Stochastic Events and Natural Selection (Split thread)

[overtone]

Imagine a population in which 100 individuals of allele A, B and C exist (each). Now a rockslide kill 25 individuals, independent of their alleles. As the population is small we will see a change in allele frequency (drift) but it is absolutely independent of what the alleles the 25 killed individuals had. Thus we have a stochastic and not a selection event (and you cannot handwave that distinction away).

That's a selection event, by definition. The random component or aspect of it is unusually large, by your specification (normally one would suspect some kind of trait influence, like a tendency to huddle with like traits or walk closer to cliffs, but you specified otherwise). Note that your claim of the shift in allele frequency being independent of the alleles carried by the 25 killed is also nonsense - it is completely and arithmetically dependent on exactly that.

 

It is also stochastic - unrealistically so, but that's the specification. One of the mechanisms of natural selection is chance - the stochastic component of events like that one, usually lower but seldom absent. The origin of the species we see cannot be explained by an evolutionary theory that excludes the random component of selection events, and if you wish to make some kind of cutoff at some level of stochastic influence you need an argument - bald assertion that this or that level of randomness crosses the line defining "selection" from "selects alleles but not for the right reasons so we need another name" is not sufficient.

Edited November 20, 2014 by overtone

[CharonY]

 

but I'll be a Devil's Advocate and ask how was this determined that it was just random and independent in the first place?

 

As I mentioned earlier, the studies have to be on populations and expanded over larger time scale. You cannot go in immediately after a landslide look at survivors and deduce that. That would be like standing in the rain and try to use that singular information to determine where the water came from and predict whether there will be more rain over the next year.

 

Rather, looking at frequency distributions you could see that at some point there was a genetic drift in that population. You will likely never know whether it was a landslide, or something else, though (unless the area has a high propensity for such effects). The reason is that ongoing selection will eventually result in a different distribution. For example, even if after the event an allele is over represented, over time, and assuming sufficient generations have passed, a equilibrium of sorts will be found. If there was positive selection of one allele it will increase in frequency and eventually become fixed (i.e. there will be no other alleles in the population). If the population was large to begin with, the stochastic events will have little to know influence (unlike selection).

 

It is true that especially in very small populations and if sudden things happen recently (i.e. not enough time has passed to have a sort of equilibrium), it will be very tricky to figure these things out. But again, just because we have trouble seeing them, it does not mean that they do not happen. They predict different outcomes which we can observe. Just not for each and every population.

In other words, it is important to look at these aspects in a population and time-dependent manner which will smooth over singular events (unless large enough that it affects many generations).

Note that I am doing extreme simplifications. Evolutionary sciences have been looking at many things which a fine comb that I have little or no knowledge of as it is not my field of expertise. Actual research and existing models are orders of magnitude more complex than what is being discussed (and sometimes dismissed) here. And I am certain that my simplifications will not hold when we actually start looking at the complex matters (rather than trying to get some basic misunderstandings out of the way). If someone really wants to learn a bit more about what it is all about and why the stuff you learned in high school (and even sometimes basic courses in college) are not nearly enough to claim understanding, I warmly recommend Evolutionay Biology by Futuyma, in which the explanatory framework is explained nicely and accessibly (and I admit, that I should re-read it at some point, too).

Starlarvae, as I said before, these distinctions and uses are important for evolutionary research and have provided a massive amount of research. Just because the models are not fine enough to explain everything, they still yield useful predictions that can and have been tested. Claiming something else is ignoring the existing body of literature. "There is no way to know" is simply false as you can build models in which either drift dominated or selection, run it over genomic data and see where the best fit is. You will not know in detail what happened in the history of every organism, but that is not what evolutionary sciences is about. It is like trying to through out all physical models on fluid flow because they are very unreliable to predict turbulent flow. Context matters.

Overtone, I am tired of repeating myself. It appears that others got the salient points, feel free to redefine things to better fit your world view rather than learning about the world. For some that is the easier approach to get them through life.

Edited November 20, 2014 by CharonY

[overtone]

Overtone, I am tired of repeating myself.

Consider it a learning opportunity. You seem to be projecting the concepts of cause and effect into Darwinian evolutionary theory, where they mislead you. Selection by "cause" and selection by "luck" do not play different roles in the theory, and essentially no selection event in the real world is one or the other anyway.

 

Try this take: The allele neither knows nor cares what got rid of the competition.

 

 

It appears that others got the salient points, feel free to redefine things to better fit your world view rather than learning about the world. For some that is the easier approach to get them through life.

Tell you what: if I ever need a lesson in Darwinian evolutionary theory from someone who thinks Darwinian natural selection never happens by chance, you'll be the first guy I call.

