The irony of natural selection

Although most mutations in the DNA that affect fitness are harmful, without mutations there would be no evolution. Evolution depends on the genetic variation created by mutation, and although there are other ways to change DNA beyond conventional mutations (horizontal gene transfer is one, though in effect it acts like a big mutation), in general evolution would pretty much come to a standstill without those random errors in the DNA. That a small fraction of the random errors increase the propagation of their gene copies (usually by improving the reproductive output of individuals carrying those good mutations) is why we have all the species and adaptations on Earth today.

The irony to which I refer in the title is that natural selection would in principle—and has in practice—actually tried to reduce the frequency of mutations to zero. But if this process were perfectly successful, natural selection would put itself out of business by totally eliminating the creation of genetic variation. We know that selection has “tried” to do this, for all the intricate mechanisms for repairing DNA damage, and excising new mutations, are products of natural selection. Those mechanisms operate not only in the “somatic” cells of the body, but also in the cells that ultimately produce sperm and eggs.

So natural selection acts on the DNA-repair level to put itself out of business. Why hasn’t it? Why are organisms still evolving? I see only four answers, one more likely than the other three.

The most probable explanation is that evolution does not produce perfect adaptations. In the case of mutations, though natural selection favors individuals most able to repair any changes in DNA (although a small percentage of these might be adaptive), this level of perfection cannot be achieved because of constraints: the cost of achieving perfection, the fact that all errors are impossible to detect or remove, or that some cells (i.e., sperm or eggs) may not even have DNA-repair mechanisms because of genetic or physiological constraints.

A less likely explanation is that the imperfection of DNA-repair mechanisms is itself an adaptation. That is, selection has acted to favor imperfect repair because such perfect repair would lead to organisms that are maladapted when the environment changes and new mutations are required to adapt. I don’t see this as likely because most mutations would still be deleterious (DNA-repair processes have no way to distinguish between useful and harmful mutations), and because this kind of selection would require frequent changes of the environment.

The third mechanism, conceptually related to the second, is that selection could favor a reduced level of DNA repair, or a higher rate of mutations, when the organism senses that the environment is changing. We know that stressed bacteria have a higher mutation rate, but that doesn’t seem to be a result of natural selection; it’s likely an epiphenomenon of stressful conditions like heat or a change in the chemistry of the substrate. But we can show theoretically that if the environment changes often enough, natural selection could favor a general increase in mutation rate because the generation of lots of bad mutations is more than counterbalanced by the few good mutations needed to survive. This is called selection for “adaptability” or “evolvability.” But there’s little evidence that a general increase in mutations under stress or changed environments is either a general phenomenon or, when present (as in bacteria) has resulted from natural selection.

Finally, there is a group-selectionist explanation. This posits that some species have indeed managed, via natural selection, to achieve near perfection in eliminating mutations, but those species went extinct because they couldn’t respond to environmental change. That would leave us with only those species having imperfect mechanisms for detecting and repairing mutations—what we see today.

I see this form of group selection as improbable, because although group selection would oppose a reduction of mutation rates toward zero, individual selection would oppose that trend. Thus, selection for imperfect DNA repair would require that the rate of group extinction or differential reproduction outpaces outweigh the rate of different reproduction of individuals that favors ever-reduced mutation rates. When group and individual selection act in different directions, as they do for traits like altruism, it requires a substantial rate of group extinction or group propagation to fix a trait; and even when if imperfect repair became the group norm, natural selection on individuals would again start driving mutation rates toward zero.

In the end, the irony of natural selection is that it tends to put itself out of business. But it hasn’t been able to, because, in my view, natural selection can’t create absolute perfection. In the case of mutations, selection isn’t able to completely weed out new errors in the DNA.

I may be wrong in these musings, or may have missed some explanations, but I’m largely unable to brain today and so am just offering this for your consideration.




  1. GBJames
    Posted December 8, 2015 at 11:38 am | Permalink


  2. Justicar
    Posted December 8, 2015 at 11:41 am | Permalink

    I’ve often thought about this as well and I’ve always tended to assume that if indeed any organisms’ progeny was born with the exact mutations required to prevent further mutations, then whilst its’ own progeny might well continue for a few more generations, ultimately it would be unable to compete in its’ environment. Whether that be against predators, prey, viral agents etc. This seems similar to your fourth option which you tag as being part of “group-selection”. I don’t see why?

    If we used a hypothetical time machine to pull an ancient population of organisms into the present (crudely simulating the effect of a species not mutating over time), would they be expected to reproduce sustainably?

