Selection/Drift

Natural selection is a simple theory because it can be understood by anybody; to misunderstand it requires special training.

Graham Bell, The Masterpiece of Nature

Interest has been expressed in a thread on selection and drift, so I thought I’d start one, and offer my own 2-cent summary of the concepts.

Evolution, as commonly understood in biology, simply involves change in a lineage. Through mechanisms of change – principally, mutation and the insertion of ‘foreign’ DNA sequence – offspring frequently contain DNA sequences that do not derive by simple copying from their parent(s). This change is inevitable and, iterated, inexorable. There is no memory, no externally-stored blueprint for organisms; the specification of a species ‘floats on the breath of the population’, as Doctor Johnson wrote of the unwritten Gaelic language. Unless there is some kind of boundary blocking all possible avenues, a continuing source of variation is sufficient to keep a lineage exploring never-before-seen genetic space ad infinitum.

Among close ecological competitors (among, for example, the genetically similar members of a species), at a given locus in a finite world one individual ancestor’s genetic sequence is headed towards being the ancestor of every instance of that locus in a future population, and all the others with which it shared a population are heading for extinction. This derives from two facts: samples are more likely to deviate from the frequencies in the wider group than match them, and the probability of fixation of an allele is equal to its current frequency. The distortions on generational sampling tend to reinforce, through to extinction of all but one variant. This same tendency underlies the ecological principle of competitive exclusion between non-interbreeding competing species.

If a particular locus is invariant in a population, fixation has already happened. An original mutation, occurring in a single ancestor, has been passed to every member of the current population. Looking forwards, the mechanism of this concentration continues to operate, and so one particular individual from the present population will become the ancestor at that locus of all members of a future population. From any given starting point, a population of N diploid individuals will take a mean 4N generations to achieve fixation of one ancestor’s copy, and the probability for any diploid locus of being that copy is 1 in 2N. This doesn’t mean that large populations cannot fix neutral alleles, however – the number of mutations occurring scales with population size, so mutations will be fixed at the same rate they are generated, completely irrespective of population size. Doubling the population gives twice as many mutations taking twice as long to fix – the result is the same number of mutations being fixed per generation.

At the point of fixation, all instances of that locus descend from the same ancestor – they coalesce upon that ancestor. The case described – where there is no variation at all at the locus, ie there is just one allele – is the baseline process, the neutral case. If there is no variation, there is nothing for Natural Selection to ‘see’. The only process in operation is random genetic Drift – even though in this instance, it effects no evolutionary change because there are no variant alleles. The change occurred with the original mutation. This latter fact leads me to prefer the view of ‘descent with modification’ over the population geneticist’s ‘change in allele frequency’. It is true that allele frequency change is also evolution, the only part over which selection and drift have a role, but as far as each lineage is concerned, the change occured at the moment of mutation. The lineage changed at that point; the population changed somewhat later, when this mutation became the norm.

Suppose we could uniquely label the locus for every member of the population, in a heritable manner. Now, we have essentially created 2N alleles. If we allow them to operate neutrally, just as when there was no variation, evolution will now occur because allele frequencies must change in the population. Because our labelling has had no effect on the neutral ancestry-fixation process, the label itself will surf to fixation on this process, while all others become extinct.

If, instead of labelling every instance, we simply labelled one, we would find that it still had the same 1 in 2N chance of becoming fixed. And this is the situation for any neutral mutation: 1 in 2N neutral mutations will become fixed; the neutral mutation simply functions as a label.

So now, having laboured the neutral case, where all is Drift, we can look to introduce a differential between alleles. If a new allele consistently performs better or worse than the existing one – meaning that it enhances or hinders the survival and/or reproduction of its bearers – then Natural Selection has come into play. It is a simple and obvious and non-tautological!) truth that a consistent increase in survival/reproduction – in fitness – will tend to favour such alleles over the purely neutral case, and render fixation more likely and speedy, while a reduction will increase the likelihood and speed of elimination.

Unlike the purely neutral case, in which population size is cancelled out, the behaviour of selectable alleles is affected by population size. In smaller populations, random factors have a greater influence than in larger ones, and hence alleles may behave as effectively neutral despite possessing an advantage which would see them selected in a larger population.