Edited November 20, 2014 by overtone

[Ophiolite]

Consider it a learning opportunity. You seem to be projecting the concepts of cause and effect into Darwinian evolutionary theory, where they mislead you. Selection by "cause" and selection by "luck" do not play different roles in the theory, and essentially no selection event in the real world is one or the other anyway.

 

Try this take: The allele neither knows nor cares what got rid of the competition.

 

 

Tell you what: if I ever need a lesson in Darwinian evolutionary theory from someone who thinks Darwinian natural selection never happens by chance, you'll be the first guy I call.

Overtone, your thinking is, to use a technical term, screwed up. Selection by cause is wholly different from selection by luck. If you cannot see this, then I recommend a crash course in evolutionary biology. I can recommend several books. Charon Y's professional understanding of the concept would provide better examples.

 

I was going to try explaining again, but Charon has done a better job of that than I can and still you misunderstand. I am at a loss. Good luck.

Edited November 20, 2014 by Ophiolite

[overtone]

Selection by cause is wholly different from selection by luck.

If you are going to insist on that, I'm perfectly willing to simply bow out of the debate. It's a serious, basic, fundamental error in your conception of Darwinian evolutionary theory (and surprised as I am by uncovering its prevalence, it does clarify some past oddities of incomprehension I've been confronted with here, such as the obliviousness to GMO risk), but it has no bearing on whether Darwinian theory employment "follows the scientific method" in general, and it interrupts the thread.

 

So I will agree to let it slide, and you guys with professional understanding can take a break from educating me on this topic, especially with posting like this:

 

Imagine a population in which 100 individuals of allele A, B and C exist (each). Now a rockslide kill 25 individuals, independent of their alleles. As the population is small we will see a change in allele frequency (drift) but it is absolutely independent of what the alleles the 25 killed individuals had.

and this

 

If the shift is independent on the alleles and their properties (i.e. allele A is not more likely to be eliminated as allele B, unless by pure chance) this is non-selective, but stochastic. Genetic drift is a stochastic effect. Founder effect is more a starter position that affects the spread of alleles and both, stochastic and selection events can be involved.

and this

 

It's going to favor some alleles over others, by chance, with near certainty.

Then what follows is selection and the mechanism is not stochastic anymore.

and this is pretty depressing. Over and out.

Edited November 20, 2014 by overtone

[Robittybob1]

Consider it a learning opportunity. You seem to be projecting the concepts of cause and effect into Darwinian evolutionary theory, where they mislead you. Selection by "cause" and selection by "luck" do not play different roles in the theory, and essentially no selection event in the real world is one or the other anyway.

 

Try this take: The allele neither knows nor cares what got rid of the competition.

 

 

Tell you what: if I ever need a lesson in Darwinian evolutionary theory from someone who thinks Darwinian natural selection never happens by chance, you'll be the first guy I call.

Can you give an example where luck appears to have been involved in meaningful way in the evolution of a species?

For I tend to think luck is going to be evened out over a longer period of time. Luck isn't going to hold up all the time and in the end one with an advantage would generally be the winner. But if you have an example, I'll stand to be corrected.

[CharonY]

Of course that reasoning is just plain silly. I am going to provide a real example, but will use some reasonable but made up narratives to explain them (simply because I have no time to dig out the actual available data).

Sickle cell anemia. Physiologically this disorder and can result in negative heatlh outcomes. Thus as a whole one would expect that the involved alleles would be under negative selection. Thus the occurrence in a population should be mostly driven by mutations, for example ( there are other contributing mechanisms that could stabilize an otherwise detrimental allele, but I will ignore them for now).

In many populations this is the case, the incidence of this allele is fairly low. However, there are notable populations that have a higher frequency of that allele than others. Now there a number of possible mechanisms, but let us focus on two alternatives that have been the subject of this thread:

- drift: the populations with high sickle alleles may have been randomly started with sub-population that had a high amount of sickle allele carriers and not enough time has passed for selection to kick in and reduce the frequency by much, or:

- selection: for some reasons in these populations the sickle cell is under less negative selection than in the other populations.

 

This can then be investigated by looking at population variance (to see whether there was indeed a small starting population), migration patterns and distribution of incidence over the world. What we see is that distribution is quite wide, spans many somewhat separate populations and as whole does not fit the assumption of drift.

Thus one can speculate that some balancing selection is going on that allows higher frequencies than in other populations. Further studies have correlated incidence of high frequency populations with malaria incidence. I.e. in regions with high sickle cell incidence, there is also strong selective pressure from malaria (or vice versa).

 

This is an example where the data clearly favors selective forces rather than stochastic ones. It also (hopefully) illustrates that a) one need quite a bit of data and has to look into population rather than focusing on individual events to understand the way selection and other elements shape populations and b) that there are means to do it, but it requires more than just simple narratives. And again, I am using simplifications (but which follow existing data).

