Thursday, March 12, 2009

Examples of Accelerated Human Evolution

Gregory Cochran and Henry Harpending claim that human evolution has accelerated in the last 10,000 years. In one sense this has to be correct since the number of humans is increasing exponentially and that means far more mutations are occurring every generation. Many of those new mutations are contributing to a significant increase in variation.

But that's not what they mean. They claim that adaptations have increased. When they talk about accelerated human evolution they are mostly talking about an increase in natural selection.

For those of you who have not read the book, I thought I'd give you some of the examples that feature prominently in the opening chapter.
... when humans hunted big game 100,000 years ago, they relied on close-in attacks with thrusting spears. Such attacks were highly dangerous and physically taxing, so in those days, hunters had to be heavily muscled and have thick bones. That kind of body had its disadvantages—if nothing else, it required more food—but on the whole, it was the best solution in that situation. But new weapons like the atlatl (a spearthrower) and the bow effectively stored muscle-generated energy, which meant that hunters could kill big game without big biceps and robust skeletons. Once that happened, lightly built people, who were better runners and did not need as much food, became competitively superior. A heavy build was yesterday's solution: expensive, but no longer necessary. (p. 3)

With the invention of nets and harpoons, fish became a more important part of the diet in many parts of the world., and metabolic changes that better suited humans to that diet were favored. (p. 4)

Close-fitting clothing provided better protection against cold, allowing people to venture farther north. In cool areas, people needed fewer physiological defenses against low temperatures, while in the newly settled colder regions they needed more such defenses, such as shorter arms and legs, higher basal metabolism, and smaller noses. (p. 4)

With the advent of new methods of food preparation, such as the use of fire for cooking, teeth began to shrink, and they continued to do so over many generations. Pottery, which allowed storage of liquid foods, accelerated that shrinkage. (p. 4)

As the complexity of human speech approached modern levels, there must have been selection for changes in hearing (both changes in the ear and in how the brain processes sounds) that allowed better discrimination of speech sounds. Think of the potential advantages in being just a bit better at deciphering a hard-to-understand verbal message than other people: Eavesdropping can be a life-or-death affair. (p. 4)

... we believe that the obvious difference between racial groups are linked to gene variants that have recently increased in fitness and had major fitness effects. Blue eyes, found only in Europeans and their near neighbors, are a result of a new version of the DNA that controls the expression of OCA2 that has undergone strong selection, at least in Europe. (p. 18)

Dry earwax is common in China and Korea, rare in Europe, unknown in Africa: The gene variant underlying dry earwax is the product of strong recent selection. (p. 18)

We can confidently predict that many (perhaps most) as yet unexplained racial differences are also the product of recent selection. For example, we argue that the epicanthic eyelid found in the populations of northern Asia is most likely the product of strong and recent selection. (p. 18)


18 comments :

  1. The just so stories are pretty frustrating, as is the cover -- surely they are not implying that upright walking evolved <100Kya?

    ReplyDelete
  2. In one sense this has to be correct since the number of humans is increasing exponentially and that means far more mutations are occurring every generation. Many of those new mutations are contributing to a significant increase in variation.

    I have to correct this, since some readers may assume that you are describing population genetics accurately.

    Genetic variation from new mutations increases incredibly slowly in an exponentially growing population. A larger population does have a higher equilibrium expectation of genetic variation, but it takes on the order of N generations to get there. Because we have been exponentially increasing for fewer than 2000 generations, genetic variation in living humans has not "significantly increased" compared to the long-term average. We still have the same level of variation as expected for a Wright-Fisher population of 10,000 individuals.

    There is one way that new mutations could rapidly contribute to variation: if they were positively selected.

    ReplyDelete
  3. There is one way that new mutations could rapidly contribute to variation: if they were positively selected.

    A second idea... repeated population bottlenecks (plagues, migration, war) over the last 10,000 years, and differential reproductive success for non-genetic traits ("wealth", royalty, on the winning side of the war) allowed rare alleles to become more common after the bottleneck, then fix by genetic drift.

    ReplyDelete
  4. A second idea... repeated population bottlenecks (plagues, migration, war) over the last 10,000 years

    This is a common misperception, which geneticists have done too little to correct.

