, your calculated selection coefficients are not very impressive (2 or 3%).

Have you thought about the implications? About 5,000 of our genes are required for basic metabolic processes and structure. These include genes for ribosomal proteins, glycolytic enzymes, muscle fibers, and spliceosomes. I'd like to know if any of these genes are part of the set of genes undergoing "selection" in your study. Unfortunately, I don't have access to your data so I can't tell. If you're picking up regions where there are only very highly conserved genes then I'd be skeptical about your conclusions.

I understand completely the strong desire to give genetic drift its due. We can make very clear predictions about drift in recent human populations. The substitution rate of neutral alleles should have remained precisely the same (population size being irrelevant to substitution rate). That seems a less important measure on the time scale of the last 10,000 years, since the mean fixation time of a new neutral mutation in the Neolithic would be on the order of 10 million generations.

You mention in your paper that the are 40,000 amino acid substitution differences between chimps and humans. This works out to about 330 substitutions every 100,000 years. It's reasonable to conclude that most of these are neutral.

Contrast this with your claim that there are about 3000 alleles being fixed by natural selection in each of your populations, and all of this is happening in the past 40,000 years. It would only take about six events like that in the past 6 million years to account for all of the amino acid substitutions in the human lineage. And every one of them would be a selected allele.

But what we see are no more than 10,000 amino acid substitutions that could be attributed to natural selection (assuming a generous 50% of all known aa substitutions). This means that the events in the past 40,000 years overwhelm all adaptations that have occurred since chimps and humans last shared a common ancestor.

This is a bit hard to swallow.

-DG said...

Amanda:

Genetic Drift/Allelic Drift is a purely random stochastic process that is the result of sampling errors. If you take a population of say diploid individuals then you have two alleles at every gene locus. At neutral loci these alleles are randomly picked (essentially) during each successive generation, which over a sufficient length of time will cause an allele variant to be either fixed or lost in the population. Wikipedia has a decent primer article.

Anonymous said...

I have a question for Dr. Hawks:

What do you believe the implications of your paper are for putative cognitive and temperamental differences between human populations?

The paper mentions adaptations in such traits since the late Pleistocene. You noted that much adaptive genetic evolution has been with genes related to brain function, corresponding to phenotypic evolution in the brain. Cochran has earlier speculated that this is related to functional changes in the brain.
('Human evolution, radically reappraised', World Science, 3/26/07): "growth in more ad­vanced brain ar­eas might have made up for the shrinkage, Coch­ran said; he spec­u­lated that an al­most break­neck ev­o­lu­tion of high­er fore­heads in some peo­ples may re­flect this.")

The paper also notes that human populations maintained different population sizes, adopted agriculture at different times, and had different rates of adaptive evolution. One implication of the research, according to Harpending, is that human populations are diverging.

John Hawks said...

But what we see are no more than 10,000 amino acid substitutions that could be attributed to natural selection (assuming a generous 50% of all known aa substitutions). This means that the events in the past 40,000 years overwhelm all adaptations that have occurred since chimps and humans last shared a common ancestor.

This is a bit hard to swallow.

Some of the alleles overlap between populations -- we are looking at more like a total of 8000 than 11000.

Otherwise you are quite correct. The current estimates suggest that at most a third of the amino acid substitutions between chimps and humans were selected. That goes down from 40000 to 13000. We can add to this some number of non-coding selected substitutions. At present, we essentially do not know this number but it probably is not a larger number than the selected amino acid substitutions. So let's say at most 30,000 selected substitutions between chimpanzees and humans.

That means that the last 40,000 years have seen as many new adaptive alleles as the previous 4 million, assuming a constant rate across the earlier time span. I wouldn't make that assumption; things may have been changing faster in Homo than in Australopithecus, for instance.

I agree that this looks on the surface like a surprising number, but why should it be? The fundamental dietary, disease, and behavioral changes of the last 40,000 years -- and in particular the last 10,000 are much larger than the Australopithecus-Homo transition, or any equivalent episode in human evolution. I see no need to posit that this kind of event had ever happened earlier in human evolution -- after all, if the hominid population had suddenly ballooned to 50,000,000 or more in the Pliocene, we would have noticed!