    If I can make an analogy to physics, the rate of mutation is like an oscillating spring. Too slow a rate and selection pressure will produce a restoring force to increase the rate. Too fast a rate, and selection pulls it back the other way, forever oscillating around a “sweet spot” (that may itself change with the environment).

    • divalent
      Posted December 8, 2015 at 2:31 pm | Permalink

      Ya, the “environment” is never stable, so it would be impossible to become perfectly adapted (where environment is physical conditions, other species it competes against, those it preys on, those that prey on it (predators, parasites, bugs that want to eat it, etc; there really is no such thing as a “top predator”).

      Any sub-popoulation of a given species that acheived a no-mutation state would soon be extinguished by a combination of reduced ability to prey on it’s normal food-source species, and an increased vulnerability to predators, parasites, and disease, as those other species will continue to adapt to this species.

  3. daveyc
    Posted December 8, 2015 at 11:42 am | Permalink

    “Although most mutations in the DNA are harmful…”

    Is this actually the case? Would it, perhaps, be more accurate to say that most mutations that have detectable fitness consequences are harmful?

    • Posted December 8, 2015 at 11:43 am | Permalink

      Yes, that would be more accurate, as most mutations may be “neutral”. I’ll amend the above.

    • gravelinspector-Aidan
      Posted December 8, 2015 at 1:16 pm | Permalink

      Hmmm, thinking out loud here.
      There are three DNA bases per codon at 4 options so 4^3 = 64 possible states per codon. Allowing one permutation as a “STOP” leaves 63 possible amino acids that can be specified per codon. Terrestrial life uses 20-ish amino acids (the exact acids vary slightly fomr one taxon to another, with about 23 in total being used naturally). So, on average, each amino acid can be coded for in 3 different ways. Now, the statistics get a bit hairy now – more than I can do in my head, but with 4 bases to choose from, and 3 possible sequences for a particular amino acid per codon, I think that means that, on average 2/3 of random mutations in a codon would lead to no change of amino acid, and 1/3 would lead to a change of amino acid.
      Does that make sense? I recall having thoughts like this the first time I wrote the matrix of codon-> amino acid into my pocket braincell , and it’s by no means a new thought. I’ve remembered the result, but not the exact derivation.

      • Ralph
        Posted December 8, 2015 at 1:35 pm | Permalink

        I’m not sure what you’re thinking about here, but the principal reason that most mutations are neutral (no phenotypic effect) is because most DNA does not code for protein.

      • Richard Bond
        Posted December 9, 2015 at 5:51 am | Permalink

        I once worked through the entire list of codons, and calculated the probability of a change for each amino acid plus stop. Nine changes are possible: three at each base, but some of these make no difference. For example, the codons for glycine are CCA, CCG, CCT and CCC. All three changes to each of the first and second bases would change the amino acid, but a change to the third would have no effect. Thus six out of nine possible changes would constitute a mutation, or a 2/3 probability. Averaged over the 21 codons, the probability of a single base change corresponding to a mutation came out at 0.78.

        I realise that there are some symmetry assumptions there, but, those aside, I would welcome comments on my methodology by people more expert than I.

  4. Dominic
    Posted December 8, 2015 at 11:46 am | Permalink

    I find interesting – but off the cuff cannot think of examples – why some species at one time adjudged close on morphological grounds, are in fact far apart genetically. Then there is the whole realm of convergent evolution – there seem to be certain niches that demand certain ‘shapes’ and habits… alas, I have no time to muse now though!

  5. Paul D.
    Posted December 8, 2015 at 11:48 am | Permalink

    Is there an inverse correlation between the mutation rate in a species and the size of its genome (or some sort of “effective” size, taking into account junk)? If the mutation rate per base pair is driven down, perhaps this could just let the species accumulate more working genes.

    • Posted December 8, 2015 at 3:03 pm | Permalink

      Yes, if I remember correctly, mutation rates in prokaryotes are higher than in eukaryotes. Moreover, eukaryotes lack the fast and large-scale but inaccurate emergency repair mechanisms of bacteria, known as SOS response.

    • Posted December 9, 2015 at 3:52 am | Permalink

      Yes, there is, at least for errors introduced by DNA replication in single-cell organisms. There is a trade-off between replication speed and accuracy, as the proof-reading and error correction step of replication takes time. Organisms therefore evolve to minimise the error rate as far as they need to – but no more. If you are accurate enough to, on average, produce less than one error per genome per replication, there is little selection pressure to reduce it further – and quite strong selection against slowing down replication. For multicellular organisms with big genomes, things get a bit more complex – but here the likely impact of mutations is smaller due to all of the non-coding DNA and diploidy, which masks many mutations.