    Let's take one well-known plague: the Black Death. At one point, it killed up to a third of the European population. The population size rebounded within four generations. That's simply not long enough to have a noticeable effect on gene frequencies, even in the presence of strong differential reproduction (as you suggest).

    When geneticists talk about bottlenecks capable of influencing gene frequencies, they're talking about events that shrink population by a factor of ten or more, that last hundreds or thousands of generations.

    ReplyDelete
  5. That all depends on who's modeling you believe as to whether the bottleneck needs to be extremely drastic or last for 100's to 1000's of generations. For instance, from MARUYAMA & FUERST. Genetics 111: 691-703 (1985).

    "Allelic diversity is most affected by the initial sampling that occurs when the population is reduced; the effects of additional sampling in the generations following the bottleneck are less important."

    The black plague is a graphic example, but by no means the only bottleneck, nor is it typical of all population bottlenecks.

    "Selection only" thinking is difficult to reconcile with the population-specific disease alleles that we are reminded of in the population genetics textbooks. (Tay-Sachs, diastrophic dystrophy in Finland, etc). These disease alleles illustrate how human populations can move deleterious alleles to relatively high frequencies.

    ReplyDelete
  6. These disease alleles illustrate how human populations can move deleterious alleles to relatively high frequencies.

    So diseases kill people, and these alleles mysteriously increase in frequency because of the bottleneck?

    I can think of a more parsimonious explanation....

    For instance, from MARUYAMA & FUERST. Genetics 111: 691-703 (1985).

    Here's a quote from p. 676 of that article:

    "Our studies focus on a small time period, of the order of N or 2N generations, where N may be 50-100, or even less"

    If you'd like to think of some bottlenecks in the last 10,000 years where human populations have been dropped to 50 to 100 individuals or less, we can talk more. Not even Iceland is there. New World is not impossible, but it's earlier. Polynesia -- there you have a lot of gene flow after initial founding populations.

    ReplyDelete
  7. So diseases kill people, and these alleles mysteriously increase in frequency because of the bottleneck?

    After re-reading your comment, I realized that you meant "disease alleles" to mean genetic disorder-causing alleles, not "disease-resistance alleles". Sorry.

    Tay-Sachs is selection. There was no substantial bottleneck in early Ashkenazim.

    Diastrophic dysplasia may be drift; I wouldn't rule selection out. But its current frequency is 0.8 percent. It doesn't contribute substantially to variation even in Finland.

    Variegate porphyria in South Africa is up to around 0.2 percent. These are about as extreme founder effects as you're going to find in the history of a present-day large population. They're just not that important to genetic diversity.

    ReplyDelete
  8. Tay-Sachs is selection. There was no substantial bottleneck in early Ashkenazim.

    I'm sorry, but you will have to describe to me how a 4-base insertion that disrupts a reading frame in an allele, leading to a disease when homozygous, is fixing through selection. A biochemical explanation would be most helpful. Complicating this explanation will be fitting in how the other 90 or so different disease alleles also arose, and are also being maintained by selection.

    Closer to home for me, this sounds a lot to me like mitochondrial polymerase gamma mutations (check out pubmed ID 9010300). They lead to a dysfunctional enzyme (the story is little more clear on the pol-gamma story, IMHO, so that's why I point it out... the alleles often deplete or disrupt mitochondrial DNA through impaired replication). For yet unknown reasons, there are a lot of diseases alleles (mutational hotspot?). To start invoking a story to say breaking the enzyme is good, and positively selected, seems like an immense stretch.

    ReplyDelete
  9. John Hawks says,

    I have to correct this, since some readers may assume that you are describing population genetics accurately.

    Genetic variation from new mutations increases incredibly slowly in an exponentially growing population. A larger population does have a higher equilibrium expectation of genetic variation, but it takes on the order of N generations to get there. Because we have been exponentially increasing for fewer than 2000 generations, genetic variation in living humans has not "significantly increased" compared to the long-term average. We still have the same level of variation as expected for a Wright-Fisher population of 10,000 individuals.