If we count the number of people we estimate to have lived in the last 40,000 years, it is within an order of magnitude of the number who lived in the previous 4 million years. To the extent that positive selection has involved rare mutations, the acceleration ought to be around the extent that we are estimating from data.

A. Vargas said...

I hope Hawks is not thinking "this may be as good as the evidence gets", as he does with the proposal that some microcephalin genes we have inherited from sporadic ancient copulations with...neanderthals.

Truth is, many genes that supposedly are highly positively selected thereafter fail when tested for an effect o the phenotype; take microcephalin. Having or not a "positively selected" alelle does not have any evident effect on the brain or intellectual capacities.

These "selection detection" studies may truly go intellectually bankrupt if they fail to correlate in a satisfactory way with phenotypic effects that conferr selective advantages. Concrete biological stories.

They MUST trascend mere detection of selection which is, after all, an inference, a calculation on paper that may well be missing something important. Is saying "selection was important" all that we care about? Sometimes it seems so.

I myself remain skeptical that positive selection is what is going on. I think the idea that culture "stopped" human evolution though popular has never been entertained by most smart evolutionary biologists; on the contray, the notion is that culture, by DECREASING selective pressures, may well have allowed several otherwise lethal mutations to become admissable. LESS selective pressure, more genetic variation: MORE evolution.
And then of course, biogeographic and demographic expansion alone (without selection) MUST have a great effect in explaining the increase in genetic variation.

As we see more and more of these studies, if there continues to be a the great disproportion between the actual evidence (calculated selection of genes) and the speculative inferences drawn form them (high recent competition among humans, hybrid crosses with neanderthals, chimps, etc) , we SHOULD expect some level of discredit to develop (but then, adaptationism is..."immortal")

A. Vargas said...

take microcephali again. No clear phenotypic effect, and the supposed "selected" alleles are the populations outside of africaa. Looks just like a foundation effect for coming out of africa. What a coincidence, huh?

Yet despite the absence of concrete phenotypic effects, we are required to believe that
some selective process on microcephalin has being going on only outside of africa....

Joe Pickrell said...

Well-characterized genes under positive selection include lactase (adult milk-drinking), G6PD (falciparum malaria), Fy (vivax malaria), CCR5 (HIV now, plausibly smallpox in the past)

john, your paper cites sabeti et al. for evidence that ccr5 is under selection, but they actually argue against selection at this locus.

"While such results can not rule out the possibility that some selection may have occurred at C-C chemokine receptor 5 (CCR5), they imply that the pattern of genetic variation seen atCCR5-Δ32 is consistent with neutral evolution."

there are plenty of clear examples of selection without that one.

Anonymous said...

John,

The fundamental dietary, disease, and behavioral changes of the last 40,000 years -- and in particular the last 10,000 are much larger than the Australopithecus-Homo transition, or any equivalent episode in human evolution.

Does this kind of raise the spectre of the Upper Palaeolithic Revolution? Something key going on around 40ka?

Larry Moran said...

John Hawks asks,

I agree that this looks on the surface like a surprising number, but why should it be? The fundamental dietary, disease, and behavioral changes of the last 40,000 years -- and in particular the last 10,000 are much larger than the Australopithecus-Homo transition, or any equivalent episode in human evolution.

I find this difficult to accept. We have encountered hunter-gatherer groups existing in modern times. Presumably they never went through the environmental changes that you postulate and they never experienced the huge population increase that accompanied urbanization.

What you're saying is that the difference between these small modern hunter-gatherer groups and the average commuter living in Tokyo is greater than genetic difference between the modern hunter-gatherers and chimpanzees.

You aren't surprised that some of us are a wee bit skeptical of such extraordinary claims?

If we count the number of people we estimate to have lived in the last 40,000 years, it is within an order of magnitude of the number who lived in the previous 4 million years. To the extent that positive selection has involved rare mutations, the acceleration ought to be around the extent that we are estimating from data.