  6. Posted December 8, 2015 at 11:51 am | Permalink

    So “evolvability” and “non-evolvability” have both evolved! This shows how the lack of foresight for sure!

  7. ECS
    Posted December 8, 2015 at 11:56 am | Permalink

    THere are temperature sensitive mutations of the phage T4 DNA polymerase gene that have lower spontaneous mutation rates than the wild type. However, there is a cost. They copy DNA very slowly having an overactive 3′ -> 5’proof reading function that removes many correct nucleotides as well as mismatches.

  8. YF
    Posted December 8, 2015 at 11:58 am | Permalink

    It appears that A Wagner has done some work related to this issue:

    Proc Biol Sci. 2008 Jan 7;275(1630):91-100.
    Robustness and evolvability: a paradox resolved.

    Wagner A

    Understanding the relationship between robustness and evolvability is key to understand how living things can withstand mutations, while producing ample variation that leads to evolutionary innovations. Mutational robustness and evolvability, a system’s ability to produce heritable variation, harbour a paradoxical tension. On one hand, high robustness implies low production of heritable phenotypic variation. On the other hand, both experimental and computational analyses of neutral networks indicate that robustness enhances evolvability. I here resolve this tension using RNA genotypes and their secondary structure phenotypes as a study system. To resolve the tension, one must distinguish between robustness of a genotype and a phenotype. I confirm that genotype (sequence) robustness and evolvability share an antagonistic relationship. In stark contrast, phenotype (structure) robustness promotes structure evolvability. A consequence is that finite populations of sequences with a robust phenotype can access large amounts of phenotypic variation while spreading through a neutral network. Population-level processes and phenotypes rather than individual sequences are key to understand the relationship between robustness and evolvability. My observations may apply to other genetic systems where many connected genotypes produce the same phenotypes.

  9. rickflick
    Posted December 8, 2015 at 11:59 am | Permalink

    This appears to be an example of dynamic equilibrium, where a tendency toward perfect repair is opposed by the difficulty in achieving it. A significant difficulty might be the number of different kinds of encodings or the sizes of generated proteans, and other measures of complexity.
    I wonder if there is a way of testing these hypotheses? I would think that forces that resist perfection in DNA would be least in very simple organisms like bacteria and algae. So, if one could measure DNA perfection across species with different levels of complexity it might reveal a trend.

  10. ThyroidPlanet
    Posted December 8, 2015 at 12:02 pm | Permalink

    Do epigenetic things like methylation matter in evolution?

    • Posted December 8, 2015 at 12:17 pm | Permalink

      Just search for “epigenetics” on this site to see my take on it. Short answer: epigenetic modification has been important in evolution, but the kind of modification produced by genes, not solely by the environment.

      • ThyroidPlanet
        Posted December 8, 2015 at 12:49 pm | Permalink

        methylation is important in some (all?) mechanisms of DNA repair, was just wondering if there was more to add about this when talking about “imperfection of DNA-repair mechanisms” or “reduced level of DNA repair” in the two reasons given, for instance. It sounds like the answer is “no”.

    • Posted December 9, 2015 at 3:59 am | Permalink

      Ironically, despite the claims of Lamarkian evolution etc., epigenetics generally acts as a buffer to evolution by natural selection and slows it down. This is because epigenetic changes can mask the effects of mutations by changing gene expression to restore the desired endpoint. (A lot of gene regulation is buffering the environment – and the genome is part of that environment, essentially.) We see this with protein-coding changes too, which are buffered by proteins called chaperones (and chaperonins). Under stress, when these buffering mechanisms begin to fail, a lot of “cryptic variation” suddenly become “visible” to selection.

      • rickflick
        Posted December 9, 2015 at 7:56 am | Permalink

        So the buffering effect of epigenetics adaptive because it can restore function after deleterious mutation…and is that it’s only contribution?

        • Posted December 18, 2015 at 5:33 pm | Permalink

          I would not say that buffering is the “only” contribution of epigenetics. Without epigenetics, we would have no cellular differentiation – no different tissues and cell types – and thus no multicellular life as we see it. These genetically programmed epigenetic changes are clearly adaptive in many cases.

          My argument is really only against the claims of epigenetics as a source of some kind of magic “Lamarckian” evolution, in which organisms can direct their own evolution. This is nonsense in my opinion. Most environmentally-influenced epigenetic changes are there to buffer the effects of variation, not promote/accelerate it. Furthermore, if an epigenetic change is environmentally directed, it is an evolved genetic trait and should be reversed if the environment changes (hence isn’t really heritable in the true sense). If it is a *random* heritable change that has a adaptive effect in the offspring then it is not directed, so cannot really be considered Lamarckian. (There is no mechanism of getting random changes that occur in somatic cells during the life of the organism into the germline to make them heritable. Unless such a mechanism is found, the only real explanation for epigenetic changes that effect both parent AND offspring is independent *programmed* changes occurring in both soma and germline in response to some trigger – unless I am missing something.)