Drift does not simply disappear when you start to turn up the selective ‘heat’. Drift essentially derives from random sampling, the tendency of subsets to deviate from the distribution of the complete set, and such sampling is in effect almost all the way along the continuum of selective advantage (apart from alleles that are so strongly detrimental that they never gain a foothold). Even a favourable allele can disappear through Drift, likewise a deleterious allele can become fixed through the same mechanism. But more often, progress will go with the expectation, not against it. The large-number tendency is for genomes to become enriched in advantageous alleles and impoverished in detrimental ones. Because this process is environmentally conditioned, it allows populations to adapt to their circumstances, by purging the traits that do worst in the recent environment.

There continues to be a debate about the relative importance of Selection and Drift in evolution generally, and in driving speciation among sexual forms. Only selection can be adaptive, because it is the only component that is responsive to the environment. But they both have significant contributions to make, and cannot readily be teased apart. Both tend to reduce the variation in a population, which variation is only restored by mutation, recombination or immigration.

243 thoughts on “Selection/Drift

  1. phoodoo:
    Allan Miller,

    Would heterozygote brown eyed people also be on a downward trend?

    Is blue eyes deleterious, neutral or beneficial?

    Welcome to the thread. There is someone called phoodoo who has asked this and he has been answered in previous comments. you could start by checking answers addressed to him.

  2. phoodoo:
    Alan Fox,

    Why would sexual selection favor a deleterious mutation you think?

    Why do you ask? Do you interpret something I said as implying sexual selection could favour a deleterious mutation? Remember the niche, phoodoo, remember the niche! Deleterious in one niche could be advantageous in another. It’s all about context.

    OM says:

    Just as an intellectual exercise, why don’t you try and answer your own question first? Then your answers can be corrected, if necessary.

    It’s a good way to learn.

    You could have a look at the role of melanin in hair, eye and skin colour. You could look at what might be the benefit of having particular levels of melanin in particular latitudes. You could ask why “gentlemen prefer blondes“.

  3. phoodoo,

    Why would sexual selection favor a deleterious mutation you think?

    Sexual selection has the power to oppose the other determinants of reproductive success – a trait favoured by females may well decrease its bearers’ survival prospects. But if that decrease is offset by increased offspring, ‘cos girls like it, then it will spread, because the net offspring accruing to carriers is greater than that accruing to non-carriers. If increased deaths are more than offset by more matings, it’s of net benefit, though we might regard the costs as indicative of a detriment.

  4. phoodoo,

    Would heterozygote brown eyed people also be on a downward trend?

    I wouldn’t get too sidetracked by references to heterozygosity. I expressed allele increase by reference to it, but need not have. The term is also used in haploid populations, which cannot have a diploid phenotype.

    In the context I was using it, heterozygosity references the probability of two single alleles drawn at random from the population differing. If you pick a diploid individual, its parents have effectively already done the drawing for you (under the assumption of random mating). Clearly, as an allele spreads, this probability decreases.

    It’s a measure of variation at population level. The role of the heterozygote phenotype is a different issue.

    Is blue eyes deleterious, neutral or beneficial?

    Dunno.

  5. Joe Felsenstein,

    If you have neutral alleles, heterozygosity declines on average.

    Yes, I try to frame my descriptions in probabilistic terms, but often slip. And clearly, when a new allele starts to spread against a fixed incumbent, heterozygosity must be on the way up. But this can only be a temporary state, without selective maintenance/mutation.

  6. Allan Miller: though we might regard the costs as indicative of a detriment.

    I think a useful analogy might be the economic effect of borrowing.

    You borrow money, and you pay interest. This means that if you buy stuff with borrowed money, you are paying more than if you saved and paid cash.

    So is borrowing beneficial (there’s a loan company called Beneficial) or detrimental (no company called Detrimental Finance)?

    It depends on the context. Paying more is always objectively undesirable, but the net effect is dependent on context. If you can achieve a net gain by investing borrowed money, then paying more is beneficial.

    In this sense, slightly detrimental variations may survive in a population long enough to pay off, as in the Lenski experiment. One could view slightly detrimental mutations as high risk investments. The effect is objectively detrimental in the short run, but could be beneficial in the long run. There is no contradiction and no paradox.

