Commenter nails the problem with neutral theory of evolution

_{Denyse O'Leary

May 12, 2014

Evolution, News}

Share: Facebook; Twitter; LinkedIn; Flipboard; Print; Email

Comment of the day (re Darwin’s natural selection acting on random mutation vs. neutral evolution/genetic drift:

The real problem is not that they are “mutually exclusive”. Obviously, they are not.

The problem is that they are different things, and that there is a repeated shift from one to the other when we ask which of them is responsible for functional information in biology.

The traditional view, defended also by Matzke and by you, is that NS is responsible for that. In that case, neutral evolution is irrelevamt for biological function.

Others, like Moran and Nei, seem to suggest an important role for neutral evolution in generating function. Even if they do not exclude the importance of NS.

Shifting from one model to the other is a smart way to elude analysis. It’s certainly easier to analyze and falsify a well defined model, rather than a slipping one.

The simple truth is that neither can explain functional information, but for different reasons.

Follow UD News at Twitter!

Comments

Jehu: That paper about cortisol receptor "evolution" is not only "weak". It is no refutation of anything at all, least of all of Behe's irreducible complexity concept. It is indeed a very good example of a good research paper which says interesting things, which are then inflated, in the conclusions, to the level of simple propaganda, completely unjustified by the facts in the paper. First of all, it has nothing to do with Behe's irreducible complexity. Behe's concept is about complex machines whose function is determined by the interaction of different complex parts, different molecules, different proteins, which are not in themselves functional out of the "meta-machine" of which they are part. Nothing of that is investigated in the paper. The paper is about the supposed derivation of GRs form an ancestor which was mainly sensitive to aldosterone (or rather to DOC or some other similar molecule, because aldosterone did not exist yet). The reasoning is fine, and the paper is well made. But it has nothing to do with Behe's irreducible complexity. Here we only have a protein which changes its ligand affinity (but always in the same group of ligands) by mutations at the active site. So, is it an example of a complex functional transition? Not at all. The transition is certainly functional (GRs are certainly used in specific pathways), but it is not complex. Guess how many mutations are necessary to transform the ancestor receptor into the GR? Two mutations. Two AAs.
One combination—replacement of Ser106 with Pro (S106P) and Leu111 with Gln (L111Q) (numbered by position in AncCR-LBD)—conferred a GR-like phenotype: The receptor_s median effective concentration (EC50) for aldosterone increased by three orders of magnitude, but moderate cortisol and DOC sensitivity were retained (Fig. 4C). None of the other mutants showed this pattern (table S4). Structural studies of the human GR have shown that these two residues change the architecture of the ligand-binding pocket and alter contacts with steroid in ways that exclude aldosterone and facilitate cortisol activation (18, 25). Our data thus indicate that the aldosterone specificity of MR has a simple and conserved mechanistic basis—two crucial replacements in the GRs that wiped out ancestral sensitivity to aldosterone.
Like chloroquine resistance. Like nylonase. With an important difference. While in the case of chloroquine resistance, and probably of nylonase, the mutation itself is enough to give a new function, in the case of GRs the scenario is not so simple. It is true that we have a new receptor, which has a different, specific affinity with a specific ligand (cortisol). But that means only that we have a possible new messenger for cell to cell interactions. What this new messenger does, what kind of new adaptations and cellular pathways are elicited by this particular interaction, is in no way explained by the existence of a new affinity for a ligand. That meta-function, which uses the new local biochemical function, is in no way simple like the mutation of two AAs, and in no way it is explained by it. That's where Behe's concept of irreducible complexity would be pertinent. But, obviously, the paper which was conceived to refute Behe says nothing about that.gpuccio_{May 15, 2014
May
05
May
15
15
2014
04:29 AM
4
04
29
AM
PDT}

Thanks gpuccio, I am not familiar with Axe. With respect to Behe, I think it should be further elaborated that he places the edge at different places depending on the reproductive capacity of the system where the protein in evolving. For example, 2 AAs may be the edge with respect to p falciparum but not HIV. I am interested to see what Piotr comes up with. So far, the experiments that I am aware of that attempt to refute Behe are shockingly weak. I would think they could do better. It makes me wonder if I am still giving too much credit to the power of RM+NS.Jehu_{May 14, 2014
May
05
May
14
14
2014
03:56 PM
3
03
56
PM
PDT}