          One of the big problems that the “epigenetic inheritance” field has is distinguishing between epigenetic states that somehow escape the normal epigenetic resetting that occurs during reproduction (and there are a few known loci that do this but they are the minority) and states are wiped and then re-set by the same mechanisms and triggers that set them in the first place.

          Masking variation can itself affect evolution and may ultimately speed it up, as “cryptic” variation becomes “visible” to selection at “boundary conditions” where the environment is challenging enough for organisms to be generally stressed. However, the important point here is that the epigenetic changes themselves were not driving the subsequent adaptation by natural selection – this is being driven by the genetic mutations that the epigenetic changes were masking. If that makes sense?

          • rickflick
            Posted December 18, 2015 at 6:33 pm | Permalink

            OK, I think I’m with you on this. Except the wording,

            “epigenetic changes are there to buffer the effects of variation”

            I don’t think you mean “there to” in the sense of being “selected in order to…”

            This suggests teleology.

            I think you mean a “side effect of epigenetics is to…”

            Am I right?

            • Posted December 18, 2015 at 7:47 pm | Permalink

              I mean it in the sense that natural selection has resulted in evolved feedback systems that promote/maintain homeostasis via epigenetic mechanisms. So yes, these epigenetic mechanisms are selected in order to buffer the effects of variation in the system, whether said variation derives from the external environment of the organism or the internal environment of the cell, including the genetic background of that individual. A side effect of this is that some novel mutations are buffered by these mechanisms unless the feedback systems begin break down. I don’t mean “variation” in that sentence to exclusively mean “genetic variation”.

              • Wayne Tyson
                Posted December 21, 2015 at 1:00 am | Permalink

                The context (all relevant variables) for any organism is dynamic to some degree, and in the case of organisms where change is slight and slow (homeostasis), not much evolution will happen. Conversely, where the change is great and fast (chaotic), evolution (or extinction) will drive evolution.

                I can’t remember the citation, but forty or fifty years ago there was a simple but elegant study done on grasses at the edge of a mine somewhere in the British Isles. The researcher collected seed from the grasses (same species) growing on the acidic soils and planted them on the undisturbed basic soils and vice versa. On each plot, there were few survivors, but their progeny survived better (or very well–I can’t remember exactly) than the first generation. I also can’t remember how many generations were planted, but the point was made. Elementary.

  11. Posted December 8, 2015 at 12:14 pm | Permalink

    It seems likely to me that it is impossible to eliminate all mutations.

    • Kevin
      Posted December 8, 2015 at 12:51 pm | Permalink

      As long as there is environmental changes there will be likely be mutations.

      • Posted December 8, 2015 at 3:08 pm | Permalink

        The oxygen content normal for aerobes is mutagenic. Same for the normal body temperature – actually, for any temperature above Celsium -70. If you want to eliminate all mutations, put your DNA in a sealed container, deep-freeze it and get a life.
        What I regret is that during the nocturnal bottleneck our ancestors carelessly lost the photolyase gene. Would be useful for us, their diurnal descendants.

        • Kevin
          Posted December 8, 2015 at 5:28 pm | Permalink

          My guess is the second law of thermodynamics (arrow of time) is mutagenic, full stop. And even limits on quantum computers are suggested to be unavoidably environment-decoherence limited in order to corrected for errors. There may be no way to outrun entropy.

  12. eric
    Posted December 8, 2015 at 12:15 pm | Permalink

    A less likely explanation is that the imperfection of DNA-repair mechanisms is itself an adaptation. That is, selection has acted to favor imperfect repair because such perfect repair would lead to organisms that are maladapted when the environment changes and new mutations are required to adapt. I don’t see this as likely because most mutations would still be deleterious (DNA-repair processes have no way to distinguish between useful and harmful mutations), and because this kind of selection would require frequent changes of the environment.

    Isn’t this, though, the standard argument for the evolution of sex? That selection has (at least in multicelled eukaryotes) acted to favor genetic mixing over perfect duplication? Granted errors will occur in both, but from the ‘selfish gene’ perspective, sex is a form of duplication with a 50+little bit% rate of imperfection, while cloning is a form of duplication with a +little bit% rate of imperfection. If selection in general favors the organisms with the lower rates of duplication imperfection, then sex should never have caught hold.