    There is one way that new mutations could rapidly contribute to variation: if they were positively selected.


    Every single human has about 130 new mutations that were not present in either of their parents. When the population was about 10,000 individuals that meant that there were roughly 1.3 million new alleles being created each generation. Most of these were lost. Assuming, for the sake of argument, that they were all neutral, there were 1.3 million being fixed every generation.

    Today there are 7 billion humans and we produce 910,000,000,000 new mutations every generation. We aren't fixing that many each generation but we're sure doing better than 1.3 million. More importantly, since the present population was, until recently, subdivided into thousands of demes consisting of 10,000 or less, there were all kinds of variants persisting, and becoming fixed, in different subpopulations.

    This is why you can use DNA tests to determine where your ancestors were born. You could have done that 100,000 years ago but there would only have been a few dozen locations. Today there are hundreds.

    There is much more variation in the human species today that there ever has been. And the vast majority of those variants have nothing to do with selection.

    ReplyDelete
  10. I'm sorry, but you will have to describe to me how a 4-base insertion that disrupts a reading frame in an allele, leading to a disease when homozygous, is fixing through selection.

    Same way sickle cell did. Not fixation. Overdominance.

    If Ashkenazim had been bottlenecked, then it would have affected more than a handful of genes. But their genome-wide SNP variation is the same as non-Ashkenazi Europeans. Besides that, a recessive lethal should have been declining constantly in frequency during the last 1200 or so years, when we have historical records of large Ashkenazi populations.

    ReplyDelete
  11. Assuming, for the sake of argument, that they were all neutral, there were 1.3 million being fixed every generation.

    Larry, thanks for giving me the chance to clarify some of these issues in population genetics.

    I'll call the mutation rate per diploid genome u -- which is around 130 per generation for point mutations. The effective size is N. The number of mutations per population per generation is Nu, as you say.

    The fixation probability of each new mutation is 1/2N.

    Hence, the number originating in each generation that will ultimately fix is u/2. This is called the substitution rate, and it is independent of population size for neutral mutations.

    For this reason, there were approximately 130 neutral mutations fixed each generation in the past -- that's four orders of magnitude less than your estimate.

    This algebra underlies the molecular clock hypothesis.

    Today there are 7 billion humans and we produce 910,000,000,000 new mutations every generation. We aren't fixing that many each generation but we're sure doing better than 1.3 million.

    On expectation we will fix 130 per generation, once the population comes to equilibrium. At present, the fixation rate is rather lower because the population is expanding.

    That only applied to neutral mutations. Selected mutations may be fixing at a much higher rate at present. That's because out of those 910 billion new mutations you point out, even if a small fraction are advantageous, they may still be many more than the measly 130 that will fix under genetic drift.

    And they can fix much faster. A neutral mutation will take around 2N generations to fix. Assuming the present effective size is a third the census size, we'll be waiting an average of


    More importantly, since the present population was, until recently, subdivided into thousands of demes consisting of 10,000 or less, there were all kinds of variants persisting, and becoming fixed, in different subpopulations.

    Fst is around 0.1 in humans, which means (among other things) that around 10 percent more variants persisted than expected in the same size population under panmixia.

    Different variants that became fixed in different subpopulations would show up as loci with very high Fst. Since this is almost impossible by drift under the actual human population history, high Fst is a strong indication of selection.


    This is why you can use DNA tests to determine where your ancestors were born.

    DNA tests that determine where your ancestors were born belong to two general classes. The first use non-recombining loci, in which we can observe hundreds of linked variable sites. The most powerful is the mtDNA, which has a mutation rate more than an order of magnitude higher than the autosomes.

    The second class uses genes with very high Fst -- so called "ancestry informative markers". These are mostly genes like Duffy (Fy), which have a history of strong regional selection making different parts of the world different.

    Neither variety of test is particularly aided by the present large population size. That's why we can regularly apply similar tests to other mammalian species that number in the thousands.

    You could have done that 100,000 years ago but there would only have been a few dozen locations. Today there are hundreds.