Jehu: Everything can qualify as "C", but only transitions that generate new complex functional information would be considered for design inference. A transition of two AAs which generates a new function (in this case chloroquine resistance) is in the range of RV. It is much rarer of the more common 1 AA transitions which generate some other types of antibiotic resistance, even in palsmodium (see Behe), but it can happen by RV. The functional complexity of this transition would be 2 AAs, that is 8.64 bits. Behe considers that as the observed hedge of evolution. Axe considers 4-5 AAS (let's say 22 bits) as a computed "hedge" based on different observations about protein functionality. I have suggested 150 bits (35 AAs) as a very safe threshold to infer design in any biological system on our planet. So, let's say that is the transition from A to C implies more than 150 bits of functional information, we can be rather safe in inferring design for that transition, provided there is no evidence of an explicit path through some well demonstrated "B" (and there is no evidence, for all known basic protein superfamilies). All the scenarios we know of imply one aminoacid, maybe two. Classic simple antibiotic resistance, the expansion of Hemoglobin S in populations exposed to malaria, nylonase. These are good examples of molecular "microevolution": single or double mutations which, for particular environmental conditions, confer some remarkable advantage under strong selective pressure. That is all that pure RV can accomplish. As for the natural selection of intermediates as a step to complex function, I am aware of no observed example of such a mechanism. Let's see what Piotr can dig up. If two connected AAs is the best RV can do in decades of strong selection (see Behe), I suppose that a 35 AAs transition (my threshold) should be deconstructed at least into 17-18 two AAs transitions, each of then naturally selectable, to explain the complete process. Who can really believe that's the way 2000 protein superfamilies, most of them well beyond that threshold, were found? (In his paper, Durston computes the functional complexity for 35 protein families. 28 of them are beyond my threshold of 150 bits, with a range of 156 - 2,416 bits. The remain 7 families are short proteins, less than 100 AA long, and have functional complexity of 46 - 123 bits, well beyond Axe's threshold of 5 AAs - 22 bits. We could probably infer design for them too, considering carefully the context of their emergence, but let's say that they are not our priority).gpuccio_{May 14, 2014
May
05
May
14
14
2014
02:24 PM
2
02
24
PM
PDT}

Of course, I guess mutant PfCRT does not qualify as "C," a new protein.Jehu_{May 14, 2014
May
05
May
14
14
2014
01:05 PM
1
01
05
PM
PDT}

Most of us are familiar with the mutations in the PfCRT gene of P. falciparum that confer chloroquine resistance. As I recall, two mutations are required at the 76 and 163 position in order for there to be chloroquine resistance and a selective advantage where chloroquine is in use. The mutations can happen in any sequence but the first would be an example of a neutral mutation and the second a mutation that would be selected.Jehu_{May 14, 2014
May
05
May
14
14
2014
12:57 PM
12
12
57
PM
PDT}

Piotr: Do your work. I am not impatient. I just wanted to make more arguments more explicit. B needs to be an intermediate at sequence level, if we want to explain what we observe with that model. There is no hindsight, we are only truing to build a model that works to explain something we observe. And My A, B and C are not abstract. C is any new observed protein in the course of natural history. And A and B are necessary actors in the context of a darwinist explanation based on RV + NS. Indeed, I believe that B simply does not exist, and A can be any unrelated sequence which is modeled by design. For example, the non coding regions which become ORFs, ad observed in many cases.gpuccio_{May 14, 2014
May
05
May
14
14
2014
10:04 AM
10
10
04
AM
PDT}

Can you really believe that we can find no trace of all that in the existing proteome?
Nah, I'm sure we can find such traces if we know what to look for.
You yourself say that “it may be still present as a simpler homologue in the genomes of more distant relatives”. But not even that happens.
Doesn't it? We shall see. Please don't be impatient, I need to do a little reading.Piotr_{May 14, 2014
May
05
May
14
14
2014
09:48 AM
9
09
48
AM
PDT}