This brings up another issue that puzzles me. Perhaps you can clear it up?

Your paper discusses "selection events" in four different populations of humans. One of them is the Yoruba of Nigeria who currently number about 30 million individuals.

As you point out in your paper, there are 3486 selection events in this subgroup of which only 509 are common to other groups.

What that means is that you've detected about 3000 alleles undergoing fixation by natural selection in a population of 30 million individuals. Your model suggests that this group has expanded about 100-fold in the past 5000 years and that means there were about 300,000 Yorubas back in the time when the mutations arose.

That number compares quite favorably with the estimated size of the population of all humans over a period of 6 million years, don't you think?

So, it seems to me that all this talk about a huge population of several billion people is a red herring since you are actually looking at small, genetically isolated, subsets of that population that don't have anywhere near the number of mutations you suggest.

Am I missing something?

TheBrummell said...

I apologize for my earlier snarkiness: sorry. Thank you, Dr. Hawks, for appearing here and writing some interesting clarifications. I shall have to take the time to read through this thread more carefully than my brief skimming right now. Interesting stuff.

A. Vargas said...

I guess my complaint is that I think evidence for rapid evolution can be confused with evidence for adaptationist stories. Sure, we label "faster than neutral" as "positive selection" but is it really evidence that adaptation to the environment is improving?

John Hawks said...

john, your paper cites sabeti et al. for evidence that ccr5 is under selection, but they actually argue against selection at this locus.

We cited both sides of that argument to be fair; I think the evidence for selection is compelling, since Sabeti et al. clearly used a bogus demographic model in their simulations.

That is one problem with demographic simulations; they are not run with relevant archaeological or historical estimates. We often see papers simulating population size as 10000 up to the present, which has me wondering who built the pyramids.

The problem with this is that evidence for selection is often explained away as "consistent with demographic expansion or bottlenecks." But the only pattern of demographic expansion that can create these LD blocks is rapid expansion from a very small number of founders, isolated for a long, long time. This kind of demographic event is very unrealistic in human history -- not the expansion, but the long, long isolation of a small group of founders. It's as if the ship that brought porphyria to South Africa had sailed through the Bermuda Triangle for 50,000 years before landing at Cape Town. If you think I'm exaggerating, look at the demographic simulations used by Currat et al. to argue against selection on ASPM.

What that means is that you've detected about 3000 alleles undergoing fixation by natural selection in a population of 30 million individuals. Your model suggests that this group has expanded about 100-fold in the past 5000 years and that means there were about 300,000 Yorubas back in the time when the mutations arose.

This sounds like an indulgence in this kind of Bermuda Triangle-like demography. What exactly would account for the complete isolation over 5000 years of these 300,000 Yoruba from the rest of the subsaharan African population? Clearly, from an anthropological perspective it is ridiculous -- neighboring populations are not very genetically different from each other, because they have exchanged a large fraction of genes over their histories. And selected alleles are the most likely to spread rapidly from one population to another.

What you're saying is that the difference between these small modern hunter-gatherer groups and the average commuter living in Tokyo is greater than genetic difference between the modern hunter-gatherers and chimpanzees.

Not at all; I am saying that with respect to selected (not neutral) alleles, they are as different as chimpanzees and bonobos, or Western and Eastern gorillas, or Bornean and Sumatran orangutans.

Now, since their crania are more different from each other than chimpanzees are from bonobos, and their skin pigmentation is more different from each other than Eastern and Western gorillas, and their hair form is more different than Bornean and Sumatran orangs, and their diabetes risk, cardiovascular disease risk, osteoporosis risk, all these differ substantially between these human groups ... I don't see why this should be surprising.

John Hawks said...

Let me just finish by saying this, because the thread is long, although I would be happy to come back.

Our theory assumes

1. The number of new adaptive mutations in a population is 2Nu, where u is unknown to us and possibly variable.

2. The fixation probability is 2s (where s varies among selected mutations).

From this it follows that the average product of these numbers, 4Nus, is the adaptive substitution rate. This value clearly increases linearly with N, u, and s.