[Robittybob1]

Charon were you answering, with your sickle cell anaemia example, my post http://www.scienceforums.net/topic/85957-does-evolution-follow-the-scientific-method-if-so-how/page-6#entry838229

 

I could imagine if there were only two male members (plus females) of a species left if might be luck which of the two survive even if the better characteristics is liable to outperform his rival could well fall victim to predation.

 

But in larger populations chance (what I called "luck") would tend to even itself out, sometimes you are lucky but then sometimes not.

Whereas the general benefit of an advantage is progressively passed through the generations, unstoppable.

Edited November 20, 2014 by Robittybob1

[overtone]

Again, selection is the force which results in differential reproductive success. Measuring a change in allele frequency and comparing it to a null model is one of the ways of identifying the impacts of selection.

That won't work, according to the professional understanding given us on this forum. You have to discover and subtract the influence of stochastic events - and no, a null model does not do that automatically.

 

So this needs qualification:

 

 

Selective forces are environmental conditions which lead to differential reproductive success. Therefore, natural selection describes the process by which environmental phenomena result in differential reproductive success within a population, and thus, evolution.

Environmental phenomena which are stochastic events do not qualify as selective forces, and thus natural selection describes only a part of the process by which environmental phenomena result in differential reproductive success.

 

 

 

The comment you quote goes back to my characterizing the whiteness of bones as a trait, and I was informed that the bones are a trait and the whiteness merely an aspect of that trait.

You were not "informed" of any such thing. Review the posting. You were informed that any dictionary validated use of "trait" and "aspect" is fine with anyone here, and all such uses of either term fit into Darwinian evolutionary theory without difficulty.

 

 

 

 

But you cannot observe natural selection.

It's observed frequently, all the time, by thousands of mainstream biologists and other researchers, in the course of their research. It's quantified, often, and used to assign values to the variables in the ecological models people employ when considering - say - the nature of the effort necessary to save an endangered species or deal with a destructive invader.

 

If you want a detailed example for a layperson, there is a popular book describing many years of continuous observation of the finches on the Galapagos Islands. In this book you will find records of direct counts of survivors used to derive differential death rates correlated with incremental variations in beak size, and the influence of this natural selection on the differential reproduction of different sizes and shapes of beak.

Edited November 20, 2014 by overtone

[Arete]

That won't work, according to the professional understanding given us on this forum. You have to discover and subtract the influence of stochastic events - and no, a null model does not do that automatically.

 

The field of population genetics will be interested in how you've proven Hardy-Weinberg exact tests, Fisher's exact tests, the MacDonald-Kreitman test, Dn/Ds ratio tests, Tajima's D, the LD likelihood ratio test, etc etc, etc to be flawed.

 

Testing against a null model is how selection is identified in population genetics.

 

 

Environmental phenomena which are stochastic events do not qualify as selective forces, and thus natural selection describes only a part of the process by which environmental phenomena result in differential reproductive success.

 

Well, yes. If an event has an equal probability of eliminating all genotypes in a population, it will not cause the population to deviate from Hardy-Weinberg equilibrium - barring certain demographic changes which violate selection tests such as population subdivision. Therefore such an event will not show up in a test for selection, and is not generally considered a selective pressure. If all genotypes have an equal chance of being eliminated, none are being selected for.

 

Ergo, such an event is not considered a selective force. Changes in allele frequency caused by an event which has an equal chance of eliminating all alelles from a population is rather, by definition, stochastic rather than selective.

Edited November 21, 2014 by Arete

[CharonY]

 

Well, yes. If an event has an equal probability of eliminating all genotypes in a population, it will not cause the population to deviate from Hardy-Weinberg equilibrium - barring certain demographic changes which violate selection tests such as population subdivision. Therefore such an event will not show up in a test for selection, and is not generally considered a selective pressure. If all genotypes have an equal chance of being eliminated, none are being selected for.

 

Ergo, such an event is not considered a selective force. Changes in allele frequency caused by an event which has an equal chance of eliminating all alelles from a population is rather, by definition, stochastic rather than selective.

 

Very clear explanation, hopefully this will clear things up (considering that previous attempts obviously failed). It is sometimes amusing (and slightly worrying) how convinced people can be about their misconceptions.

[overtone]

The field of population genetics will be interested in how you've proven Hardy-Weinberg exact tests, Fisher's exact tests, the MacDonald-Kreitman test, Dn/Ds ratio tests, Tajima's D, the LD likelihood ratio test, etc etc, etc to be flawed.

I can't take the credit. My recommendation is and has been that all such changes as you defined to be "selection" be included as selection events, and that we return to those very useful tests without complications.