I have argued many times that, if mere survival and reproduction were really the driving engines of evolution, the process should have simply stopped at the first, remarkable achievement of prokaryotes.
It did stop there for at least two billion years; and two billion years later most life on Earth is still bacterial and archaeal (and, well, viral). The eukaryotic "revolution" succeeded only once (perhaps there were other false starts, but we can't know for sure). I don't intend to deny that the rise of Eukaryota was extremely unlikely, given that they had to compete against extremely well-adapted and efficient life forms. Prokaryotes have vast effective population, so even very slight selective pressures force them to adapt. "Junk DNA? No, thank you, sir; I have to keep fit."
So, what is B? It is supposed to be a functional, naturally selectable intermediate step.
Intermediate with hindsight, as I said. B is not a step towards anything. It's a derivative of A, full stop. It doesn't know where it's going (figuratively speaking). If selection fixes and conserves B, the further course of evolution is constrained by the structure and functionality of B, not the other way round. Constrained, but not strictly determined. B may in due course become C, or possibly D, or E, or simply remain B. There is no unique target.
To work as that, B must have two different properties: a) It must be related, at sequence level, both to A and to C. IOWs, in the sequence space it must be somehwere “in the middle” between A and C.
Please let me do my homework and I'll be back with concrete examples. When I speak in abstract terms like A, B and C, you accuse me of telling fairy tales.Piotr_{May 14, 2014
May
05
May
14
14
2014
09:44 AM
9
09
44
AM
PDT}

Piotr:
As soon as a more functional version appears, the old one is outcompeted and no longer protected by conservative selection. We probably won’t see it in the same species (unless it manages to survive by acquiring a new function), but it may be still present as a simpler homologue in the genomes of more distant relatives.
That is the old answer: "They were eaten!". But it is not possible. Remember, B is a functional protein which was expanded to the whole population at some time. And you need many Bs for each complex transition (we can discuss that later). And you need them for each transition which generated (at least) each of the 2000 superfamilies which we observe in the proteome. Can you really believe that we can find no trace of all that in the existing proteome? You yourself say that "it may be still present as a simpler homologue in the genomes of more distant relatives". But not even that happens. No, to believe that new functional sequences arise through a long chain of functional expanded intermediate proteins of which no trace remains in the proteome is really beyond any credibility. Only a dogmatic attitude can justify that kind of faith in so many self professed "skeptics".gpuccio_{May 14, 2014
May
05
May
14
14
2014
09:06 AM
9
09
06
AM
PDT}

Piotr: So, what is B? It is supposed to be a functional, naturally selectable intermediate step. To work as that, B must have two different properties: a) It must be related, at sequence level, both to A and to C. IOWs, in the sequence space it must be somehwere "in the middle" between A and C. b) It must be functional and naturally selectable (confer a reproductive advantage), so that it may be expanded/fixed by NS. The second point is fundamental. If B is not fixed by NS, the following variations that have to take place to get C must happen in the same individual in which the transition from A to B has taken place (or its individual clone). There is no multiplication of the probabilistic resources, and the probabilities of getting to C remain the same. Only the expansion of B multiplies the probabilistic resources of the system, and lowers the probabilistic barriers. But the first point is fundamental too. B must be related to A, and it must be related to C, at sequence level. Otherwise, it is of no help. If B is "near" A, but is unrelated to C, the transition from B to C is as improbable as the transition from A to C, even after B expands. If B is "near" C, but unrelated to A, the transition from A to B will simply never happen, because it has the same probabilities as the transition from A to C. So, B must be "in the middle" to be of some help. Please, remember that we are discussing the sequences here. That has nothing to do with structures and functions. RV happens at the sequence level. It knows nothing of structure and function. The random walk is a walk in the sequence space. All unrelated states have the same probability to be reached. Only related states at sequence level have higher probabilities. So, for B to be useful, it has to be a "step" from A to C at sequence level. And it has to be functional and selectable, and it must expand / be fixed. Otherwise, it is of no utility in the process.gpuccio_{May 14, 2014
May
05
May
14
14
2014
08:58 AM
8
08
58
AM
PDT}