    Taking the per-site per-generation mutation rate to be m and the length of a locus to be l, the average difference between two gene copies 200,000 years ago (before any sign of recent expansion) would have been around 20000lm. Today, the average difference will be around 22000lm. If lm were on the order of 1/1000, there would be dozens of mutations separating the average two people in the past, and baker's dozens now.


    There is much more variation in the human species today that there ever has been. And the vast majority of those variants have nothing to do with selection.

    There are more neutral alleles now in the human population than ever before. Almost all of these new alleles have frequencies less than 0.00001. All but a tiny fraction of them become extinct in fewer than 8 generations. Only 1/2N of them will become fixed.

    Measures of genetic variation such as the average difference between two individuals, the heterozygosity, the number of segregating sites in a sample of 1000 individuals, should be slightly larger but close to the same values 200,000 years ago. The number of segregating sites in the entire species should have increased more markedly than the statistics that we can measure, but again most of those sites have frequencies less than 0.00001.

    The major exception includes mutations that have a fitness advantage. In which case, the chance of fixation goes to 2s. A substantial number of such mutations that happened in the last 40,000 years have proceeded up to frequencies over 20 percent. They account for most of the high Fst SNPs, and by means of hitchhiking, for most changes in the frequencies of neutral polymorphisms.

    ReplyDelete
  12. Assuming the present effective size is a third the census size, we'll be waiting an average of

    Sorry about that. An average of over 100 billion years.

    ReplyDelete
  13. For yet unknown reasons, there are a lot of diseases alleles (mutational hotspot?). To start invoking a story to say breaking the enzyme is good, and positively selected, seems like an immense stretch.

    Recurrent mutation does not require bottlenecks to get a disorder up to a high collective frequency. Neurofibromatosis is another example.

    ReplyDelete
  14. John Hawks says,

    Hence, the number originating in each generation that will ultimately fix is u/2. This is called the substitution rate, and it is independent of population size for neutral mutations.

    For this reason, there were approximately 130 neutral mutations fixed each generation in the past -- that's four orders of magnitude less than your estimate.


    You are correct and I was completely wrong. Anyone reading this comment thread should ignore most of what I said in my earlier comment about fixation. I don't have an excuse for getting it wrong.

    There are about 130 mutations fixed per generation regardless of effective population size.

    Thus, in the ancient past when there were only 10,000 humans there were 130 neutral alleles fixed per generation. Today, with 7 billion people, if that is the effective population size then, at equilibrium, there will still only be 130 neutral mutations fixed per generation.

    However, if the species is currently subdivided into many different populations then each one could be fixing 130 mutations per generation. I don't think there's any doubt that our species is subdivided. That's why different races have different characteristics. There was very little gene flow between them until recently.

    When John says,

    We still have the same level of variation as expected for a Wright-Fisher population of 10,000 individuals.

    That statement flies in the face of common sense. Many of the variants we see today simply didn't exist 50,000 years ago. If we traveled back in time to visits our ancestors I doubt very much whether any of them would have red hair, blue eyes, or any of the defining characteristics of today's Asians.

    ReplyDelete
  15. That statement flies in the face of common sense. Many of the variants we see today simply didn't exist 50,000 years ago. If we traveled back in time to visits our ancestors I doubt very much whether any of them would have red hair, blue eyes, or any of the defining characteristics of today's Asians.

    Yes. Isn't it amazing? It's almost as if some force had pushed these functional variations to high frequencies!

    ReplyDelete
  16. However, if the species is currently subdivided into many different populations then each one could be fixing 130 mutations per generation. I don't think there's any doubt that our species is subdivided. That's why different races have different characteristics. There was very little gene flow between them until recently.

    It takes very little gene flow to maintain our very low Fst -- only 2 migrants per population per generation. The heterozygosity in human subpopulations is reduced only 10 percent from the panmictic expectation. Besides that, the fixation time for new mutations is on the order of 2N.

    ReplyDelete
  17. Regarding Tay-Sachs and overdominance, is there any evidence of this? Or is this a post-hoc explanation to fit your hypothesis. I've done some digging and cannot find the evidence of of a heterozygote advantage with any of the HEXA alleles.

    ReplyDelete
  18. Well, this has been an interesting exchange.

    ReplyDelete