    In the economic arena, investments can fail, and one can suffer financial disaster by over-borrowing. Analogous possibilities exist in biology. A species can invest in sexually selected tail feathers that in the long run lead to extinction. Or slightly detrimental alleles can simply fade out of a population.

  7. petrushka: (there’s a loan company called Beneficial)

    I was a branch manager for them in my misspent youth (for about five minutes, honest!) and I think I can safely say the only people that benefited from their activities were the shareholders. I see it no longer exists, having been acquired by HFC who were subsequently swallowed by HSBC

  8. I use the economic metaphor, because I suspect phoodoo is more likely to understand investment and risk. Most business investments fail, and yet business continues to exist. How can that be?

    And Most known species are extinct, and yet life continues.

  9. Alan Fox: I was a branch manager for them in my misspent youth (for about five minutes, honest!) and I think I can safely say the only people that benefited from their activities were the shareholders. I see it no longer exists, having been acquired by HFC who were subsequently swallowed by HSBC

    Possibly the shareholders didn’t benefit either.

    Hindsight is sharper than foresight. Successful entrepreneurs look like geniuses in retrospect. They get fawning biographies and such. Not so much, the innovators that go bust.

    In a similar sense, extent species look designed, and their features look specified.

  10. Allan Miller: The role of the heterozygote phenotype is a different issue.

    I think phoodoo already danced around sickle-cell anaemia but so far seems not to have grasped that the likelihood of exposure to malaria is a factor in the prevalence of the sickle-cell trait.

  11. Alan Fox: I think phoodoo already danced around sickle-cell anaemia but so far seems not to have grasped that the likelihood of exposure to malaria is a factor in the prevalence of the sickle-cell trait.

    I’m sure phoodoo doesn’t want to understand that a trait can be detrimental and advantageous at the same time. Perhaps he should watch the leper scene in Life of Brian.

  12. Alan,

    I was a branch manager for them in my misspent youth (for about five minutes, honest!)…

    If you don’t mind my asking, what is/was your career? I know what almost all of the TSZ regulars do (or did), but not you. (Feel free not to answer if you’d prefer not to.)

  13. Allan Miller: a trait favoured by females may well decrease its bearers’ survival prospects. But if that decrease is offset by increased offspring

    Wouldn’t that be an adventageous mutation, since it increases fitness?

    Bear in mind that phoodoo’s question is very tricky. Alan Fox said:

    “Sexual selection could be part of the equation for eye pigmentation”

    Nobody said that blue eyes are product of a deleterious mutation (except from phoodoo), so Alan fox did not say that sexual selection could favour a deleterious mutation.

  14. petrushka: A species can invest in sexually selected tail feathers that in the long run lead to extinction

    Alan Fox: Or antlers!

    Wouldn’t this be beneficial traits? A bird with more desirable tail feathers will leave more offspring.

    Could suicidal attitude in male spiders be a better example?

  15. Guillermoe: Wouldn’t this be beneficial traits? A bird with more desirable tail feathers will leave more offspring.

    If the tail is not so long as to result in the bird being more vulnerable to, say, predation. I tentatively suggest there may be a length for peacock tails beyond which the bird would have reduced chance of surviving before getting to mate. The dynamics between attractiveness and survival in order to exploit that attractiveness might have reached some kind of evolutionary equilibrium. There is a suggestion that Irish Elks became extinct because their antlers became unmanageably large, sometimes spanning almost 12 feet.

    Could suicidal attitude in male spiders be a better example

    Yes, indeed. Also some species of female spiders allow their newly hatched offspring to eat her.

  16. Alan Fox: If the tail is not so long as to result in the bird being more vulnerable to, say, predation

    Usually this traits develop in places where predators are absent, as far as I know. So, they doesn’t really represent a danger. Anyway, if successful mating compensates the danger of being predated, I would say the trait is adventageous.