Piotr:
Only of you think change is teleological (we are searching for a combination that allows us to achieve F. But an “intermediate” step is not really intermediate except with hindsight. A camera obscura eye is fine if you can’t build anything more sophisticated with the tools available. It isn’t an unsuccessful attempt to build an eye with a lens. Evolution doesn’t predict or anticipate its future solutions. But if a small improvement becomes possible, it outcompetes the old version which then and only then begins to look “primitive”. There is no “nearer F” until F appears (and F may itself be far from perfect — as is the human eye, notwithstanding its complexity). Sorry for using an example from morphology — it’s only for illustration.
You have to use an example from morphology, because the reasoning would not work if we pass at the level of molecular sequence. I will try to be more clear. Let's say that we observe the appearance of a new protein (domain, superfamily, as you like) at some time in natural history, for example in jawed fish. That's what we have to explain, and it's not important if we think it is teleological or not: it's what we observe. Also, it is not important if it happened in one day, one million years or 100 million years. We have a window of time in which it must have happened. Speaking of jawed fish, 4 billion years are certainly too many, so let's say we have a window of 100 million years. Now, let's say that we model what we observe as a transition form A to B to C. What is C? It is the new protein we observe in jawed fish. By definition, it is a new protein of a new superfamily. What is A? It is the original sequence, present in the immdediate ancestors, from which we believe that the transition towards C starts. It can be another protein gene, more or less duplicated or inactivated, or a non coding sequence, more or less functional. It does not matter. The important point is: A and C are unrelated at sequence level. Why? Because C is a new protein of a new superfamily, and therefore by definition it is unrelated at sequence level to all the other existing proteins. The sequence relation between superfamilies is lower than 10% identity, in the range of random identity. Even if A is a non coding sequence, there is absolutely no reason to assume that it has some relation with the future protein coding sequence C. Unless you are assuming design or, again, sheer luck. So we say that, if C is long enough and functionally complex, the transition from A to C is practically impossible by mere RV, because of the probabilistic barriers which I have discussed in detail. It is a random walk which will never happen. That's where B should help. More in next post.gpuccio_{May 14, 2014
May
05
May
14
14
2014
08:44 AM
8
08
44
AM
PDT}

Piotr:
Only as long as the selective pressure continues. It may continue almost indefinitely (as in the case of some really fundamental vital functions), but if, say, environmental pressures change, selection may be relaxed or even change its sign.
OK. I tend to think more in terms of fundamental functions, but what you say is correct, and we see it happening in antibiotic resistance, for example. But I think that darwinist interpretations tend to overemphasize the role of environmental pressure, and completely forget the importance of basic needs. For example, the adaptive immune system appears in jawed fish, and it can be explained as a result of the necessity of better defense of the new complex animals from their "simpler" competitors. Complexity makes biological beings frailer, and they need to defend themselves more and better, in order to be able to express new functions. That has not changed after that because of any environmental fluctuation, because it is a basic need. I have argued many times that, if mere survival and reproduction were really the driving engines of evolution, the process should have simply stopped at the first, remarkable achievement of prokaryotes.gpuccio_{May 14, 2014
May
05
May
14
14
2014
08:30 AM
8
08
30
AM
PDT}

Gpuccio: Sorry, I'm teaching today and have little time for pleasures. I agree with much of what you say in the last post, but some caveats are in order:
So, in NS the new state (allele) is fixed and it is is then conserved.
Only as long as the selective pressure continues. It may continue almost indefinitely (as in the case of some really fundamental vital functions), but if, say, environmental pressures change, selection may be relaxed or even change its sign.
a) A must confer a reproductive advantage, so that it can be fixed by NS. b) A must be a step which, at sequence level, takes us nearer to F (A + B).
Only of you think change is teleological (we are searching for a combination that allows us to achieve F. But an "intermediate" step is not really intermediate except with hindsight. A camera obscura eye is fine if you can't build anything more sophisticated with the tools available. It isn't an unsuccessful attempt to build an eye with a lens. Evolution doesn't predict or anticipate its future solutions. But if a small improvement becomes possible, it outcompetes the old version which then and only then begins to look "primitive". There is no "nearer F" until F appears (and F may itself be far from perfect -- as is the human eye, notwithstanding its complexity). Sorry for using an example from morphology -- it's only for illustration.
That is logic and reasonable, but there is further proof: those intermediate steps to a complex function, each of them functional and naturally selectable, have never been observed in biology.
As soon as a more functional version appears, the old one is outcompeted and no longer protected by conservative selection. We probably won't see it in the same species (unless it manages to survive by acquiring a new function), but it may be still present as a simpler homologue in the genomes of more distant relatives.Piotr_{May 14, 2014
May
05
May
14
14
2014
07:02 AM
7
07
02
AM
PDT}