    We know that stressed bacteria have a higher mutation rate, but that doesn’t seem to be a result of natural selection; it’s likely an epiphenomenon of stressful conditions like heat or a change in the chemistry of the substrate. But we can show theoretically that if the environment changes often enough, natural selection could favor a general increase in mutation rate because the generation of lots of bad mutations is more than counterbalanced by the few good mutations needed to survive. This is called selection for “adaptability” or “evolvability.” But there’s little evidence that a general increase in mutations under stress or changed environments is either a general phenomenon or, when present (as in bacteria) has resulted from natural selection.

    AIUI there are some organisms that switch from cloning to sexual reproduction under environmental stress. That would seem to me a pretty clear example of an ‘evolvability’ adaptation and consistent with points 2 and 3 that there are cases where less perfect duplication is evolutionarily favored over more perfect duplication. I’m not claiming its a smoking gun, but is there any better hypothesis?

    • Ralph
      Posted December 8, 2015 at 2:07 pm | Permalink

      Diploidy per se (separately from considerations of sexual recombination) has an important impact.

      The frequency of mutations, with perhaps an order of magnitude or so them, is:
      Neutral (no phenotype)
      Loss of function
      Gain of function – deleterious
      Gain of function – beneficial

      In other words, the majority of mutations have no effect; of those that do, the majority result in a simple loss of function, a protein that just doesn’t work. That often means recessive genetics – in other words, if a single good copy of the gene is present in a diploid organism, there is no phenotype. Outbred parents may both have a many recessive genetic defects, but at different loci. Provided that between them they can contribute at least one good copy of each gene, the offspring are fine.

      So, separately from considerations of DNA repair and replication fidelity, diploid organisms tolerate more mutation.

    • Posted December 8, 2015 at 3:13 pm | Permalink

      As far as I know, sex is thought to have evolved to eliminate the accumulation of harmful mutations (the Mueller’s ratchet) by mixing genes up, obtaining some combinations with fewer mutations and others with more mutations and allowing natural selection to certify the former combinations.
      Diploidity is thought to have evolved to allow recombinational DNA repair even before replication – again, to safeguard the genome against mutations.
      Someone said once, “Evolution is NOT the function of any gene or any other structure of any organism.” Maybe wrong, but I don’t know counter-examples.

  13. Posted December 8, 2015 at 12:15 pm | Permalink

    I like the first explanation. Presumably DNA repair comes at a cost, and natural selection will act to reduce mutation frequency to the point where the resulting extra benefit just compensates for the extra cost. This hypothesis should have testable implications.

    • Mark Sturtevant
      Posted December 8, 2015 at 12:32 pm | Permalink

      A cost factor is a possibility. For example, I would expect that higher fidelity repair would require a slower cell cycle. Mayhaps that would decrease fitness since it slows growth or reproduction or stem cell renewal.

      • Posted December 9, 2015 at 4:11 am | Permalink

        I’m pretty sure this is the main answer. (And people have studied it, particularly in viruses.) There also probably some DNA lesions that are simply impossible to repair accurately under certain conditions, so there is probably a non-zero minimum theoretical mutation rate, even without trade-offs.

  14. Posted December 8, 2015 at 12:21 pm | Permalink

    It’s especially interesting because these repair mechanisms are themselves the results of mutations. Hmm.

    • Posted December 8, 2015 at 2:57 pm | Permalink

      I wonder if replication is a result of mutation. I imagine life itself began because of mutation. No way to prove that I suppose.

      • Posted December 8, 2015 at 3:14 pm | Permalink

        No need to prove. It cannot have been otherwise.

  15. Ken Elliott
    Posted December 8, 2015 at 12:27 pm | Permalink

    While my gleaning of all bits of this is minimal, it is still quite fascinating. I daresay that my understanding of evolution and it’s mechanisms has increased greatly due to Professor Coyne and Professor Dawkins, but nowhere near a level that would allow me to pass a biology class on the subject.

    • Marilee Lovit
      Posted December 8, 2015 at 3:57 pm | Permalink

      I am also a beginner and have learned a lot from Jerry Coyne’s website. But I highly recommend Dr. Mohamed Noor’s on-line course “Intoduction to Genetics and Evolution.” He teaches at Duke University and the course is available through Coursera. Truly wonderful introductory class.

      • Ken Elliott
        Posted December 8, 2015 at 4:57 pm | Permalink

        Thanks, Marilee. I will keep that in mind. It’s going into my Evernote as we speak.

      • Torbjörn Larsson
        Posted December 9, 2015 at 7:13 am | Permalink

        I agree on Noor’s course! [Disclaimer: former student.]