 

Testing against a null model is how selection is identified in population genetics.

And I think that should continue to be the custom. It always worked for me, and I never saw the need to set aside "stochastic" events and screw it up.

 

 

Well, yes. If an event has an equal probability of eliminating all genotypes in a population, it will not cause the population to deviate from Hardy-Weinberg equilibrium - barring certain demographic changes which violate selection tests such as population subdivision.

There are complications: 1) an a priori equal probability of eliminating all genotypes is not the same thing as an equal elimination of all genotypes in fact, which is quite unlikely in many circumstances. Your calculation of "selection pressure" without stochastic component - above - demands that you know the history, not merely the theoretical probability, of these events. 2) An equal probability of eliminating any given individual of a certain genotype - the usual criterion for determining "advantage" of a given genotype - is not the same as an equal probability of eliminating that genotype. 3) Almost all populations are "subdivided", in a reproductive sense, by (or in the face of) significantly effective stochastic events.

 

 

 

 

Ergo, such an event is not considered a selective force.

Yeah, I caught that. But its effects are almost certain to be, in fact, changes in allele frequencies and reproductive success, including in some cases extinctions. So you see the issue: The outcome is indistinguishable from "selection" as you defined it, without specific historical information allowing one to correctly classify the changes in allele distribution (including information adequate to extricate such effects from their role as components of the professionally "considered" selection events).

 

They're playing baseball and poker, not boxing and chess. In the case of serious but fairly rare events, one has a random walk with very large steps to handle. Do you see? Think of a poker game in which one or more players disproportionately suffer from a couple of large stochastic losses midgame - a robbery and then a fire, unconnected with the game.

 

The issue is how this affects one's formulation and theoretical handling of specifically Darwinian evolution (in particular, how one explains it to laymen). Clearly "natural selection" in that limited sense, without stochastic event, explains only some fraction of most observed evolutionary changes - even events correctly classified as "selective" normally involve large "stochastic" components, and many features of living species seem to have been established almost entirely in response to what one might call "stochastic pressure" rather than "selective pressure".

Edited November 21, 2014 by overtone

[Arete]

And I think that should continue to be the custom. It always worked for me, and I never saw the need to set aside "stochastic" events and screw it up.

 

A population exposed only to stochastic elimination of alleles IS the null model. This is what is referred to as genetic drift. Hence the distinction between selection and drift. Stochastic effects on allele frequency are generally always considered separately.

 

There are complications: 1) an a priori equal probability of eliminating all genotypes is not the same thing as an equal elimination of all genotypes in fact, which is quite unlikely in many circumstances. Your calculation of "selection pressure" without stochastic component - above - demands that you know the history, not merely the theoretical probability, of these events. 2) An equal probability of eliminating any given individual of a certain genotype - the usual criterion for determining "advantage" of a given genotype - is not the same as an equal probability of eliminating that genotype. 3) Almost all populations are "subdivided", in a reproductive sense, by (or in the face of) significantly effective stochastic events.

 

1) Well, no. Generally, one would look to see if the observed data deviated from a drift model to a statistically significant degree. A population genuinely deviating from a drift model while only experiencing drift would be a statistically anomalous type 1 error. 1a) No again, it demands that you know what a stochastic model of genetic data looks like (i.e. Hardy-Weinberg equilibrium) 2) Yes, that is what I stated - the elimination of a genotype from a population via drift is proportional to the frequency of the genotype in a population: rare alleles are more likely to be lost than common ones. 3) I don't see the relevance - or the blanket applicability of the statement; many populations are panmictic. However significant population structure within a sample or significant non-random mating will generally violate the assumptions of a selection test and potentially yield an erroneous result, sure. Often HWE tests are used to determine deviations from these assumptions, rather than to look for selection.

 

 

Yeah, I caught that. But its effects are almost certain to be, in fact, changes in allele frequencies and reproductive success, including in some cases extinctions. So you see the issue: The outcome is indistinguishable from "selection" as you defined it, without specific historical information allowing one to correctly classify the changes in allele distribution (including information adequate to extricate such effects from their role as components of the professionally "considered" selection events).

 

Drift is distinguishable from selection - in fact that is precisely what the aforementioned tests achieve. Drift can cause changes in a population and extinction events, sure. But it's distinct and definable outside of the effects of selection.

 

The issue is how this affects one's formulation and theoretical handling of Darwinian evolution. Clearly "natural selection" without stochastic event explains only some fraction of most observed evolutionary changes - even events correctly classified as "selective" normally involve large "stochastic" components, and many features of living species seem to have been established almost entirely in response to what one might call "stochastic pressure" rather than "selective pressure".