    Alan Fox: Also some species of female spiders allow their newly hatched offspring to eat her

    Wooooow!!! Did not know that…

    http://www.huffingtonpost.com/2014/02/01/spiders-eat-mom_n_4703235.html
    http://en.wikipedia.org/wiki/Amaurobius_ferox

  17. There is a general ‘trade-off’ aspect to many traits. Seeking food can be dangerous, but starving more so.

    There is a question, which phoodoo may well be limbering up to ask, as to why a female might prefer a cumbersome trait in mates. Is there a gene expressed in females governing that choice, and if so, how did it spread, given that it inhibits survival in sons? One answer is that the two are inherited together. Sons inherit their mothers’ predilections and pass them onto daughters, likewise daughters inherit their fathers’ tail genes and pass them on to sons. This has the potential to create a feedback loop, at least until counterbalancing selection takes a hold.

  18. Guillermoe: Usually this traits develop in places where predators are absent, as far as I know. So, they doesn’t really represent a danger. Anyway, if successful mating compensates the danger of being predated, I would say the trait is adventageous.

    I think this is treading close to circular reasoning.

    I think biologists like to talk about the costs of attributes and traits. We can avoid circularity by talking about costs and benefits. That’s what I tried to introduce with the economic model.

    Big brains, for example, require a disproportional amount of energy. This cost is offset by the possibility that smarter animals can find more to eat.

    But sexual selection doesn’t consider the happiness or long term welfare of individuals. It just selects those that reproduce. And in some species, that means a one time delivery of sperm, followed by death of the individual.

    So we can split the definition of fitness into multiple dimensions. Individual health is one dimension. Reproduction is another.

    Phoodoo’s error — and one made by most Idists — is inability to understand that fitness is multidimensional. The ability to live long enough to mate, and the ability to attract a mate, are not circularly defined.

  19. Allan Miller:

    One answer is that the two are inherited together. Sons inherit their mothers’ predilections and pass them onto daughters, likewise daughters inherit their fathers’ tail genes and pass them on to sons. This has the potential to create a feedback loop, at least until counterbalancing selection takes a hold.

    Alan Miller did not give any citations, but let me add two. R.A. Fisher, in his 1930 book The Genetical Theory of Natural Selection put forward this argument. It is often called Fisher’s Runaway Process. Russell Lande did one of the first theoretical analyses of this in the 1980s, using models of quantitative characters. It did work in those models.

  20. petrushka: I think biologists like to talk about the costs of attributes and traits. We can avoid circularity by talking about costs and benefits. That’s what I tried to introduce with the economic model

    Definetely. Besides, we were talking about adventageous and deleterious mutations, but mutations don’t necessarily fall in one of these categories but in the grey space between them.

    petrushka: Phoodoo’s error — and one made by most Idists — is inability to understand that fitness is multidimensional.

    And there’s also the difference between our expectation and the actual outcome. We might that perceive certain mutation would have a harmful effect, but then discover a benefit that overcompensates the harm. So, our judgement or opinions should not be the reason to consider mutations as deleterious or not.

  21. While Joe is here, I’d like to ask if there are multidimensional fitness simulations in wide use. I hear about fitness coefficients, but not about multidimensional fitness models.

  22. Alternative models (eg Zahavi, Trivers) show a different mode of sexual selection. In these, survival despite possession of the ornament is a signal of genetic quality, or ‘honesty’. These don’t tend to be greeted with quite the enthusiasm of the earlier Fisher model, and are sometimes seen as a potential secondary force, but they can apparently give sexual selection independently of the Fisher mechanism. eg Grafen 1990

  23. Another thing that makes fitness a sometimes slippery concept is that environments are heterogenous. Across a range, and through time, certain characters will be favoured more or less in different places and at different times. But in many models, these variations are flattened; we talk of s as if it were an intrinsic quality of an allele, whereas any effect it has on its own copying is entirely circumstantial.

    In any one life, there is a largely unknowable contribution of each genetic locus to the success or otherwise of that life. The same sequence could easily help at one time, hinder at another, and be of no consequence in a dozen others, and the same goes for that sequence multiplied up across many lives. We are happy to accept the intuitive correctness of the view that net ‘helping’ will cause increase, but the causality of that ‘helping/hindrance’ itself can be complex.

  24. I used to say it is impossible to design without evolution. Nothing short of an omniscient poofster could juggle all the dimensions of fitness. And do it over time, with the landscape constantly changing.