N has increased exponentially in humans over the last 40,000 years, by three orders of magnitude or more.

s has likely increased also, based on the idea that humans were hunter-gatherers for a long, long time and agriculture, sedentism, and disease took the population further from a fitness optimum.

That leaves u, which we still don't know. If u was always large, then every possible adaptive mutation would have happened many times in a small population, so that demographic growth would make little difference.

Some mutations may be like that, but many (maybe most) are rare -- they don't happen that often. Certainly many of the malaria adaptations are rare: even though they are strongly selected they don't tend to happen in multiple populations. So for these mutations we can infer that u has been very small.

For the set of mutations with small u -- the rare mutations -- the acceleration effect of population growth should have been very powerful -- up to the three orders of magnitude that population size grew. Only if s and/or u decreased by a combined three orders of magnitude could acceleration have been avoided.

If our analyses of the empirical data are consistent with theory, I don't think we should be surprised by them. I think we should take our theories seriously.

Joe Pickrell said...

We often see papers simulating population size as 10000 up to the present, which has me wondering who built the pyramids.

john, that's Ne. the recent population growth in humans has left relatively little effect on neutral genetic variation, because it's not that much coalescent time. Ne-~10000 is a good approximation when simulating data from modern humans.

see here:
"Late expansion to a large size, roughly coincident with the advent of agriculture, was included for the sampled populations in all calibrated models, but had little effect on results"

Larry Moran said...

John Hawks asks,

This sounds like an indulgence in this kind of Bermuda Triangle-like demography. What exactly would account for the complete isolation over 5000 years of these 300,000 Yoruba from the rest of the subsaharan African population?

I don't know enough about African populations to answer this question.

However, I do know that there are a great many Caucasian groups that have been somewhat isolated for that period of time. Note that the operative work here is "somewhat" not "complete." You don't need complete isolation in order to recognize different races or tribes.

In any case, I think I can discern your answer. You think that the dataset for the Yoruba stands in for the entire sub-Saharan black population just as you think that the dataset for 90 white people from Utah stands in for all Caucasians.

How do explain the differences between your Chinese and Japanese groups? Each of them have thousands of "selection events" that are unique to the group. This suggests that these two groups are evolving independently during the time frame of your analysis (40,000 years).

Is this also an examples of "Bermuda Triangle-like demography"?

Your response leads to a solid prediction. You predict that looking at 90 individuals from, say Mozambique, would reveal the same 3000 selection events that you saw with the Yoruba dataset, right?

Larry Moran said...

John Hawks says,

Our theory assumes

Yes, we know all this. It's standard textbook population genetics.

1. The number of new adaptive mutations in a population is 2Nu, where u is unknown to us and possibly variable.

Of course it's variable. It depends on a lot of things. A mutation may be beneficial under some conditions and not under others.

We would love to know, on average, how many mutations are adaptive.

N is the effective population size "Ne". This is defined as "the size of an idealized population that would have the same effect of random sampling on allele frequencies as that of the actual population" (Graur and Li, 200).

Ne depends on many things including fluctuations in the size of the population over time (i.e. bottlenecks) and who is mating with who.

The average value of Ne for humans over the past 2 million years is about 10,000.

2. The fixation probability is 2s (where s varies among selected mutations).

Agreed.

From this it follows that the average product of these numbers, 4Nus, is the adaptive substitution rate. This value clearly increases linearly with N, u, and s.

Assuming that you meant "Ne", there's nobody who disputes this value. The debate is over the value of the effective population size at the time mutations arose (40,000 years ago) and what that size is today. (Because the fixation time is getting longer as the population grows.)

The other debate is over the value of "u" - the rate of adaptive mutations. You have concluded that this rate increased considerably when farming and urbanization began.

John Hawks said...

The other debate is over the value of "u" - the rate of adaptive mutations. You have concluded that this rate increased considerably when farming and urbanization began.