 

Yes, the other major force we are discussion here is stochastic events, which are generally defined as genetic drift. The issue I think is that it's fairly basic population genetics to statistically separate the effects of both selective and stochastic forces when assessing the evolution of a population, and to say that you can't would seem to contradict basic population genetics.

[chadn737]

I can't take the credit. My recommendation is and has been that all such changes as you defined to be "selection" be included as selection events, and that we return to those very useful tests without complications.

 

And I think that should continue to be the custom. It always worked for me, and I never saw the need to set aside "stochastic" events and screw it up.

 

 

There are complications: 1) an a priori equal probability of eliminating all genotypes is not the same thing as an equal elimination of all genotypes in fact, which is quite unlikely in many circumstances. Your calculation of "selection pressure" without stochastic component - above - demands that you know the history, not merely the theoretical probability, of these events. 2) An equal probability of eliminating any given individual of a certain genotype - the usual criterion for determining "advantage" of a given genotype - is not the same as an equal probability of eliminating that genotype. 3) Almost all populations are "subdivided", in a reproductive sense, by (or in the face of) significantly effective stochastic events.

 

 

 

 

Yeah, I caught that. But its effects are almost certain to be, in fact, changes in allele frequencies and reproductive success, including in some cases extinctions. So you see the issue: The outcome is indistinguishable from "selection" as you defined it, without specific historical information allowing one to correctly classify the changes in allele distribution (including information adequate to extricate such effects from their role as components of the professionally "considered" selection events).

 

They're playing baseball and poker, not boxing and chess. In the case of serious but fairly rare events, one has a random walk with very large steps to handle. Do you see? Think of a poker game in which one or more players disproportionately suffer from a couple of large stochastic losses midgame - a robbery and then a fire, unconnected with the game.

 

The issue is how this affects one's formulation and theoretical handling of Darwinian evolution. Clearly "natural selection" without stochastic event explains only some fraction of most observed evolutionary changes - even events correctly classified as "selective" normally involve large "stochastic" components, and many features of living species seem to have been established almost entirely in response to what one might call "stochastic pressure" rather than "selective pressure".

 

Quite frankly little of what you have argued makes sense to me....particularly the argument earlier that basically said that all tests against null distributions were wrong...

 

Especially when we know how to distinguish natural selection from genetic drift as Arete has already pointed out.

 

In the case of where you have a single large stochastic event....say a meteor hitting the Earth or a volcano going off....we can actually distinguish this from normal genetic drift and natural selection. Thats because events like this force populations into a bottleneck. There is a huge reduction in the effective population size....something that can be directly estimated and timed using both genetic and other sources of data. Such stochastic events like this have genome wide consequences that make them distinguishable from normal genetic drift and natural selection.

 

Quite frankly, I am not sure what you are arguing, unless its to say that all of population and evolutionary genetics is bunk.

[overtone]

Quite frankly little of what you have argued makes sense to me....particularly the argument earlier that basically said that all tests against null distributions were wrong...

? Where was that?

 

I agree that little of what I have posted made sense to you, if you read that in it. I'm at a loss as to how to improve, however.

 

 

 

 

1) Well, no. Generally, one would look to see if the observed data deviated from a drift model to a statistically significant degree. A population genuinely deviating from a drift model while only experiencing drift would be a statistically anomalous type 1 error.

And you would not want to make that error.

 

1a) No again, it demands that you know what a stochastic model of genetic data looks like (i.e. Hardy-Weinberg equilibrium)

You need to know what the stochastic environmental events and effects were, in order to exclude their influence on allele frequencies from your calculations of natural selection.

 

 

 

 

However significant population structure within a sample or significant non-random mating will generally violate the assumptions of a selection test and potentially yield an erroneous result, sure.

And significant stochastic environmental events will typically create such structure. So if you want to exclude their influence on allele frequency from your "natural selection", you have to find out about them.

 

 

 

A population exposed only to stochastic elimination of alleles IS the null model. This is what is referred to as genetic drift.

Which would exclude volcanic events, meteor strikes, rockslides, jokelhaups, and the other examples of stochastic environmental event specified above. So are we starting over?

Edited November 22, 2014 by overtone

[Arete]

And you would not want to make that error.

 

Hence the existence of P values, confidence intervals, false discovery rates... in fact, really, the existence of statistics to limit type I and II errors.

 

You need to know what the stochastic environmental events and effects were, in order to exclude their influence on allele frequencies from your calculations of natural selection.

 

No, you need to know what the alellic frequencies of a population are expected to look like in the absence of selection.

 

 

And significant stochastic environmental events will typically create such structure. So if you want to exclude their influence on allele frequency from your "natural selection", you have to find out about them.

 

As a theoretical technicality, - no population is in absolutely perfect HWE. However, the purpose of a selection test is to detect statistically significant deviations from the HWE (As in "We're X% sure this population is not in HWE and thus under selection") - thus a quantifiable estimation of the impact of selection on a population. To do so categorically does not require you to know about individual stochastic events.