    • HaggisForBrains
      Posted December 9, 2015 at 5:04 am | Permalink

      Agreed; fascinating post.

  16. Mark Sturtevant
    Posted December 8, 2015 at 12:29 pm | Permalink

    DNA repair may be imperfect because the current success rate of DNA repair is ‘good enough’. I know of no trait that is the product of natural selection that is perfect. The very similar process to DNA repair is DNA replication, and it too is imperfect.

    • Kevin
      Posted December 8, 2015 at 1:06 pm | Permalink

      In relation to quantum efficiencies, nothing in our warm soups of thermodynamic organic mush comes close to perfection.

      Individually, we also rather suck at passing information on to ancestors. My desktop computer stores 95% of my life that is no longer accessible to me (and I am not even half way done).

      As a species, information is preserved pretty well. This makes the species successful despite the microscopic (e.g. DNA replication) imperfections.

      • Mark Sturtevant
        Posted December 8, 2015 at 2:15 pm | Permalink

        And to cap all that off, organisms are competing with other organisms that face the same constraints on physiology and genetics and anatomy. It’s like the old evolution joke: “You cannot outrun a bear!” “No, I only have to outrun you.”

      • Gregory Kusnick
        Posted December 9, 2015 at 1:01 am | Permalink

        “Individually, we also rather suck at passing information on to ancestors.”

        Yeah, my kids never call me either.

    • Posted December 8, 2015 at 3:16 pm | Permalink

      I think DNA repair is less perfect in somatic than in germ cells, and this becomes evident in somatic cell nuclear transfer.

  17. Ken Pidcock
    Posted December 8, 2015 at 12:45 pm | Permalink

    I think there’s some agreement that increased mutation frequency can be adaptive to clonal agents, like certain viruses, that have to evolve around host defense.

    And, while it’s not quite the same thing, we have the example of somatic hypermutation in B lymphocytes. Surely there are a lot of useless cells generated by that process, but we agree it’s adaptive.

  18. Burt Simon
    Posted December 8, 2015 at 12:48 pm | Permalink

    Wouldn’t a (slightly) increased mutation rate lead to more speciation? If so, perhaps one could add “species selection” to the list. Or would this be part of choice 4?

  19. BogiT
    Posted December 8, 2015 at 12:53 pm | Permalink

    Here’s a link to a lecture by Michael Lynch:

    Among other things, he talks about evolution of mutation rate. First 30 minutes are devoted to evolution of protein structure, but the second part of the talk is about his drift-barrier hypothesis for mutation rate evolution.
    If I got it right, improvements of repair machinery eventually become so small that the power of drift outweighs the power of selection.

    • BogiT
      Posted December 8, 2015 at 12:54 pm | Permalink

      Oops, sorry! I didn’t know it was going to embed the video.

    • Posted December 9, 2015 at 2:21 am | Permalink

      Here’s a paper from Lynch and colleagues:

      Sung, W., M. S. Ackerman, et al. 2012. Drift-barrier hypothesis and mutation-rate evolution. Proceedings of the National Academy of Sciences, USA 109: 18488–18492.

      The key evidence is in Figure 1C:

      There is a negative correlation, across organisms, between the number of mutations per coding DNA per generation versus effective population size.

      The hypothesis is that natural selection reduces mutation rates. Eventually further improvements in the fidelity of replication are so weakly selected that they become impossible in the face of genetic drift.

  20. merilee
    Posted December 8, 2015 at 12:55 pm | Permalink


  21. Randy Schenck
    Posted December 8, 2015 at 12:58 pm | Permalink

    The changes that are likely tested at length would be the viruses for flu since it is almost a sure bet and happens pretty fast. Also, the resistant weeds to herbicides are probably tested.

    It is an unfair thing in evolution if the weeds can change to survive the herbicide but the butterfly cannot.

  22. Simon Hayward
    Posted December 8, 2015 at 1:20 pm | Permalink

    So my best guess – and that’s what it is, is along the lines of some of the comments above.

    First, we need a system that is efficient enough that we don’t all die of somatic malignancies before reproduction can occur.
    And second, there is a cost to being significantly more efficient than that (obviously in species where parents/grandparents might play a role in rearing offspring the curve shifts compared to a lay eggs and die approach). We have developed pretty sophisticated mechanisms for ensuring genomic integrity and killing cells with DNA damage so clearly there is considerable investment in this area.

    So my take is that it is a simple cost/benefit analysis performed by natural selection.

  23. peepuk
    Posted December 8, 2015 at 1:47 pm | Permalink

    Evolution can only create local optimums; it has no foresight.