This is where I think your confusion may be. We assume that the per-genome rate of adaptive mutations, u, is constant.

It may have increased, indeed I think this is most likely, but I have no information to show that it did. We predict acceleration only on the basis of increased population size. It will be true as long as u and/or s have not decreased by the same amount as populations have increased.

A. Vargas said...

So we detect a "positively selected" gene. We may find no phenotype at all related to its variation, and later realize it was not positive selection after all, but drift (my prediction with microcephalin)

OR, we may find some phenotypic change (for instance, sickle cells) with some "adaptive value" (malaria resistance). Despite the fact that selection for these genes is described as "strong" this does not mean they're taking over the population. Sickle-cell disease is still a quite small minority even in malaria zones. No peppered moth- like thing here. Just higher frequency than expected for a neutral allele: That's all.

So, if we find many such genes that show "strong" positive selection, does this mean that recent human evolution is, in essence, a story of adaptation by natural selection?

To begin with, human evolution is clearly being spearheaded by non-genetic, cultural evolution. If we can detect any improved fitness between us and early agricultors, to assume these improvements are the result of natural selection of genes seems to obviate quite obvious improvements of fitness that must have been attained as cultural evolution itself became more sophisticated.

Anonymous said...

A minor quibble:
"This created a new environment that was less fit than the previous one. The human population responded to this less fit environment by expanding rapidly in numbers."

I've always understood fitness to be a property of individuals relative to other individuals within the context of their shared environment and not a property of the environment.

John Hawks said...

Your response leads to a solid prediction. You predict that looking at 90 individuals from, say Mozambique, would reveal the same 3000 selection events that you saw with the Yoruba dataset, right?

I expect that earlier things have dispersed more widely -- many of them along with the Bantu dispersal some 2000-4000 years ago. The most recent selected alleles have not had time to disperse widely. Some things have purely local advantages: there are fewer genes under selection for malaria resistance in Mozambique than in Nigeria, and none of them have a 30 percent fitness advantage.

So I predict that older events will be more likely shared. This is true of the Africa-Europe comparison, for example; the shared events are older. This is the opposite prediction from the hypothesis that recent bottlenecks or drift account for many of the apparent events, because the ones most likely to be consistent with drift are those that we infer to be older, because the LD decay is greater.


History was complicated and our model of population growth is simple. But the direction of our error is equally simple: we will miss many events, we will miss many more recent events than ancient ones (because the ancient ones should have dispersed farther), and we will miss many locally advantageous events. The Chinese and Japanese have significantly more shared events than the other population pairs, but also remember that Japan was initially settled more than 20,000 years ago. Despite the later gene flow, again many recent selected alleles have not had time to become common in both populations.

John Hawks said...

To begin with, human evolution is clearly being spearheaded by non-genetic, cultural evolution. If we can detect any improved fitness between us and early agricultors, to assume these improvements are the result of natural selection of genes seems to obviate quite obvious improvements of fitness that must have been attained as cultural evolution itself became more sophisticated.

I think Sanders is an anthropologist. Although, instead of "spearheaded", I would say that ecological changes (including ones induced by culture) must be present before any selected response could have happened.

But rather than view this as a one-way relation, I would point out that selection tends to maximize the intrinsic growth rate of the population. As new alleles entered the population and increased in frequency, they also forced cultural responses and changes of various kinds, if not by direct behavioral effects then certainly through population growth.

A. Vargas said...

You don't need to be an anthropologist (I'm not, I'm an evolutionary biologist)to realize that further cultural development, which is crucial to fitness and adaptation, is simply being ignored.
Under right environmental AND cultural condtions, we can have increases of fitness ("prosperity" if you will( that most definitely are NOT increases due to gene selection.

Are biologists as a class supposed to gloss over the obvious influence of culture? That seems pretty silly to me.

Anonymous said...

Both the CEU and YRI dataset consists of 30 trios, i.e. father-mother-child. If you want to characterize LD patterns in the respective population, you normally use only the 60 founder individuals from each of the samples.