 

Which would exclude volcanic events, meteor strikes, rockslides, jokelhaups, and the other examples of stochastic environmental event specified above. So are we starting over?

 

No, it would not - and I'm not following your reasoning as to why you came to that conclusion. A reduction in overall allelic diversity does not cause the same genetic signal as selection.

 

The only reasoning I can come up that's even close with would be that substantial population bottlenecks or founder effects can result in a heterozygote deficiency which can look like selection. However, there are specific statistical tests, e.g. http://www1.montpellier.inra.fr/CBGP/software/Bottleneck/pub.html to seperate the effect of bottlenecks from selection.

 

 

If selection is a quantifiable "force," and its unit of measure is a change in allele frequency, and the signifcance of that change is statistical (does it exceed or not some statistically defined threshold relative to the null hypothesis?), then the question remains as to cause and effect. The difference between natural selection and genetic drift rests on a statistical threshold. The assignment of causation (selection or drift) is done by convention (by defining statistical thresholds and seeing whether the thresholds are exceeded or not). Statistics is a descriptive tool, and useful for that. It can produce suggestive correlations. But causality is something else.

 

Such tools will tell you if the observed change in allele frequency statistically deviates from a null model. If observed changes in allele frequency deviate from a null model (barring some caveats) the population is, by definition, under selection at the genetic loci which are being tested. That's not a confusion of causality and correlation, that's measurement. As an analogy, if readings on a thermometer are going up over time, it's getting hotter.

 

If what you're saying is the detection of selection doesn't immediately allow one to determine the source of selection, that's quite right - and I don;t think anyone has suggested that.

 

 

It's not clear how much "environment" should be credited with the diversity of life. Early photosynthesizers dramatically reconstituted the proportions of gases that composed the atmoshphere. Bacteria nucleate raindrops and fix nitrogen in the soil. Calcium sea shells help regulate the ph of the oceans. Beaver dams.

 

And so on. Creatures construct their niches, and those of neighbors and descendants, as much as they adapt to whatever situation they're born into.

 

 

You're describing changes in environmental selection - i.e. the creation of a dam, either by landslide, beaver or human is a change in the aquatic environment which can subsequently alter selective pressures on species living in such an aquatic environment. In relation to your previous discussion of epigenetics, this is quite a confounding example.

Edited November 30, 2014 by Arete

[overtone]

No, you need to know what the alellic frequencies of a population are expected to look like in the absence of selection.

That would be "yes", not "no" - you need to know what the allele frequencies would have been in the absence of the stochastic events, which in most cases (such as those listed) would require fairly detailed knowledge of those events.

 

 

 

 

thus a quantifiable estimation of the impact of selection on a population. To do so categorically does not require you to know about individual stochastic events.

You have to know about individual stochastic events that have had selective impact - such as would be expected from the list above, and others - in order to remove their influence from your calculation of the impact of selection.

 

 

 

 

A reduction in overall allelic diversity does not cause the same genetic signal as selection.

But the differential elimination or reduction of specific alleles does. And those events listed will accomplish that, by presumption.

 

It's a semantic problem, in essence - if you don't want to call those founder and bottleneck and event effects you can test for "selection", if you don't want to call the allelic filtering often accomplished by the kinds of events listed above "selection", for example, that's OK in a sense - but the language in which you describe Darwinian evolution needs some reference to this kind of factor. The origin of the species is not describable as differential reproduction mediated by "natural selection" alone, any more, if you insist on excluding these events.

Edited December 1, 2014 by overtone

[Arete]

That would be "yes", not "no" - you need to know what the allele frequencies would have been in the absence of the stochastic events, which in most cases (such as those listed) would require fairly detailed knowledge of those events.

 

No, I meant no. By definition, the stochastic elimination of alleles from a population is random, thus over time, it is predictable, thus can be modeled quite accurately using a distribution. Each allele has a frequency in a population, and scaled to frequency, an equal chance of persisting to the next generation, or being eliminated. That does not require any specific knowledge of specific events, even in the event of fluctuating Ne.

 

 

You have to know about individual stochastic events that have had selective impact - such as would be expected from the list above, and others - in order to remove their influence from your calculation of the impact of selection.

 

An event which as a higher chance of eliminating certain alleles from a population than others is selective, not stochastic. It would rightfully show up in a selection test as selection.

 

But the differential elimination or reduction of specific alleles does. And those events listed will accomplish that, by presumption.

 

Again, an event which as a higher chance of eliminating certain alleles from a population than others is selective, not stochastic.