  25. petrushka:
    While Joe is here, I’d like to ask if there are multidimensional fitness simulations in wide use. I hear about fitness coefficients, but not about multidimensional fitness models.

    There can be many traits contributing to fitness, but in the end fitness is unidimensional (in simple models). There are many models, for example, where we have multiple traits, each controlled by many loci, and then a fitness function which favors an optimum multidimensional phenotype and falls away in all directions from that.

    I think maybe you were thinking of situations like overlapping generations where one has tradeoffs between how much you reproduce now and how well you survive into the next period. Such as semelparous versus iteroparous reproduction (should the salmon give its all this year or save some energy to swim back out to sea and return next year, should the plant species be annual or perennial).

  26. Mung,

    I’ll read it if you buy it for me, Mung. 🙂 Not sure why Provine is the ID-promoted authority on this subject – except, perhaps, that anyone questioning a current paradigm is bound to be right.

    eta: A summary here of Provine’s case, going deeper than the virtually content-free paragraph in Mung’s link.

  27. Provine’s beef appears to be that linkage effects mean that (in recombining populations) neutral sites cannot behave neutrally, if there is anything nearby that is under selection. However, even under selection drift is a factor – it is the departure from expectation due to sample error.

    It is true that models can be unrealistically naive, but it does not follow that the behaviour they point to is biologically irrelevant.

    The boundary of an evolutionary allele is not clearly delineated. Crossovers occur at a rate of not much more than 1 per chromosome per meiosis, so sites can remain linked for many generations. They are not evenly distributed, for several reason, so some regions will experience internal severance at greater or lesser rates than others. If, within a region of greater generational stability, a sector is under selection, all linked sectors can effectively be considered under selection as well, even if they make no contribution to fitness themselves. But does this make drift an illusion? Not if the genome is significantly non-functional, it doesn’t.

    But even in organisms with a high proportion of function, and permanent linkage, drift occurs. It is for example the causal reason why chemostats purify bacterial cultures. They don’t recombine, have permanent linkage, high population numbers and most of their sites are subject to selection – they cannot afford the luxury of nonfunctional DNA. And yet they drift. Not least, the selective effect of individual sites can be masked by clonal interference – ie, by linkage.

    If homologous recombination were induced in such organisms, what would happen to drift?

  28. Guillermoe,

    Indeed. Yet here’s Mung, and Joe G quotes him frequently too. That’s the problem with argument from authority, of course. My authority supports me on this issue. That issue? Nah, he doesn’t have a clue, the dozy evolutionist bastard.

  29. I should not comment because Will Provine sent me copies of his chapters for comment some months ago and I have not yet gotten comments back to him. I still will, because his book is self-published online and he can revise it in an ongoing fashion.

    The summary and excerpt linked to by Alan Miller do not persuade me. It’s like saying that if molecules are affected by Brownian Motion but also by gravity, then there is no such thing as Brownian Motion. Selective sweeps at nearby loci do not eliminate neutral evolution at nearby neutral sites. In effect they bottleneck the population at the neutral sites, at least if there is little recombination between these loci. Which makes genetic drift at the neutral locus all that more severe.

    Also, I’ve known Will for almost 50 years. No supporter of ID he.

  30. Allan Miller,

    The entire argument for evolution is one great big argument from authority, with zero substance to back it up. Its a bunch of college professors fighting for grant money.

  31. phoodoo:
    Allan Miller,

    The entire argument for evolution is one great big argument from authority, with zero substance to back it up.Its a bunch of college professors fighting for grant money.

    Ok. Let’s discuss ID, then.

  32. phoodoo,

    The entire argument for evolution is one great big argument from authority […]

    No it isn’t. You don’t know what argument from authority means, either.

  33. Joe Felsenstein,

    Selective sweeps at nearby loci do not eliminate neutral evolution at nearby neutral sites. In effect they bottleneck the population at the neutral sites, at least if there is little recombination between these loci.

    Also, a selectively fixed allele loses its power to ‘drag’ the linked locus with it. Once the combination is fixed, the ‘neutral attachment’ is free to drift, even if the selectable part of the combination has reduced freedom to vary. Which is, essentially, the situation for silent substitutions within a selected sequence.

Leave a Reply