    Evolution is extremely wasteful; 99.9999999% of all species went extinct already, for example on average mammals vanish after just 4 million years.

    Isn’t evolution just some sort of slow (chemical) warfare with billions of parallel arms races? If this is true it pays to be a moving target.


    “.. most scientists agree this is more to do with luck than some inherent superiority of modern forms.”

  24. kelskye
    Posted December 8, 2015 at 3:22 pm | Permalink

    As a layman, my take is Darwinian. If DNA optimised itself too much, it would mean organisms wouldn’t as effectively respond to changes in their environment. So organisms with the ability to change win out over organisms that are fixed in place.

    No idea how accurate this is though.

  25. Ken Kukec
    Posted December 8, 2015 at 3:47 pm | Permalink

    Could DNA-repair mechanisms ever eliminate not just transcription errors, but mutations caused by environmental factors, such as cosmic radiation?

    Has Templeton jumped on the second possible explanation, that “that the imperfection of DNA-repair mechanisms is itself an adaptation”? Seems that, with some motivated pushing and prodding, that explanation might lend itself to the type of teleological explanation the JTF might like.

  26. Frank
    Posted December 8, 2015 at 4:34 pm | Permalink

    “The third mechanism, conceptually related to the second, is that selection could favor a reduced level of DNA repair, or a higher rate of mutations, when the organism senses that the environment is changing. We know that stressed bacteria have a higher mutation rate, but that doesn’t seem to be a result of natural selection;”

    But there may be mechanism at play. Experiments with stressed, non-dividing E. coli suggest that such cells sometimes undergo gene amplification in response to stress – including interactions with copies on plasmids. This transient increase in copy number leads to a higher probability that one copy will undergo the “right” mutation, even without any change in the actual mutation rate. I believe a well-studied system is the reversion from lac- to lac+ in cells that are starved but are in the presence of lactose. The rate of the reversion mutation SEEMS to go up, but in fact it is more likely due to an amplification of copy number in these stressed cells, with no change in mutation rate. BUT, having the gene-amplification mechanism in place for times of stress could be viewed as an adaptation to increase “evolvability”.

  27. DiscoveredJoys
    Posted December 8, 2015 at 5:08 pm | Permalink

    I try(!) to avoid the teleological bias in calling things by their ‘function’ – like DNA repair. However these ‘repair processes’ appear to be more like stabilisers. They not only partially ‘repair’ favourable mutations, they also partially repair unfavourable mutations (and neutral mutations), perhaps avoiding (for a while) the population being locked into a permanent state, which cannot then adapt (using previously favourable and unfavourable mutations) to a new environment.

    The whole set of processes, mutations *and* repairs, is a delicate dance which will eventually go wrong. There’s an awful lot of death in Natural Selection, more than the amount of life. Perhaps we should think of it as Natural Winnowing?

  28. Nicolas Perrault
    Posted December 8, 2015 at 5:17 pm | Permalink

    Here is an interesting thought experiment. Would life on earth go extinct if all DNA always copied perfectly well? If yes how long would it take?

    • Diana MacPherson
      Posted December 8, 2015 at 5:31 pm | Permalink

      I would suspect that would depend on how static/dynamic the environment was.

  29. Diana MacPherson
    Posted December 8, 2015 at 5:30 pm | Permalink

    The next time someone tells me how wrong and imperfect I am (which will probably be tomorrow), I’m going to respond “yeah but without me, evolution would stop so….”

  30. Steve Gerrard
    Posted December 8, 2015 at 9:24 pm | Permalink

    Since even computers can not attain perfect error correction with digital data, I think the idea that perfect DNA replication is possible is wrong. Billions of bases, millions of individuals, and not a single missed base ever? Not going to happen. If you could show what specific changes would lead to perfect replication and repair, you might have an argument, but I don’t think that can be demonstrated.

    It is not as if there are any species that are perfectly adapted in every other way, so why not perfect replication as well. No species is perfectly adapted on any measure, so there is no reason to thing one would be with respect to DNA replication. Is there a species in which no individuals ever die early? I don’t think so. As Jeb so eloquently put it, stuff happens.

    Maybe I just arguing that nothing is ever perfect, so why expect DNA repair to be perfect. Natural selection seems to do a good job of keeping the fidelity very high. If you have eliminated almost all the sources of mutation to select from, it becomes very difficult to stumble onto a more accurate repair mechanism, so it seems like a self limiting process that can never actually go all the way.