Piotr and all: In the meantime, I want to profit of this quiet moment to clarify some important differences between drift and NS. To avoid confusion, let's calrify in the beginning that the term "fixation" is usually used in population genetics in the sense of "spread". I often use the term "expansion". However, "fixation" means "complete expansion". From Wikipedia:
In population genetics, fixation is the change in a gene pool from a situation where there exists at least two variants of a particular gene (allele) to a situation where only one of the alleles remains.[1] The term can refer to a gene in general or particular nucleotide position in the DNA chain (locus). In the process of substitution, a previously non-existent allele arises by mutation and undergoes fixation by spreading through the population by random genetic drift and/or positive selection. Once the frequency of the allele is at 100%, i.e. being the only gene variant present in any member, it is said to be "fixed" in the population.[1]
Now, we have already clarified that the spread/expansion of an allele is random in the case of drift: of all the possible mutations that happen, some are fixed, and many are lost. Which are fixed does not depend on the properties of the mutation itself, least of all on its function. That is the first and most important difference between drift and NS. The second difference is that, as most mutations are neutral or slightly deleterious, most mutations can be involved in drift, while only a tiny subset of mutations (those which can confer a reproductive advantage) can be involved in NS. That's why, even if in theory a functional mutation could be fixed by drift, if we observe a functional mutation that is fixed we can safely assume that it was fixed by NS. But there is a third, important difference between drift and NS that we must remember. I would sum it up as follows: Drift can only expand (fix) a mutation. NS not only expands (fixes), but also conserves. IOWs, drift is a single process, while NS is a double process: a) Positive selection: a mutation is expanded (fixed) because it confers reproductive advantage (IOWs, though selective elimination of the old form). b) Negative (purifying) selection: The expanded (fixed) mutation is conserved, because new mutations are selective eliminated because of the loss of function (reproductive disadvantage). So, in NS the new state (allele) is fixed and it is is then conserved. In drift, only the first thing happens. The new state is as subject to new changes as any other neutral sequence. A final note is that expansion/spread/fixation need not be complete: at each moment, a trait can be present in different percentages of the population (polymorphism). However, as the origin of each mutation happens in one copy only, any gene which is polymorphic, that is present in significant percentages of the population, must have undergone some important expansion, either through drift or NS. A last reflection: is expansion/spread/fixation important to understand the mechanism of generation of new complex information? The answer is: it depends. The only possible role of expansion/spread/fixation is to lower the probabilistic barriers. I will try to explain how. Let's say that mutation A is a step to some function F, but is not enough to achieve the function. Let's say that mutation B, added to A, achieves the function. Let's say that pA and pB are the probabilities of getting each of the two mutations in time t in a population of N numerosity. The probability of having both A and B in the same individual in time T is therefore pA X pB, which is much lower than either pA or pB. But, if A happens in one individual, and then in a short time is fixed, the probability of getting B together with A in one individual is now pB, and not pA x pB. IOWs, the probability of getting F in time t has become much higher. That means, in few words, that is A (a step to F) has a special reason to be expanded preferentially (vs other possible mutations that can be expanded), than the probability of getting F increases significantly. Fixation by drift cannot do that. Why? Because all mutations have the same probability of being expanded/fixed by drift. Therefore, A can get no preferential expansion versus all other mutations. And all other mutations are not a step to F, and their expansion will prevent the walk towards F, or at least not help it. IOWs, drift does not change the probabilities of getting F. What about NS? The situation is not very different. Why? Because, for NS to work as a facilitator in getting F, two things must be true at the same time: a) A must confer a reproductive advantage, so that it can be fixed by NS. b) A must be a step which, at sequence level, takes us nearer to F (A + B). Now, there is absolutely no reason why a) and b) should be connected, and therefore be true at the same time. If A is functional for reasons different from those for which F is functional, why should it also be a step towards the sequence of F? For sheer luck? But again, invoking sheer luck does not improve probabilities. That's another way of saying what I often say: complex functions cannot be deconstructed into simpler steps, each of them functional and naturally selectable. That is logic and reasonable, but there is further proof: those intermediate steps to a complex function, each of them functional and naturally selectable, have never been observed in biology. And yet, if the theory were true, each of them would have been fixed in its population at some time. So, why isn't there any trace of them in the proteome?gpuccio_{May 14, 2014
May
05
May
14
14
2014
05:12 AM
5
05
12
AM
PDT}

Section 2.3 of Nei's book Mutation-Driven Evolution is Difficulties of Defining and Estimating Selection Coefficients It opens as follows:
Although it is easy to develop mathematical theories of natural selection, it is very difficult to estimate genotype fitnesses or selection coefficients in natural populations.
Subsections: Estimates of Selection Coefficients and their Reliability Fluctuation of Selection Coefficients General Considerations For those who are interested it's worth a read.Mung_{May 13, 2014
May
05
May
13
13
2014
06:02 PM
6
06
02
PM
PDT}

Piotr: Give your models with references to real proteins, and I will try to analyze them from my point of view (the emergence of functional information). And you are right, it isn’t population genetics, its models and their parameters that really bother me, but the origin of new genes and functions (as opposed to their spread and fixation). And you are right, it's a nice discussion. :) Take all the time you need.gpuccio_{May 13, 2014
May
05
May
13
13
2014
02:20 PM
2
02
20
PM
PDT}

Piotr Start with an Orphan gene.Jehu_{May 13, 2014
May
05
May
13
13
2014
09:48 AM
9
09
48
AM
PDT}

Gpuccio: Now I could go on showing how the selection coefficient of de numerous novo genes is determined in real populations of fruit flies, but it's obvious that it isn't population genetics, its models and their parameters that really bother you, but the origin of new genes and functions (as opposed to their spread and fixation). It isn't quite fair to dismiss my abstract scenario as a fairy tale (especially if your alternative is spooky action by otherworldly mental powers). It is abstract, but it can easily be illustrated with actual examples of known gene families. I'll try to do that, but I'll need a little time to prepare my case. I can't hope to convince you, I suppose, but what the hell, It's a nice discussion.Piotr_{May 13, 2014
May
05
May
13
13
2014
06:28 AM
6
06
28
AM
PDT}

Piotr, No one even knows if the transformations required for a dinosaur to become a bird are even possible via genetic changes. Not only that but unguided evolution can't even account for the origin of Eukaryota, never mind Metazoa.Joe_{May 13, 2014
May
05
May
13
13
2014
06:28 AM
6
06
28
AM
PDT}

The modality of the interaction between the designer and biological matter is at present mainly the object of speculation. As you have asked my ideas about that, I have offered them. But I have no specific model beyond what I have already said.
OK. I will comment on that later.
The observations about the appearance of new superfamilies in natural history suggests a pattern with sudden increase of information at focal points, like OOL and the appearance of eukaryotes and of metazoa.
You don't know how sudden they were. You are taking advantage of gaps in the fossil record. The origin of Eukaryota and Metazoa can be dated only within a margin of error of the order +/- a few hundred million years. In the case of "focal points" which are well documented by fossils, e.g. the origin of birds, there is no real discontinuity and no "sudden increase" of anything. You can't even tell where "non-avian dinosaurs" end and "real birds" begin.Piotr_{May 13, 2014
May
05
May
13
13
2014
06:16 AM
6
06
16
AM
PDT}

Piotr:
While we are at it: how about a formal mathematical model of guided evolution by quantum-level manipulations conducted by an unobservable designer? Does it yield any quantifiable predictions? Can you please refer me to any literature on the subject, containing real-life case-studies? Or does it all boil down to the contemplation of wonderful molecular machines doing incredible things?
Some brief clarifications (which should already be clear, after our long exchanges): a) A design origin is inferred by observing: - Functional information which - Is vastly beyond the probabilistic resources of the system and - Cannot be explained algorithmically. The design inference is an inference, not a model. b) Once design is inferred, we can try to model the details of the implementation, as far as the data allow it. That has different aspects: - The modality of the interaction between the designer and biological matter is at present mainly the object of speculation. As you have asked my ideas about that, I have offered them. But I have no specific model beyond what I have already said. - On the contrary, it is certainly possible to model different aspects of how and when functional information appears in biological beings. For example, the observations about non coding regions which gradually become ORFs suggests gradual guided mutation. Another possibility is RV + IS, but at present I don't think that empirical observations support that model. The observations about the appearance of new superfamilies in natural history suggests a pattern with sudden increase of information at focal points, like OOL and the appearance of eukaryotes and of metazoa. Another aspect that can be studied is the role of NS in expanding the designed information, after it appears. And another one is the possible role of designed adaptation algorithms.gpuccio_{May 13, 2014
May
05
May
13
13
2014
04:53 AM
4
04
53
AM
PDT}

Piotr: a) It is true that I don't always read the whole article that is proposed (although generally I read more than the abstract), but it's only for lack of time. It is in the interest of those who suggest a reference to point to the relevant aspects, if possible. b) OK, they developed a simulated model and then they tested it against some datasets of the frequency of moth alleles, and similar. This is an example, if not of the measure of a coefficient, at least of an empirical validation of an assumed one in certain circumstances. OK, thank you for providing that. So, we know that NS probably works for those moth alleles. Which was never a point of controversy. c) About shifting posts, who is doing that? Which were the points of controversy? In my posts 4 and 12 I made many points to which you have not answered. In particular, in #12 I say: "Well, I think we agree that both drift and NS exist. That is not the point. The point is: what is their supposed role and relevance in trying to explain complex functional information, if you want to exclude design? Let’s suppose that we want to explain how some complex functional protein emerges for the first time at some point in natural history." Emphasis added. And I go on detailing the problem. In #15, I explain why population genetics cannot help solve those problems. I quote myself again: "No. We can calculate the expected effects of drift. But I am not aware of models which take into account realistic parameters for NS, for the simple reason that those realistic parameters don’t exist, because we know of no case of NS leading to complex functions. So, the models just assume some reproductive advantage, and just calculate how likely it is that such a reproductive advantage will be fixed. I am fine with that, but of what use is it? We know nothing of: a) How often some variation of defined complexity happens, and how often it realistically confers a reproductive advantage, and how great it can realistically be. b) If it is even possible that such variations can add up to give a complex function. So, population genetics is modeling its abstract assumptions, not reality." It is rather clear that the problem is not if NS can expand an existing tract which is functional (which I have always admitted), but of how it can contribute to its generation by expanding its precursors and creating new complex functional information. You have not answered any of that. Your only answer, in #19, has been to propose a fairy tale model, completely disconnected from any real data, about some imaginary G protein. I have detailed my objections to that in #20. You have "answered" in #28 with other vague reasoning, none of which was pertinent to my arguments, stating among other things that: "The selection coefficient may be variable (e.g. seasonally or regionally) and has to be determined empirically, because it is not a simple function of something we can easily measure. This is no problem in science." That was still about you example of the G gene, of its functions and of how protein families can evolve, as shown by the following paragraph: "Once the different functions have been segregated and proteins begin to specialise, you already have the first step in the growth of a protein family. It’s normal for a protein to have many potential uses, so G1 and G2 may acquire new secondary functions (which may already have been present to a minor degree in the “universal protein” G, but could not be honed to perfection because P and Q were the main targets of positive selection). Then the whole cycle is repeated and the family grows." In #35, I detail my objections to what you are saying. I quote myself again: "I am sorry for you, but your argument is meaningless. Please, give real examples, and not only words. Some of the problems: a) Proteins must be biochemically functional to be of use, in most cases. Again, what use is ATP synthase if it does not sinthesizes ATP? b) “The selection coefficient may be variable (e.g. seasonally or regionally) and has to be determined empirically, because it is not a simple function of something we can easily measure.” That is pure philosophy. Again, proteins are wonderful molecular machines. They do incredible things, or they do nothing. In many serious genetic diseases (the Mendelian diseases) a single mutation in the sequence inactivates the protein completely, wit tragic consequences. An enzyme is a wonderful machine, which achieves what can never be achieved by “natural” biochemical reactions. The activity of an enzyme can be measured in the lab. Maybe selection for an enzymatic activity can depend by environmental changes, but if the activity is not there, there is nothing to be selected. And please, give real examples of how the “selection coefficient” is measured in real cases. c) “Once the different functions have been segregated and proteins begin to specialise, you already have the first step in the growth of a protein family. It’s normal for a protein to have many potential uses, so G1 and G2 may acquire new secondary functions (which may already have been present to a minor degree in the “universal protein” G, but could not be honed to perfection because P and Q were the main targets of positive selection). Then the whole cycle is repeated and the family grows.” With all respect, these are pure fairy tales. Please, give empirical examples. I have tried, up to now, to debate facts and give facts. Please, do the same." Of all that, you only answer in #49 to the single phrase: "And please, give real examples of how the “selection coefficient” is measured in real cases." And you give me a reference where a model of selection is tested against data of existing alleles frequencies in moth. Which was never the object of the discussion. So, who is "shifting the goalposts"?gpuccio_{May 13, 2014
May
05
May
13
13
2014
04:04 AM
4
04
04
AM
PDT}

Gpuccio: Yes, it's a real case. Did you read beyond the abstract? See the section "Real data" (pp. 19-20) and the discussion that follows. You wanted an example of selection coefficients being estimated from empirical data and I provided one. Now you are shifting the goalposts. Selection coefficients can also be estimated for macroevolutionary changes using phylogenetic data, but of course the margin of error is wide and what you get is a distribution rather than an exact value. Population genetics makes no predictions as to how often "complex functions" arise (the rate of a given mutation is a parameter of the model), but, of course, once they do appear, a selection coefficient can be computed for them. You can take any species in which newly evolved proteins (or, say, recently duplicated gene families) are still getting fixed under positive selection (various species of Drosophila are strongly polymorphic in this respect), collect some data on the distribution of the competing alleles and do the calculations. Needless to say, it's being done too. While we are at it: how about a formal mathematical model of guided evolution by quantum-level manipulations conducted by an unobservable designer? Does it yield any quantifiable predictions? Can you please refer me to any literature on the subject, containing real-life case-studies? Or does it all boil down to the contemplation of wonderful molecular machines doing incredible things?Piotr_{May 13, 2014
May
05
May
13
13
2014
02:40 AM
2
02
40
AM
PDT}

Axel: I love to be the bad guy. It's my favorite role! :)gpuccio_{May 13, 2014
May
05
May
13
13
2014
12:32 AM
12
12
32
AM
PDT}

Piotr: Real case?
We do this by considering evolution in a simple lattice model of spatial population structure
What about real data of real populations? I quote myself: "No. We can calculate the expected effects of drift. But I am not aware of models which take into account realistic parameters for NS, for the simple reason that those realistic parameters don’t exist, because we know of no case of NS leading to complex functions. So, the models just assume some reproductive advantage, and just calculate how likely it is that such a reproductive advantage will be fixed. I am fine with that, but of what use is it? We know nothing of: a) How often some variation of defined complexity happens, and how often it realistically confers a reproductive advantage, and how great it can realistically be. b) If it is even possible that such variations can add up to give a complex function. So, population genetics is modeling its abstract assumptions, not reality."gpuccio_{May 13, 2014
May
05
May
13
13
2014
12:29 AM
12
12
29
AM
PDT}

And please, give real examples of how the “selection coefficient” is measured in real cases.
Try this, for example: Mathieson & McVean 2013, Estimating selection coefficients in spatially structured populations from time series data of allele frequenciesPiotr_{May 12, 2014
May
05
May
12
12
2014
11:47 PM
11
11
47
PM
PDT}

What happens in the various mathematical models of population genetics if the selection coefficient is not constant?
It's still mathematically tractable (if not analytically, then at least with the help of numerical simulations). If the selection coefficient varies in time, the changes are either seasonal (i.e. periodic) and can be averaged out over long spans of time, or follow environmental change (which is usually so slow that changes of s can be considered linear (I think we can leave aside extremely rare chance events like asteroid impacts). You can even model cases when selection coefficients vary randomly around certain mean values over generations. There are also mathematical models in which selection coefficients vary geographically (as a result, you get allele frequency clines between subpopulations inhabiting different environments). You can find basic references to relevant literature in Joseph Felsenstein's Theoretical Evolutionary Genetics (which even Sal Cordova agrees is a great book ;-)). See also this article: Uecker & Hermisson 2011, On the Fixation Process of a Beneficial Mutation in a Variable Environment and references therein.
e.g., “We predict that the A2 allele will decrease in frequency over time if there is a fixed relationship between phenotype and fitness (constant selection coefficient).” How does that work with “seasonal variations” in fitness
Come on, this is a problem for students in a four-credit honours course. They have to consider a problem using simplifying assumptions.Piotr_{May 12, 2014
May
05
May
12
12
2014
11:37 PM
11
11
37
PM
PDT}

#40 lolUpright BiPed_{May 12, 2014
May
05
May
12
12
2014
06:26 PM
6
06
26
PM
PDT}

Piotr:
The selection coefficient may be variable (e.g. seasonally or regionally) and has to be determined empirically…
What happens in the various mathematical models of population genetics if the selection coefficient is not constant? e.g., "We predict that the A2 allele will decrease in frequency over time if there is a fixed relationship between phenotype and fitness (constant selection coefficient)." How does that work with "seasonal variations" in fitness?Mung_{May 12, 2014
May
05
May
12
12
2014
06:26 PM
6
06
26
PM
PDT}

Natural selection is the process by which species adapt to their environment.
Propaganda. If Douggie had some evidence he would present it in his book. But all he sez is "Natural selection is the only process known to produce adaptations- here are some adaptations" (paraphrasing).Joe_{May 12, 2014
May
05
May
12
12
2014
06:19 PM
6
06
19
PM
PDT}

1 2 3 Next

You must be logged in to post a comment.

Leave a Reply