 

It's a semantic problem, in essence - if you don't want to call those founder and bottleneck and event effects you can test for "selection", if you don't want to call the allelic filtering often accomplished by the kinds of events listed above "selection", for example, that's OK in a sense - but the language in which you describe Darwinian evolution needs some reference to this kind of factor. The origin of the species is not describable as differential reproduction mediated by "natural selection" alone, any more, if you insist on excluding these events.

 

I believe it's a problem of basic definitions rather than semantics - a founder effect and a bottleneck describe specific effects which are certainly not selection, but affect the evolution of a population in specific ways.

 

I'm also confused by your implication that large scale, stochastic events have to be considered "selection" in order for diversification to make sense - could you give an example? At the moment, the suggestion seems nonsensical to me, but it could be a lack of comprehension, unless you're making the same mistake as starlarvae and neglecting to take into account that a large scale modification of an environment (e.g. elimination of a predator, creation or destruction of a habitat, etc) results in a change in selection pressure.

Edited December 1, 2014 by Arete

[overtone]

Again, an event which as a higher chance of eliminating certain alleles from a population than others is selective, not stochastic.

1) But this is true of most events, such as the meteor impacts and earthquake lake drainages and so forth mentioned above. How do you separate out the selective meteor impacts, say, from the stochastic ones? How do you account for the development of traits whose advantage lies in their superior resilience under regimes of random catastrophic event - randomness operating as a selection pressure itself?

 

2) So after it happened, and it did in fact eliminate certain alleles and not others, was it selective?

 

 

 

 

No, I meant no. By definition, the stochastic elimination of alleles from a population is random, thus over time, it is predictable, thus can be modeled quite accurately using a distribution

 

Well, you should have meant yes. The problem is not with your definitions, which you can make any way you want to, but with the existence of events significant in the actual origin of the actual species we have that don't seem to fit them.

 

 

neglecting to take into account that a large scale modification of an environment (e.g. elimination of a predator, creation or destruction of a habitat, etc) results in a change in selection pressure.

 

And often - by chance almost always - a change in allele frequency and distribution and population structure for that change in selection pressures to work on.

Edited December 1, 2014 by overtone

[chadn737]

1) But this is true of most events, such as the meteor impacts and earthquake lake drainages and so forth mentioned above. How do you separate out the selective meteor impacts, say, from the stochastic ones? How do you account for the development of traits whose advantage lies in their superior resilience under regimes of random catastrophic event - randomness operating as a selection pressure itself?

 

2) So after it happened, and it did in fact eliminate certain alleles and not others, was it selective?

 

 

 

 

 

Well, you should have meant yes. The problem is not with your definitions, which you can make any way you want to, but with the existence of events significant in the actual origin of the actual species we have that don't seem to fit them.

 

 

And often - by chance almost always - a change in allele frequency and distribution and population structure for that change in selection pressures to work on.

 

1) Extreme events like volcanic eruptions or asteroid impacts are not selective. They have an immediate impact of eliminating huge swaths of a population indiscriminately. The effect of this is that populations typically are forced through a bottleneck very quickly. Rare or less common alleles having less probability of surviving simply as a matter of the sampling effects, while common alleles will face nearly equal chances of being eliminated. These population bottlenecks can be dated and effective population sizes at the time of the bottleneck estimated using various methods, including those based in coalescent theory. Long term changes in climate caused by such effects will be selective, but the initial event itself will not be.

 

2) You do not need to know the selective pressure to know that a trait or allele has undergone selection. The tests for selection are unbiased in this respect and independent of whatever the selection pressure is. Its one of their strengths.

Edited December 2, 2014 by chadn737

[overtone]

1) Extreme events like volcanic eruptions or asteroid impacts are not selective.

You keep repeating that, and since you have explicitly excluded all such events from those you want to call "selective" it is true by definition. You have repeated that definition many times now. I agree that your definitions match your definitions.

 

 

 

 

Rare or less common alleles having less probability of surviving simply as a matter of the sampling effects, while common alleles will face nearly equal chances of being eliminated.

 

But typically some of the rare alleles will survive, and a couple of the common alleles will - however improbably for any given one, there are many and many relevant circumstances - be aberrantly reduced, perhaps to the point of random walk or other vulnerability to extinction. So in the event, it is not at all unlikely that a formerly rare allele will end up having replaced a formerly common one among the descendants of the remnant population or newly and temporarily isolated subpopulation, largely as a result of the event. Agreed?

 

 

 

 

2) You do not need to know the selective pressure to know that a trait or allele has undergone selection.

 

You do need to know about any stochastic events that selected new alleles, to know that a trait or allele has undergone selection - since you are excluding them.

Edited December 2, 2014 by overtone

[chadn737]

You keep repeating that, and since you have explicitly excluded all such events from those you want to call "selective" it is true by definition. You have repeated that definition many times now. I agree that your definitions match your definitions.

 

In what way is a catastrophic event selective? Natural selection is a factor of reproductive success over time. Rare catastrophic events like an asteroid impact or a massive volcanic eruption are not something that can be adapted to nor do they occur over long enough time periods to impose the sort of pressure necessary to manifest itself as differential reproductive success. The long-term changes of such events, such as climate change CAN provide that sort of selective pressure, but the initial event itself is immediate, not multigenerational.

 

You have not provided anything other than your continued assertion to the contrary coupled with a profound misunderstanding of very fundamental population genetics. The ability to trace backwards the effective population size and date bottlenecks is pretty well established in the field. Yet somehow you are going to challenge all of this?

 

But typically some of the rare alleles will survive, and a couple of the common alleles will - however improbably for any given one, there are many and many relevant circumstances - be aberrantly reduced, perhaps to the point of random walk or other vulnerability to extinction. So in the event, it is not at all unlikely that a formerly rare allele will end up having replaced a formerly common one among the descendants of the remnant population or newly and temporarily isolated subpopulation, largely as a result of the event. Agreed?

 

 

No, I don't agree. It is possible for a rare allele to replace a common allele, but for that allele to survive a catastrophic event creating a bottleneck would be by pure dumb-luck, i.e. a stochastic event. This is all a matter of very basic probability. Say you have a bag of a hundred colored balls, 48 red, 48 blue, 1 green, 1 yellow. You then blindly toss 90 balls away, what is the probability of each ball surviving? With 10% remaining, you are guaranteed to have at least either a blue or red ball remain, there is absolutely no guarantee that a green or yellow one will. In fact the odds are against picking a green or yellow ball.

 

Even if a rare allele provides some selective advantage to the catastrophic event, the mere fact that it is a RARE allele makes it improbable that it will survive. Its a pretty well established in population genetics that at small population sizes, even adaptive rare alleles stand a good chance of being wiped out simply by genetic drift. The conservation genetics literature is full of this. The best way for a rare allele to survive and spread through a population is to exist in a large population under steady or at least slowly changing conditions. Catastrophic events forcing a population through a bottleneck negate both of those factors. The population is very quickly reduced (hence a bottleneck) and conditions are changing rapidly. In such cases, common alleles have the best survival.

 

You do need to know about any stochastic events that selected new alleles, to know that a trait or allele has undergone selection - since you are excluding them.

 

No you don't. You keep asserting this, but there are many methods of looking for natural selection on an allele without knowing anything about the stochastic events. You simply don't have too. There are other measures of stochasticity beyond environment which to test against. Stochastic environmental effects will treat neutral mutations the same as mutations under selection. If an asteroid randomly falls on an animals head, it kills off both the advantageous, disadvantageous, and neutral mutations that organism carries. You test against the rate of neutral mutations to determine if a particular allele is above or below some threshold, indicating selection has been at work. Knowing the actual events causing any of these things is completely unnecessary.

[Ophiolite]

Is the difficulty here one of the technical versus the colloquial meaning of selected? Overtone, you appear to be saying - for example - an asteroid strike would selectively kill all life within a certain radius and then a large percentage of life in a wider radius and so on. If that is your intent this is not the meaning of selection in evolutionary terms. Is this the issue?

Edited December 2, 2014 by Ophiolite

[Arete]

 

 

Deviation from the null is EXACTLY what differential reproductive success is. What causes differences in reproductive success? who knows?

 

A change in temperature on a thermometer is a change in the volume of the mercury inside it... what causes the change in volume? Who knows? [/sarcasm]

 

The use of HWE deviation as a measure of the strength of selection is observationally validated e.g.

http://www.pnas.org/content/98/16/9157.short

http://library.wur.nl/WebQuery/clc/1925471

 

Much like the use of a thermometer to measure temperature. The argument that we don't actually know what causes deviations from HWE is about as relevant as arguments that we don't know for sure that Zeroth's Law explains variations in the volume of a fluid, etc.

Edited December 2, 2014 by Arete

[overtone]

No you don't. You keep asserting this, but there are many methods of looking for natural selection on an allele without knowing anything about the stochastic events.

The problem is not looking, but finding, identifying, and in the context of describing Darwinian theory (as here) labeling - reliably.

 

Changes in allele frequency are always do to one of four forces....genetic drift, natural selection, gene flow, and mutation.

About the stochastic events that happen to have altered various allele frequencies in populations - are you calling them "drift"? Because this:

 

Genetic drift and mutation are inherent in the null model which selection is tested against.

is not true of many of those events unless you have included them specifically and individually in your null model. How else are you differentiating them from selection pressure of an unknown kind?

Edited December 3, 2014 by overtone