  31. Marella
    Posted December 8, 2015 at 10:56 pm | Permalink

    In order to eliminate errors in the DNA, the copying mechanism would need to be perfect, or the repair mechanism would need to be perfect, or possibly both. Since perfection doesn’t exist, this cannot happen. The cost of perfection is too high, in energy, time and other resources to achieve that last fraction of a percentage point of accuracy that needs to be achieved to create perfection. The cost of absolute perfection is probably infinite; it is not accessible.

    Whether organisms would die out if they achieved perfect gene maintenance and reproduction doesn’t much matter, since it can’t possibly happen.

  32. Gregory Kusnick
    Posted December 9, 2015 at 1:53 am | Permalink

    I don’t quite buy the premise that natural selection seeks perfect copying fidelity and has tried to achieve it. What it seeks is optimal copying fidelity, where optimality falls somewhat short of perfection, and the error-correcting mechanisms it has built are in pursuit of that lesser goal. They’re imperfect not because higher fidelity is unattainable, but because it’s unnecessary.

    • Posted December 10, 2015 at 3:30 am | Permalink

      What you say is undoubtedly true for the B lymphocyte lineage, with their supermutation. Maybe also for some viruses. However, thinking of us with our 30,000 genes, I doubt that there can be an optimum non-maximum fidelity for germline cells. Just think of the deleterious mutations of the X chromosome taking their toll of male hemizygotes in every generation.

      • Posted December 10, 2015 at 6:31 am | Permalink

        You assume that improved fidelity comes for free. If replication rate – and thus sperm production – was reduced as fidelity increased, there would indeed be a non-zero optimum.

        Even in the absence of such a trade-off, Gregory is right that there comes a point where further improvements may be theoretically possible but “not necessary” in the sense of being too weak to have a visible fitness effect. (See Mike Lynch video elsewhere in the comments.)

      • Gregory Kusnick
        Posted December 10, 2015 at 10:56 am | Permalink

        To say that there is no optimum non-maximum is to say that there is continuing selection pressure toward maximal fidelity. That would seem to be at odds with the observation that the fidelity we see in nature appears to be stably non-maximal, which implies the absence of any net selection pressure for greater fidelity.

  33. reasonshark
    Posted December 9, 2015 at 3:59 am | Permalink

    I agree with Kusnick, since it’s worth asking for whose benefit the perfect copying exists.

    Take it right down to the selfish allele. An allele in a cartel of genes should be open to genetic mutation at other loci because one of those mutations would, sooner or later, prove to be beneficial to the gene cartel, which includes itself, giving it a competitive advantage over rival cartels.

    If the allele itself mutates, then it becomes a different gene and the stage is basically reset for it.

    The closest to a counterexample I can think of is a copy of the allele mutating into a rival that competes for space in the gene pool. Even then, it’s just the corollary of my first point, since the other genes have a collective selfish interest in promoting (and thus protecting) the superior allele. Even if the original allele has every selfish interest in preserving itself at the expense of a superior allele, it will be outnumbered by the rival alleles with every selfish interest in preserving the superior allele.

  34. Torbjörn Larsson
    Posted December 9, 2015 at 7:28 am | Permalink

    Thank you, very good for a non-brained article, and good comments too! It clarified my muddled thinking on the issue.

  35. Wayne Tyson
    Posted December 13, 2015 at 12:13 am | Permalink

    Most of this stuff is waaay outside of my “pay grade,” but “What we have here” is a tangled web, “communication-wise.”

    I will again respectfully submit to my superiors that, though the trail is neither straight nor narrow, there probably is some kind of fuzzy “boundary” around the central question made up of the rather more relevant than irrelevant elements that have more than a snowball’s chance in hell of leading toward an yearning understanding than a learning bottleneck.

    I further submit that “perfection” is a human, nay, cultural construct rather than a phenomenon that exists within Nature.

    The precise mechanics continue to elude understanding, largely because, I submit, that while advances in genetics are impressive, a corresponding context “genome” does not exist with anywhere near the same state of advancement. I continue to propose that “we” set about defining with similar precision, the elements of the context for life so that we can reap compounded benefits from the advances on the genetics “side,” driving toward an understanding of the Great Dance of Life with the context(s) in which it exists—and cannot exist, as organisms, populations, “species.”

    I credit others who have before me, touched upon this point (e.g., The Goldilocks Hypothesis, the “sweet spot,” etc.). I have proposed that a database be accumulated over time through an automatic transfer from all relevant research, and programs for analysis developed, but so far, no takers.

  36. Diane G.
    Posted December 14, 2015 at 10:01 pm | Permalink

    Great subject, great comments.

    • Wayne Tyson
      Posted December 16, 2015 at 12:27 am | Permalink

      Let’s have yours.

%d bloggers like this: