Intelligent Design

Natural Selection vs Artificial Selection

Spread the love

gent-selecting-a-suit-1514073Stimulated by the nth discussion with Zachriel on this point, I would like to offer here some thoughts about the difference between Natural Selection and Artificial Selection.

First of all, the dramatic limitation of NS is the following: it works on one functional specification, and one functional specification only: reproductive advantage.

In a sense, that specification is the byproduct of the system: biological beings that reproduce, that use limited resources to do that, and that compete for those resources. So, NS is a selection made possible by the existence of a complex functional system, and it selects for improvement in a function critically predefined in that system: reproductive success. So, it is a byproduct of the functional complexity already existing in the system.

Now, the functional specification: “reproductive success” is rather generic. It can certainly include many sub-functions. That’s the point that neo-darwinists stress. They say: OK, NS can select only for reproductive success, but reproductive success can include any function, and everything which goes in that sense can be selected.

Well, this is a false reasoning, which takes into no account the nature of complex information. The simple fact is: the search engine to which NS can be applied is random variation, and only random variation (I exclude for the moment possible algorithmic adaptation mechanisms). So, NS works on variation that is random, and not purposeful. Can that mechanism build complex functional information?

The simple answer is: no. Simple functional variation can certainly be generated by random variation, and therefore selected by NS. Why? because a few bits of variation are in a search space small enough to be explored, even many times, by biological RV. Those rare instances where the variation can give an advantage, with a bit of luck, can certainly be selected. That is the case of simple forms of antibiotic resistance. We can call those cases “molecular microevolution”. The few examples we have of that are the only empirical examples of NS at work in biology.

But what about some function which can give reproductive advantage, but which appears only if at least, say, 500 bits of specific information are found?

Such a result is definitely beyond any resource of RV. Therefore, it will never be achieved, and therefore never be selected.

Neo-darwinists, like Zachriel, argue that gradual pathways exist that will build those 500 bits of specific information in small steps. That is simply a fairy tale, existing only in their imagination. Information does not work that way. If I need 100 specific aminoacids to make something work (a case very common), then there is obviously no pathway which goes to that sequence step by step. Why? Because those 100 AAs are specific to the function. Fragments of the sequence have no special meaning and function, unless the complete sequence is achieved.

 

What about AS? Let’s take the case of the ATP binding protein, quoted by Zachriel (IOWs the Szostak paper).

This is AS, as many times argued by me here. The designer starts by conceiving and defining a function: “I want a protein which can effectively bind ATP”. That is the functional specification, and it is a form conceived in the consciousness of the designer.

As anyone can see, the function is very different from the single function available to NS: reproductive success.

Then, the designer uses his cognitive understanding of protein biochemistry and lab techniques to devise a strategy to implement his goal.

First of all, he sets a system that measures and extracts those molecules which bind ATP.

This point is very important, and it shows one of the main reasons why AS is so effective, while NS is not.

AS can measure the function defined by the designer at any desired level of sensitivity. Instead, NS has a definite threshold, under which no selection happens: reproductive success must be present, and enough of it to ensure the fixation of the trait.

That means that our designer, if interested in ATP binding, can select molecules which bind ATP with any level of affinity. There may be practical limitations due to the technology used, bu in principle any level of binding can be detected.

The reason is simple:

1) In NS, the coupling between the new function and the selection is direct: it is due to the reproductive success conferred by the function itself.

2) In AS, the coupling between the new function and the selection is indirect and symbolic: it’s the designer of the procedure who connects two events completely unrelated, for example ATP binding and the selection and expansion process. (UB, are you there?  🙂 )

In our example of ATP binding, then the designer chooses to use some form of artificial RV (in that case, mutagenic PCR), and to apply it in rounds coupled to artificial selection again and again.

The results are powerful: he obtains, in a short time, a protein with strong affinity for ATP.

The important point is: while that protein satisfies well enough the functional definition for which it was artificially selected (ATP binding), in no way it confers a reproductive advantage. So, even at the end of the artificial selection procedure, still the protein is not in the range of NS.

So, to sum up, the main differences between NS and AS are:

1) AS can define any function, and select for it. NS works only on one function: reproductive success.

2) In NS, the coupling between function and selection is direct: it’s the function itself which confers the reproductive advantage, which is the reason for the selection itself. In AS, the coupling between the defined function and the selection process is indirect and symbolic: the connection is established by the designer, by definite procedures designed by him.

3) NS has a definite threshold of measurement: it can only act if enough reproductive success is present as to ensure the fixation of the trait. AS can measure and select any desired level of the defined function.

4) In NS, the only selecting procedure is tied to the reproductive success, and is in essence differential reproduction. In AS, any intelligent procedure can be used to isolate, expand and fix the desired function.

 

87 Replies to “Natural Selection vs Artificial Selection

  1. 1
    EugeneS says:

    GP,

    Nice post! NS is passive, AS is active. NS does not select for a future function (but rather from among existing functions), AS does. NS crucially depends on there being a functional replicating system. AS has no such limitation.

    Some time ago, I asked you a question but you might have missed it somehow. It was regarding repetitive sequences and the C-paradox. How can you characterize those in light of the design hypothesis? Thanks!

  2. 2
    Zachriel says:

    gpuccio: First of all, the dramatic limitation of NS is the following: it works on one functional specification, and one functional specification only: reproductive advantage.

    Yes, but reproductive advantage may depend on molecular binding, or on the size of your teeth, or on how graceful you are when running.

    gpuccio: So, NS is a selection made possible by the existence of a complex functional system, and it selects for improvement in a function critically predefined in that system: reproductive success.

    The minimum requirement is reproduction, which may be as simple as a molecular replicator.

    gpuccio: The simple fact is: the search engine to which NS can be applied is random variation, and only random variation (I exclude for the moment possible algorithmic adaptation mechanisms).

    Random with respect to fitness is not pure randomness. For instance, according to endosymbiotic theory, bacteria invaded other cells. These cells responded by trying to isolate and minimize the effect of these invaders. Eventually, the bacteria were coopted and integrated in some lineages as mitochondria. Is that random? Certainly random mutation was part of the process.

    gpuccio: So, NS works on variation that is random, and not purposeful. Can that mechanism build complex functional information? The simple answer is: no.

    … huge hand wave, then conclusion…

    gpuccio: Such a result is definitely beyond any resource of RV. Therefore, it will never be achieved, and therefore never be selected.

    There is ample evidence to support that selection can lead to complex functional adaptation, as can be seen with protein evolution, wherein simple selection can generate highly specified, complex, three-dimensional structures. For a morphological example from nature, consider mammalian ossicles.

    gpuccio: Neo-darwinists, like Zachriel, …

    Not a Neodarwinist, as should be clear from above.

    gpuccio: … argue that gradual pathways exist that will build those 500 bits of specific information in small steps. That is simply a fairy tale, existing only in their imagination. Information does not work that way. If I need 100 specific aminoacids to male something work (a case very common), then there is obviously no pathway which goes to that sequence step by step. Why? Because those 100 AAs are specific to the function. Fragments of the sequence have no special meaning and function, unless the complete sequence is achieved.

    We know that’s false because we can show selectable pathways from random sequences to highly specified proteins.

    gpuccio: The designer starts by conceiving and defining a function: “I want a protein which can effectively bind ATP”.

    The function was chosen because it is a common function in nature. Nonetheless, it still shows a pathway exists, contrary to your claim.

    gpuccio: AS can measure the functioned defined by the designer at any desired level of sensitivity. Instead, NS has ha definite threshold, under which no selection happens: reproductive success must be present, and enough of it to ensure the fixation of the trait.

    This claim has some limited merit. The limitation is the assumption that we are starting with a random sequence, which may not be the case in nature. Nor are all random sequences of the same functionality. Some may have significant function and still be reasonably probable.

    In any case, per the claim, if we were to replace a functional sequence in a living organism with a random sequence, then it should not be able to evolve a replacement because the initial level of function would be too low to be selectable. However, in this experiment, the phage with the inserted random sequence exhibited varying levels of reproductive success, meaning the modified trait is selectable.

    Hayashi et al., Can an Arbitrary Sequence Evolve Towards Acquiring a Biological Function?, Journal of Molecular Evolution 2003: “The infectivity assays of the clones revealed that the first generation comprised a population with selectable variation in view of infectivity.”

  3. 3
    Zachriel says:

    EugeneS: NS is passive, AS is active. NS does not select for a future function (but rather from among existing functions), AS does. NS crucially depends on there being a functional replicating system. AS has no such limitation.

    Good point. Artificial selection can look forward to a future state, or in gpuccio’s nomenclature, to envision the goal in the conscious mind. However, traditional artificial selection didn’t look towards the distant future, but only considered change on a generation-to-generation basis. Indeed, the traditional concept that selection was meant only to bring out the “true form” is contrary to the evolutionary idea that progressive change being possible.

  4. 4
    Virgil Cain says:

    Yes, but reproductive advantage may depend on molecular binding, or on the size of your teeth, or on how graceful you are when running.

    And natural selection has nothing to do with any of those.

    Random with respect to fitness is not pure randomness.

    In a debate over if the mutations are directed or not, it is pure nonsense, though

    There is ample evidence to support that selection can lead to complex functional adaptation,

    Not natural selection. You are lying, again.

    For a morphological example from nature, consider mammalian ossicles.

    Right and there isn’t any evidence that natural selection did it. Did you have a point?

    Hayashi et al., Can an Arbitrary Sequence Evolve Towards Acquiring a Biological Function?, Journal of Molecular Evolution 2003: “The infectivity assays of the clones revealed that the first generation comprised a population with selectable variation in view of infectivity.”

    Those “arbitrary sequences” were designed.

  5. 5
    gpuccio says:

    Zachriel:

    Welcome to the discussion, which was anyway inspired by you! 🙂

    I find your comments very reasonable in general, but scarcely relevant. I agree with many things, so I will try to point at what I don’t agree with.

    Yes, but reproductive advantage may depend on molecular binding, or on the size of your teeth, or on how graceful you are when running.

    This is a very good point, which I should have discussed more in detail, even if I have touched it briefly:

    “Now, the functional specification: “reproductive success” is rather generic. It can certainly include many sub-functions. That’s the point that neo-darwinists stress. They say: OK, NS can select only for reproductive success, but reproductive success can include any function, and everything which goes in that sense can be selected.”

    You are right: reproductive success can depend on many different things. The point is, some of those “sub-functions” are simple, but many are extremely complex. Skin color variation can be a simple thing, in some cases it could depend on a single mutation. But ATP synthase and the genetic code are complex entities. Antibiotic resistance in its simple form often depends on one or two mutations, as discussed by Behe in his TEOE. But antibiotic resistance based on enzymes like penicillinases is complex, and can only be transmitted by HGT, it is never acquired, in empirical observations, by the emergence of a completely new functional protein.

    So, if in a software you need some ordering algorithm at some crucial point to improve efficiency, you will not get it by magic: you have to develop that subfunction, which has its own complexity, and then insert it in the right context. No blind step by step procedure will generate the ordering algorithm by gradual improvements of the original software. If you try to change your software by random search, and you can only measure the original software efficiency as feedback, you will never get a complex sub-algorithm which improves the original software. The concept is simple, and I am sure that you can see what I mean.

    The minimum requirement is reproduction, which may be as simple as a molecular replicator.

    No. As I said, the minimum requirement is reproducting entities which use limited resources to reproduce and which compete for them.

    Random with respect to fitness is not pure randomness. For instance, according to endosymbiotic theory, bacteria invaded other cells. These cells responded by trying to isolate and minimize the effect of these invaders. Eventually, the bacteria were coopted and integrated in some lineages as mitochondria. Is that random? Certainly random mutation was part of the process.

    OK. Some parts of this highly hypothetical and in no way understood process could also be what I call “possible algorithmic adaptation mechanisms”. I have purposefully avoided to discuss this aspect in this post, but you will agree that in general it does not seem to be the main component of biological evolution. Anyway, any existing adaptation algorithm has its complexity, which must be explained.

    There is ample evidence to support that selection can lead to complex functional adaptation, as can be seen with protein evolution, wherein simple selection can generate highly specified, complex, three-dimensional structures.

    What selection? What evidence?

    Not a Neodarwinist, as should be clear from above.

    Well, I apologize. I use neo-darwinist in a very general sense, and your arguments have always been much in accord with what I consider current neo darwinist theory, maybe not strictly a-la-Dawkins, but almost. Maybe I miss something. Anyway, you can specify better you position, if you like.

    We know that’s false because we can show selectable pathways from random sequences to highly specified proteins.

    Naturally selectable? That’s the whole point of my OP. If you mean intelligently selectable, that is completely non relevant.

    The function was chosen because it is a common function in nature. Nonetheless, it still shows a pathway exists, contrary to your claim.

    An intelligently selectable path is not a naturally selectable path. That’s the whole point of my OP. Designed search overcomes rather easily the probabilistic barriers which are implicit in RV + NS.

    This claim has some limited merit. The limitation is the assumption that we are starting with a random sequence, which may not be the case in nature. Nor are all random sequences of the same functionality. Some may have significant function and still be reasonably probable.

    OK, but what do you mean by that? The point which I have made many times is that when a new complex protein superfamily appears in natural history, and that sequence is completely unrelated to what existed before, then what existed before is “random” with respect to the new functional sequence.

    Regarding the paper you quote, I have read the summary, and it seems similar to the paper about ragged landscape that I often quote. I could discuss it in detail, but I will not do that now, also because I should be able to read the whole article to do that well. In brief, I have no problems that a random polypeptide which reduces the efficiency of an existing protein in an existing system can “improve” its behaviour in a highly efficient mutation selection system (phage infectivity). The problem with these “function retrival” experiment is that no new function is generated: they simply “damage” the original function somewhat, and then the system acquires some “damage limitation” through RV + NS.

    The same happened in the ragged landscape paper, but there the authors inferred the very interesting result that it is impossible to retrieve the wild type efficiency by that kind of process, because we would need a starting library of about 10^70 random sequences to achieve that result.

    The abstract of the paper you quote seems to confirm that point: the initial “damage” reduces infectivity of six orders of magnitude, while the “evolved” protein shows only a 240-fold increase in infectivity.

    So, to sum up:

    a) The function is already present, and remains the same. Those experiments would not work at all if the function were suppressed.

    b) The function is only minimally restored by the RV + NS process. The wild type efficiency seems to be beyond the realistic capacities of such a system in nature.

  6. 6
    gpuccio says:

    Zachriel:

    Ah, I forgot mammalian ossicles.

    As you know, I don’t discuss morphological issues, unless the molecular basis for them is known. Variation acts at the molecular level. Only there we can evaluate the complexity of some functional variation

  7. 7
    Virgil Cain says:

    gpuccio:

    As you know, I don’t discuss morphological issues, unless the molecular basis for them is known. Variation acts at the molecular level. Only there we can evaluate the complexity of some functional variation.

    Dr Behe brings up that very point in “Darwin’s Black Box” and evolutionists have never addressed it. They think that by ignoring that aspect of reality it will somehow go away.

  8. 8
    gpuccio says:

    EugeneS:

    Thank you for the kind comment!

    You ask:

    “Some time ago, I asked you a question but you might have missed it somehow. It was regarding repetitive sequences and the C-paradox. How can you characterize those in light of the design hypothesis?”

    Yes, I must have definitely missed it.

    OK.

    1) Repetitive sequences:

    There are many kinds of them. I will comment essentially about transposable elements, which are the majority of non coding DNA.

    As you probably know, I am a big supporter of the concept that great part of non coding DNA is probably functional. That includes SINEs, LINEs, and other transposable elements.

    You may also know that I have many times defended here the idea that transposons are a tool for biological design. That is supported by many molecular data which are accumulating, where functional sequences seem to have origin from transposon activity.

    Therefore, one important possibility is tan transposons and non coding DNA are a “factory” for new functional information.

    But I also think that repetitive sequences have function in the epigenetic regulation of cell differentiation and of cell activity. They could be crucial in many processes, like the determination of chromatin 3d structure in different contexts, and the generation of functional non coding RNAs.

    Obviously, part of them can be non functional, as neo darwinists believe of the whole thing.

    2) The C-paradox:

    Well, the number of genes seems to be rather similar in most metazoa: our 20000 genes are more or less the standard. So, what about C-value differences, which seem to defy any logic?

    I don’t know. But I have no big problems with the fact. I suppose that we cannot understand anything until we have some detailed knowledge of the composition of the highest C-value organisms. I have not found many data about that. There is certainly a great variety of functional implementation in different species, and we still understand very little about that.

  9. 9
    gpuccio says:

    Virgil Cain:

    Thank you for your comments. 🙂

  10. 10
    Mung says:

    And even if it can bind ATP, then so what? Just because there’s binding doesn’t means there’s any activity. Right Zachriel?

  11. 11
    EugeneS says:

    GPuccio,

    Many Thanks!

  12. 12
    wd400 says:

    If I need 100 specific aminoacids to make something work (a case very common)

    Common? Can you name some examples?

    The rest of this post seems to amount to a long argument that breeders can chose which traits to select for. Sure. So what?

  13. 13
    gpuccio says:

    wd400:

    ATP synthase alpha and beta subuinits.

    Triosephosphate isomerase.

    Phosphoglycerate kinase.

    Leucine–tRNA ligase

    Elongation factor G, mitochondrial isoform 2

    Histone H3

    For all of them, hundreds of AAs conserved between bacteria and humans, or fungi and humans (for histone H3).

    The purpose of the post is to show the big differences between artificial selection and natural selection, because neo darwinists go on trying to conflate the two things, which have completely different powers.

  14. 14
    wd400 says:

    ATP synthase alpha and beta subuinits….

    I’m sure these have hundreds of shared residues in these proteins, that’s a very different claim that “100 specific residues” are required for a function.

    The purpose of the post is to show the big differences between artificial selection and natural selection, because neo darwinists go on trying to conflate the two things, which have completely different powers.

    Well, I don’t think you’ve shown their powers are completely different — only that natural selection only works on fitness. But no one has claimed otherwise. Artificial selection shows that where there is genetic variance selection can rapidly generate phenotype change. And important point to make, I think.

  15. 15
    gpuccio says:

    wd400:

    You say:

    Artificial selection shows that where there is genetic variance selection can rapidly generate phenotype change. And important point to make, I think.” (emphasis mine)

    Well, why not:

    Artificial selection shows that where there is genetic variance artificial selection can rapidly generate phenotype change. And important point to make, I think.”

    That’s what I mean by “neo darwinists go on trying to conflate the two things, which have completely different powers.” You have made my case.

    If artificial selection can do something, it does not ensue that natural selection can do the same thing. Because, as I said, there are “big differences between artificial selection and natural selection”.

    An important point to make, I think.

  16. 16
    wd400 says:

    These “big differences” are that breeders get to chose what to breed for? And selection just means differential surivial based on some trait — that’s true if it’s artificial or natural.

  17. 17
    Virgil Cain says:

    Natural selection could never produce the breeds of dogs artificial selection has created. And natural selection could never produce ATP synthase- only intentional design can.

  18. 18
    Zachriel says:

    gpuccio: But ATP synthase and the genetic code are complex entities.

    While a single binding protein isn’t as complex as ATP synthase, it is certainly complex. As we have seen, even random sequences can having selectable catalytic capability.

    gpuccio: As I said, the minimum requirement is reproducting entities which use limited resources to reproduce and which compete for them.

    Molecular replication entails competition for limited resources.

    gpuccio: I have purposefully avoided to discuss this aspect in this post, but you will agree that in general it does not seem to be the main component of biological evolution.

    These sorts of mechanisms are certainly important in the early evolution of life, and as your primary concern is evolution of metabolic processes, they are very relevant.

    Zachriel: There is ample evidence to support that selection can lead to complex functional adaptation, as can be seen with protein evolution, wherein simple selection can generate highly specified, complex, three-dimensional structures.

    gpuccio: What selection? What evidence?

    Hayashi et al., among others.

    gpuccio: I use neo-darwinist in a very general sense, and your arguments have always been much in accord with what I consider current neo darwinist theory

    We take a pluralistic view of evolution, suitable to a historical process.

    gpuccio: If you mean intelligently selectable, that is completely non relevant.

    You claimed there was no selectable pathway. Unless you are claiming proteins are never subject to selective optimization in nature, then the point is directly relevant.

    gpuccio: An intelligently selectable path is not a naturally selectable path.

    That’s like saying we can’t study how salt dissolves in water in nature if we put salt and water in a beaker. Are you claiming proteins are never subject to optimizing selection in nature?

    gpuccio: In brief, I have no problems that a random polypeptide which reduces the efficiency of an existing protein in an existing system can “improve” its behaviour in a highly efficient mutation selection system (phage infectivity).

    Therefore, we now know that highly-specified, complex structures can form through selection.

    gpuccio: As you know, I don’t discuss morphological issues, unless the molecular basis for them is known.

    Ignoring a primary line of evidence undermines your argument.

  19. 19
    Virgil Cain says:

    Ignoring a primary line of evidence undermines your argument.

    And continually spewing nonsense proves that you don’t have one.

  20. 20
    Virgil Cain says:

    While a single binding protein isn’t as complex as ATP synthase, it is certainly complex. As we have seen, even random sequences can having selectable catalytic capability.

    Natural selection didn’t have anything to do with producing those random sequences. You are being dishonest at best.

  21. 21
    OldArmy94 says:

    I wish I knew how to better articulate this, but it seems to me that a big problem with natural selection is that there isn’t possibly enough time to “select” for all possible survival advantages from the population. There are just too many traits along an infinite scale of variation to make such a simplistic process work. I don’t know if that makes any sense, so sorry for my stumbling.

  22. 22
    computerist says:

    Zachriel,

    In your view, do you think there is any argument at all that can be made against evolution via NS&RM as the primary driving force of biological complexity? (regardless of whether it comes out of the evo or ID camp)

    Just wondering, thanks.

  23. 23
    gpuccio says:

    wd400 #16:

    The “big differences” are analyzed in detail in my OP. Did you read it?

    They are summed up in 4 points at the end, each different from the other. Have you considered them?

    And nowhere in the OP I refer to “breeders”. I rather quote, as an example of artificial selection, a famous molecular experiment by a Nobel prize winner.

    Are you really sure that you read what I wrote?

  24. 24
    wd400 says:

    I’ve read it. The four points just use more or fewer words to say the same thing: Breeders get to chose what to breed for. You can sub in “select” for “breed” if you want, I don’t think it changes that point.

  25. 25
    gpuccio says:

    Zachriel:

    You claimed there was no selectable pathway.

    No. This is my statement:

    But what about some function which can give reproductive advantage, but which appears only if at least, say, 500 bits of specific information are found?

    Such a result is definitely beyond any resource of RV. Therefore, it will never be achieved, and therefore never be selected.

    Neo-darwinists, like Zachriel, argue that gradual pathways exist that will build those 500 bits of specific information in small steps. That is simply a fairy tale, existing only in their imagination. Information does not work that way. If I need 100 specific aminoacids to make something work (a case very common), then there is obviously no pathway which goes to that sequence step by step. Why? Because those 100 AAs are specific to the function. Fragments of the sequence have no special meaning and function, unless the complete sequence is achieved.”

    It is rather clear. No gradual pathway exists to a new function which requires, for example, 500 bits of specific sequence information to appear.

    Unless you are claiming proteins are never subject to selective optimization in nature, then the point is directly relevant.

    I did not claim that. Optimization of an existing protein is a possibility, probably even for NS. I have discussed that point many times, for example for the possible optimization of an existing active site, whose affinity can change with a small number of mutations.

    That’s why, as you may have noticed, I always refer to protein superfamilies as examples of isolated functional islands. I have always kept an open mind for optimization or limited variation inside a family. Those examples should be evaluated case by case, according to the functional information necessary to achieve the transition. Axe has debated some aspects of that kind of problem.

    gpuccio: An intelligently selectable path is not a naturally selectable path.

    That’s like saying we can’t study how salt dissolves in water in nature if we put salt and water in a beaker. Are you claiming proteins are never subject to optimizing selection in nature?

    This comment of yours is really senseless. My OP is exactly about the essential differences between NS and AS. The way salt dissolves is the same in nature and in the beaker. But the process of selection and its modalities, powers and results are completely different in NS and AS, which is exactly my point.

    Bad metaphor, or simply bad trick?

    Regarding optimization, see previous point.

    Therefore, we now know that highly-specified, complex structures can form through selection.

    Absolutely not. As I have explained, that case is not the generation of a new highly-specified and complex structure. It is simply a case of partial retrieval from damage of an existing functional structure. The partial “optimization” (which remains however a strong reduction of function) is probably achieved through a few bits of variation, completely in the range of the probabilistic resources of that system.

    Try to explain why they do not achieve the wild type efficiency, or why, according to the authors of the ragged landscape paper, a starting random library of 10^70 molecules would be necessary for that result.

    We definitely don’t know what you claim that we know.

    gpuccio: As you know, I don’t discuss morphological issues, unless the molecular basis for them is known.

    Ignoring a primary line of evidence undermines your argument.

    I am not ignoring anything, least of all a “primary line of evidence”. My whole argument is based on the evaluation of functional complexity. We cannot evaluate the functional complexity of a variation whose molecular basis is not known. That would simply be very bad scientific methodology.

    I leave those kind of things to neo-darwinists, or to those who take a pluralistic view of evolution, suitable to a historical process.

  26. 26
    gpuccio says:

    wd400:

    Then I am afraid there is some problem in my way of writing, or in your understanding.

  27. 27
    Alicia Cartelli says:

    “If I need 100 specific aminoacids to make something work (a case very common), then there is obviously no pathway which goes to that sequence step by step. Why? Because those 100 AAs are specific to the function.”

    Completely false. You could not be any more wrong.

  28. 28
    Mung says:

    Alicia Cartelli:

    Completely false. You could not be any more wrong.

    Completely false. You could not be any more wrong.

    Another weekend of your nonsense? Don’t you have some research to catch up on?

  29. 29
    gpuccio says:

    Mung:

    Just for the record: I don’t discuss with Alicia Cartelli.

  30. 30
    gpuccio says:

    To all:

    I don’t want to give the impression that I believe that AS can do everything.

    The point is: it is a form of design, even if indirect design: what is designed is the strategy to achieve a conceived result, by using some cognitive information available to the designer, or which is gradually acquired through the strategy.

    Therefore, it can do a lot more than NS. It can generate some dFSCI, while NS can’t.

    But I don’t think that a strategy based on AS only can achieve everything. More complex designs require more active input of cognitive content by the designer than just AS.

    Even the ATP binding example, indeed, is not a simple implementation of AS. It requires more: for example, the knowledge to build the initial library:

    “Because protein sequences with specific functions are expected to be quite rare in protein sequence space, we prepared a DNA library of 4 x 10^14 independently generated random sequences. This DNA library was specifically constructed to avoid stop codons and frameshift mutations, and was designed for use in mRNA display selections.”

    and all the biological understanding and technology to implement the rounds of selection:

    ” In each round the mRNA-displayed proteins were incubated with immobilized ATP, washed and eluted with free ATP. The eluted fractions were collected and amplified by polymerase chain reaction (PCR); this DNA was then used to generate a new library of mRNAdisplayed proteins, enriched in sequences that bind ATP, for input into the next round of selection”

    and then the rounds of mutation + selection:

    “In an effort to increase the proportion of these proteins that fold into an ATP-binding conformation, we mutagenized the library and carried out further rounds of in vitro selection and amplification. Three consecutive rounds with mutagenic PCR ampli®cation were performed with an average mutagenic rate of 3.7% per amino acid for each round.”

    Well, that is certainly much more design than just breeding and selecting for desired traits!

    However, even with that effort, no really biologically useful protein was attained.

    I don’t believe that a really complex and functional protein like ATP synthase, for example, cab be engineered by this kind of methodology alone. Such a sophisticated and effective result requires IMO other forms of design, including guided variation or direct building, which imply much greater understanding of protein biochemistry and of the relationship between sequence and function.

    Maybe one day we will be able to do that kind of thing, but certainly not only by AS on a starting random library.

    However, it is perfectly possible that some controlled RV followed by AS still will be a part of the whole procedure.

  31. 31
    gpuccio says:

    Nobody has commented on point number 2 in my summary. I believe it is a very important point, so I repeat it here:

    2) In NS, the coupling between function and selection is direct: it’s the function itself which confers the reproductive advantage, which is the reason for the selection itself. In AS, the coupling between the defined function and the selection process is indirect and symbolic: the connection is established by the designer, by definite procedures designed by him.

    That implies and configures, I believe, UB’s concept of semiosis.

  32. 32
    Dionisio says:

    Very interesting OP.

  33. 33
    Dionisio says:

    There is certainly a great variety of functional implementation in different species, and we still understand very little about that. [@8]

    As outstanding questions get answered, are new questions raised?

    The deeper scientists research, the more they discover, is there more information for us to understand?

  34. 34
    Dionisio says:

    “The designer starts by conceiving and defining a function: […]. That is the functional specification, and it is a form conceived in the consciousness of the designer.”

    […] the designer uses his cognitive understanding of […] to devise a strategy to implement his goal.

    Do software developers enter/modify code randomly until they reach “bingo!” moments (i.e. get something useful)? 🙂

  35. 35
    gpuccio says:

    Dionisio:

    Welcome to the discussion!

    “The deeper scientists research, the more they discover, is there more information for us to understand?”

    Absolutely! Biological complexity seems really to be a bottomless well.

    At the end of the 19th century, many scientists believed in the “end of physics”. Many still do. But physics seems still well far from its ending.

    Maybe neo darwinists (or variants) believe in the “end of biology”. They will be really disappointed.

    If there may be any disappointement in the revelation of endless, beautiful intelligence. 🙂

  36. 36
    Dionisio says:

    “AS can measure the functioned defined by the designer at any desired level of sensitivity.”

  37. 37
    bornagain says:

    gpuccio, thanks for your comments differentiating AS and NS. Here is a reference for one of your claims:

    A Man-Made ATP-Binding Protein Evolved Independent of Nature Causes Abnormal Growth in Bacterial Cells – 2009
    Excerpt: “Recent advances in de novo protein evolution have made it possible to create synthetic proteins from unbiased libraries that fold into stable tertiary structures with predefined functions. However, it is not known whether such proteins will be functional when expressed inside living cells or how a host organism would respond to an encounter with a non-biological protein. Here, we examine the physiology and morphology of Escherichia coli cells engineered to express a synthetic ATP-binding protein evolved entirely from non-biological origins. We show that this man-made protein disrupts the normal energetic balance of the cell by altering the levels of intracellular ATP. This disruption cascades into a series of events that ultimately limit reproductive competency by inhibiting cell division.”
    http://www.plosone.org/article.....ne.0007385

    Dr. Hunter humorously notes the shortfall in Darwinian explanations (even supposing the protein would have been ‘functional’ in regards to NS):

    How Proteins Evolved – Cornelius Hunter – December 2010
    Excerpt: Comparing ATP binding with the incredible feats of hemoglobin, for example, is like comparing a tricycle with a jet airplane. And even the one in 10^12 shot, though it pales in comparison to the odds of constructing a more useful protein machine, is no small barrier. If that is what is required to even achieve simple ATP binding, then evolution would need to be incessantly running unsuccessful trials. The machinery to construct, use and benefit from a potential protein product would have to be in place, while failure after failure results. Evolution would make Thomas Edison appear lazy, running millions of trials after millions of trials before finding even the tiniest of function.
    http://darwins-god.blogspot.co.....olved.html

    Of related note, it is good to realize just how big 10^12 (a trillion) actually is:

    “The largest dump truck in the world would have to carry more than nine full loads to move a trillion grains of sand. A regular dump truck will have to make 150 trips.”
    http://www.bobkrumm.com/blog/2.....-trillion/

    A few more notes:

    Protein Life Times: Just-Right Evidence for Design – Fazale Rana PhD. – biochemistry
    Excerpt: Researchers learned that the amino acid sequences are exquisitely arranged to precisely balance the need for structural stability, while minimizing aggregation propensity.,,, Yet the optimization of proteins is not limited to their aggregation propensities. A cascade of optimization characterizes protein structure and function. In The Cell’s Design, I described a number of other ways that protein structure is optimized.
    http://www.reasons.org/article.....for-design

    Strange Behavior: New Study Exposes Living Cells to Synthetic Protein – Dec. 27, 2012
    Excerpt: ,,,”ATP is the energy currency of life,” Chaput says. The phosphodiester bonds of ATP contain the energy necessary to drive reactions in living systems, giving up their stored energy when these bonds are chemically cleaved. The depletion of available intracellular ATP by DX binding disrupts normal metabolic activity in the cells, preventing them from dividing, (though they continue to grow).,,,
    In the current study, E. coli cells exposed to DX transitioned into a filamentous form, which can occur naturally when such cells are subject to conditions of stress. The cells display low metabolic activity and limited cell division, presumably owing to their ATP-starved condition.
    The study also examined the ability of E. coli to recover following DX exposure. The cells were found to enter a quiescent state known as viable but non-culturable (VBNC), meaning that they survived ATP sequestration and returned to their non-filamentous state after 48 hours, but lost their reproductive capacity.
    Further, this condition was difficult to reverse and seems to involve a fundamental reprogramming of the cell.
    http://www.sciencedaily.com/re.....143001.htm

  38. 38
    gpuccio says:

    Dionisio #36:

    Corrected. Thank you for the tip. 🙂

  39. 39
    gpuccio says:

    BA:

    Thank you for the contributions. 🙂

    The simple fact is: a protein which avidly binds ATP, but has no other enzymatic function, except maybe minimal ATPase activity, is of no use, indeed it is a damage, because it simply subtracts ATP from the environment.

    ATP is a source of energy. ATP binding proteins, the true ones, use the energy in ATP to accomplish something.

    For example, take hexokinase, the first enzyme in glycolysis. It tranfers the phophate group from ATP to glucose, and so it starts the glycolysis process.

    You are right, 10^12 is a big number. But in some way we can prepare a library of that size.

    But just think of the size of the random library which would be necessary to retrieve wild type efficiency in the ragged landscape phage experiment, according to the authors: about 10^70!

    That’s much more than the estimated number of atoms on our planet, which is about 10^50!

    I doubt that we will ever be able to prepare such a library. And even our whole planet obviously can’t do that.

  40. 40
    gpuccio says:

    wd400:

    Here are the blast results for histone H3, human vs saccharomyces cerevisiae.

    Score: 248 bits
    Expect: 4e-84
    Identities: 121/136(89%)
    Positives: 130/136(95%)
    Gaps: 0/136(0%)

    Now, I suppose that we can safely assume at least 500 million years of evolutionary separation between fungi and metazoa. Maybe more than that.

    So, how can you explain that 121 AAs out of 136 are exactly the same after such a long evolutionary time? Have you any explanation which is better than extremely strong functional constraint?

    Just to know.

  41. 41
    Dionisio says:

    […] suppose that we can safely assume at least 500 million years of evolutionary separation between fungi and metazoa. Maybe more than that.

    So, how can you explain that 121 AAs out of 136 are exactly the same after such a long evolutionary time? Have you any explanation which is better than extremely strong functional constraint?

    I’ve read in this site some interlocutors affirming that the explanation has been known since long time ago and it’s somewhere out there in the textbooks. We just don’t understand biology. They recommend we study biology 101. 🙂

  42. 42
    Zachriel says:

    OldArmy94: I wish I knew how to better articulate this …

    Your articulation is fine.

    OldArmy94: but it seems to me that a big problem with natural selection is that there isn’t possibly enough time to “select” for all possible survival advantages from the population. There are just too many traits along an infinite scale of variation to make such a simplistic process work.

    We can directly observe natural selection in the wild, so we know it works.

    computerist: In your view, do you think there is any argument at all that can be made against evolution via NS&RM as the primary driving force of biological complexity?

    Sure. The evidence indicates there are many mechanisms other than simple random mutation at work in the history of life, including recombination, speciation, canalization, and endosymbiosis.

  43. 43
    Virgil Cain says:

    We can directly observe natural selection in the wild

    Actually we can’t do that because we don’t know if the variations were accidental.

    The evidence indicates there are many mechanisms other than simple random mutation at work in the history of life, including recombination, speciation, canalization, and endosymbiosis.

    And not one of those is known to produce any complex functional systems.

  44. 44
    Zachriel says:

    gpuccio: Zachriel, argue that gradual pathways exist that will build those 500 bits of specific information in small steps.

    That is correct, as experiments with protein evolution show.

    gpuccio: Because those 100 AAs are specific to the function. Fragments of the sequence have no special meaning and function, unless the complete sequence is achieved.”

    Sure, but parent sequences can have their own functions, and existing sequences can be recombined.

    gpuccio: No gradual pathway exists to a new function which requires, for example, 500 bits of specific sequence information to appear.

    That is incorrect, as experiments with protein evolution shows.

    gpuccio: I did not claim that.

    In which case, experiments with optimizing selection are relevant.

    gpuccio: But the process of selection and its modalities, powers and results are completely different in NS and AS, which is exactly my point.

    In the case of optimizing selection for a specific, naturally occurring function, what is the difference? What if selection is for reproductive advantage, as in Hayashi et al.?

    gpuccio: The partial “optimization” (which remains however a strong reduction of function) is probably achieved through a few bits of variation, completely in the range of the probabilistic resources of that system.

    The few bits of variation result in turning a random sequence into a specific and complex three-dimensional structure.

    gpuccio: Try to explain why they do not achieve the wild type efficiency, or why, according to the authors of the ragged landscape paper, a starting random library of 10^70 molecules would be necessary for that result.

    Because they didn’t use recombination, of course.

    gpuccio: My whole argument is based on the evaluation of functional complexity.

    Functional complexity exists in morphological space as well.

    gpuccio: We cannot evaluate the functional complexity of a variation whose molecular basis is not known. That would simply be very bad scientific methodology.

    We can be quite sure that the mammalian middle ear is far too specific, complex, and irreducible, to be the result of random assembly.

    gpuccio: “Because protein sequences with specific functions are expected to be quite rare in protein sequence space, we prepared a DNA library of 4 x 10^14 independently generated random sequences. This DNA library was specifically constructed to avoid stop codons and frameshift mutations, and was designed for use in mRNA display selections.”

    That merely reduces the search space somewhat.

    gpuccio: and then the rounds of mutation + selection:

    Well, that’s the experiment. It starts with a minimal function, then the structure becomes more specific through iterative selection. This shows that functional proteins are not that rare in sequence space, and that there are selective incremental pathways to increased specificity.

    gpuccio: However, even with that effort, no really biologically useful protein was attained.

    It’s been shown that the artificial ATP-binding protein acted to bind ATP within a living cell, so it has biological function, even though it has a different sequence than the naturally occurring ATP-binding protein. In Hayashi et al., of course, the random sequence was inserted into the phage itself. Selection was due to differences in reproductive rate.

    ID After: I don’t believe that a really complex and functional protein like ATP synthase, for example, cab be engineered by this kind of methodology alone.

    ID Before: Enzymes fold into a very complex and highly specific three-dimensional shape. I don’t believe that a really complex and functional protein capable of catalytic activity can evolve through directed evolution. That’s because there are no incremental, selectable pathways. The combinatorial explosion prohibits it.

    gpuccio: In AS, the coupling between the defined function and the selection process is indirect and symbolic: the connection is established by the designer, by definite procedures designed by him.

    In Hayashi et al., the advantage is reproductive.

  45. 45
    Virgil Cain says:

    That is correct, as experiments with protein evolution show.

    Nonsense. Not one experiment with proteins demonstrates there is an unguided gradual pathway to producing CSI.

    Zachriel is being very dishonest with its posts.

  46. 46
    Alicia Cartelli says:

    “If I need 100 specific aminoacids to make something work (a case very common), then there is obviously no pathway which goes to that sequence step by step. Why? Because those 100 AAs are specific to the function.”
    Have any examples?
    Careful though, don’t confuse conserved amino acids with absolutely essential amino acids; that is rarely the case.
    Just because an amino acid is conserved, it does not mean that it is required for proper function. Swapping it out with another amino acid may reduce the efficiency of the enzyme to some degree, but function is still maintained.

  47. 47
    Virgil Cain says:

    Alicia Cartelli- Enough with your games. If you have a non-telic process for producing a functional protein consisting of 100 amino acids produce it so we can discuss it.

    cheers,
    Virgil Cain

  48. 48
    Mung says:

    gpuccio:

    Just for the record: I don’t discuss with Alicia Cartelli.

    For good reason.

  49. 49
    gpuccio says:

    Zachriel:

    Small and scarcely relevant verbal skirmishes, at this point.

    However, for what it is worth:

    In which case, experiments with optimizing selection are relevant.

    Experiments with optimizing selection are relevant to optimizing selection, not certainly to selection of a completely new sequence and function. Your obsession with conflating things is becoming really pathological.

    In the case of optimizing selection for a specific, naturally occurring function, what is the difference? What if selection is for reproductive advantage, as in Hayashi et al.?

    The difference is obviously that NS can only, at best, optimize a specific, already existing function, and only if the optimization is achieved by small variations, and only if it is such that it gives significant reproductive advantage. That’s a big difference, whatever you may say to obfuscate what is evident.

    The few bits of variation result in turning a random sequence into a specific and complex three-dimensional structure.

    The few bits of variation add some folding to a sequence already selected form a vast random library for a weak ATP binding site, so that the already existing affinity of the binding site is significantly improved. It’s not the same thing.

    Because they didn’t use recombination, of course.

    Recombination of what, please? Just throwing in words does not change things.

    Functional complexity exists in morphological space as well.

    ???

    That merely reduces the search space somewhat.

    It does, indeed.

    And if you quoted correctly, you would have said that I quoted that passage of the paper after the following statement:

    “Even the ATP binding example, indeed, is not a simple implementation of AS. It requires more: for example, the knowledge to build the initial library:”

    Are you denying that the designers of the experiment used some cognitive understanding of DNA and protein biology to build their initial library?

    Well, that’s the experiment. It starts with a minimal function, then the structure becomes more specific through iterative selection. This shows that functional proteins are not that rare in sequence space, and that there are selective incremental pathways to increased specificity.

    OK, and:

    a) The function is not naturally selectable, neither in its initial “minimal” form, nor in its final, engineered form. Therefore, the experiment is not about NS.

    b) This shows that minimal biochemical activities are not that rare in the protein space (whoever denied that?). And that AS and bottom-up protein engineering can incrementally increase the already present specificity which has been artificially selected in the beginning. And so?

    The experiment says nothing about the powers of natural selection. It says nothing about incremental pathways from one function to another completely different one.

    It’s been shown that the artificial ATP-binding protein acted to bind ATP within a living cell, so it has biological function, even though it has a different sequence than the naturally occurring ATP-binding protein.

    It has biological activity, which is not functional in the context of a cell. IOWs, it confers no advantages, either reproductive or else.

    And it definitely has a different sequence than the naturally occurring ATP-binding protein. I am happy that you are intelligent enough to realize that, because it seems that many of your fellow thinkers cannot realize such a simple fact.

    The naturally occurring ATP-binding protein has truly trivial biochemical activity. In a sense, it is better than the engineered one: at least, if introduced into a living cell, it should not create damage by subtracting ATP from the environment!

    Finally:

    ID before, after, and always:

    “Enzymes fold into a very complex and highly specific three-dimensional shape. I don’t believe that a really complex and functional protein capable of useful catalytic activity can evolve through unguided evolution. It requires design. That’s because there are no incremental, selectable pathways to new, complex protein functions. The combinatorial explosion prohibits it. Optimization of an existing function by small functional variation is in principle possible, if the variation is in the range of the probabilistic resources of the system. AS and bottom-up engineering are a form of design, and they can overcome, at least in part, some of the probabilistic barriers. However, for the most complex and efficient results, design in the form of top-down engineering is probably necessary.”

    You still say:

    In Hayashi et al., the advantage is reproductive.

    Correct. Indeed, those kinds of experiment, including the ragged landscape paper, are really about NS, something which is not true of Szostak’s esperiment. I have never said anything different.

    My comments about those experiments are different, and you have certainly read them in my posts #5 and #25.

  50. 50
    Dionisio says:

    @26
    Your writing is very clear. The problem could be your interlocutors’ unwillingness to understand it.

    @29
    Well done.

  51. 51

    Thanks for the article GP.

  52. 52
    Zachriel says:

    gpuccio: Experiments with optimizing selection are relevant to optimizing selection, …

    Something which occurs in nature.

    gpuccio: not certainly to selection of a completely new sequence and function.

    That’s exactly what happens with a new functional structure, as shown with Lenski’s Long-Term Evolution Experiment.

    gpuccio: The difference is obviously that NS can only, at best, optimize a specific, already existing function, and only if the optimization is achieved by small variations, and only if it is such that it gives significant reproductive advantage.

    As shown in Hayashi et al. It turns out that functional proteins are common enough that they can be found even in random sequences. With non-random sequences, such as recombinations of existing functional sequences, then they will be even more common.

    gpuccio: The few bits of variation add some folding to a sequence already selected form a vast random library for a weak ATP binding site, so that the already existing affinity of the binding site is significantly improved.

    The library is not so vast. Functional sequences are found once in about 10^11 sequences. There are hundreds of times that many bacteria in the average human gut. With non-random sequences, such as recombinations of existing functional sequences, then they will be even more common.

    gpuccio: Recombination of what, please?

    Genes. In genetic algorithms, it’s sometimes called crossover.
    https://en.wikipedia.org/wiki/Genetic_recombination

    See also exon shuffling.
    https://en.wikipedia.org/wiki/Exon_shuffling

    Zachriel: Functional complexity exists in morphological space as well.

    gpuccio: ???

    Not sure your question. There is a range of possible morphologies, the morphological space. There are functional and complex structures within that space.

    gpuccio: It does, indeed.

    Eliminating stop codons in random sequences only reduces the space by less than two orders of magnitude. Of course, nature doesn’t start with random sequences, but with recombinations of existing sequences, so that improves the search by far more than two orders of magnitude.

    gpuccio: Are you denying that the designers of the experiment used some cognitive understanding of DNA and protein biology to build their initial library?

    Did you know that Galileo built inclines to “guide” falling objects as part of an experiment on gravity? Of course he did! It was an experiment. What are the odds that his experiments would lead to valid generalizations about “unguided” falling objects? Lucky guesser!

    The question was how often are functional proteins found in random sequences. So they generated random sequences and tested them for function.

    gpuccio: The function is not naturally selectable, neither in its initial “minimal” form, nor in its final, engineered form. Therefore, the experiment is not about NS.

    In Hayashi et al., reproductive capability was the selection criterion.

    gpuccio: It has biological activity, which is not functional in the context of a cell.

    Showing benefit to the cell wasn’t the purpose of the experiment. It was to show that the artificial sequence would fold into a functional enzyme and bind ATP in a natural cell.

    gpuccio: IOWs, it confers no advantages, either reproductive or else.

    Hayashi et al. does show reproductive advantage.

    gpuccio: That’s because there are no incremental, selectable pathways to new, complex protein functions.

    Of course there are, as these experiments show.

    gpuccio: Indeed, those kinds of experiment, including the ragged landscape paper, are really about NS, something which is not true of Szostak’s esperiment.

    And yet Hayashi et al. shows the same result! Lucky guesser!

  53. 53
    gpuccio says:

    Zachriel:

    These are not even skirmishes anymore.

    I suppose that, when the two parties have said all that they had to say about something, the discussion inevitably goes down. You repeat your conflations, and I really have nothing more to say about that.

    I will only comment briefly on your last obsession, recombination: the new magic intended to save a dying theory.

    When I asked:

    “Recombination of what, please?”

    I was not looking for the generic and trivial answer you gave (genes). I was asking for a specific answer about the context we were discussing: the phage experiments.

    Let’s refer to this paper by Hayashi et al.

    “Experimental Rugged Fitness Landscape in Protein Sequence Space”

    http://journals.plos.org/ploso.....ne.0000096

    It is essentially a more detailed follow-up to the paper you quoted.

    So, let’s see. They replaced one of the three main domains in protein g3p of phage fd, the D2 domain, which contributes to infectivity. The D2 domain is about 180 AAs long, and they replaced it (or probably part of it) with a soluble random polypeptide, “RP3-42,” consisting of 139 amino acids.

    This is also the paper which states, in the discussion:

    “The question remains regarding how large a population is required to reach the fitness of the wild-type phage. The relative fitness of the wild-type phage, or rather the native D2 domain, is almost equivalent to the global peak of the fitness landscape. By extrapolation, we estimated that adaptive walking requires a library size of 10^70 with 35 substitutions to reach comparable fitness.”

    Then the authors add, always in the discussion:

    “Recombination among neutral or surviving entities may suppress negative mutations and thus escape from mutation-selection-drift balance. Although the importance of recombination or DNA shuffling has been suggested [30], we did not include such mechanisms for the sake of simplicity.”

    IOWs, they have nothing empirical in favor of the possible role of recombination in this context, a role which “has been suggested”.

    Just so stories? Fairy tales?

    Let’s try to understand. What we are trying to find here is a sequence of about 140 AAs (the replaced domain) which contributes to infectivity. After RV and NS, some optimization takes place, but we are still faraway from the wild type. Indeed, we have not found the wild type “island” at all, as the authors admit:

    “No convergence to the wild-type D2 domain was detected. The amino acid sequences of the clones picked randomly from the enriched population showed no significant homology to the wild-type sequence (Figure 2B). Based on detailed analysis of the fitness landscape described below, it is likely that the adaptive walk climbed to a different mountain in the fitness landscape from that where the wild-type sequence exists”

    And here comes the admission that a starting library of 10^70 would be needed to find the wild type sequence, and the generic statement that maybe recombination or exone shuffling could help.

    But recombination of what? I still ask.

    The original D2 domain, the wildtype sequence, is lost in the phage after the substitution. There are no more similar sequences in the phage genome, only completely different sequences of different genes, with no homology to what has been lost. What should be recombined, in order to help find the original sequence?

    Have you empirical experiments that show that recombination or exon shuffling can help in such a context? Which hexons, in the phage or elsewhere, would help, if shuffled, in finding the “mountain in the fitness landscape where the wild-type sequence exists”?

    I would appreciate answers, and not just verbal games. Thank you.

  54. 54
    Zachriel says:

    gpuccio: IOWs, they have nothing empirical in favor of the possible role of recombination in this context, a role which “has been suggested”.

    Population genetics demonstrates the importance of recombination, which is prevalent in everything from viruses to humans. Genetic algorithms can easily provide the structural support for recombination. Without recombination, each linage will find the first local peak and then stop. With recombination, it will explore much more of the landscape. It’s easy to show.

    This abstracted example of homologous recombination only shows the relevant areas of the gene. We start with

    xxxx

    Through simple mutation, end up with two strains on two local peaks.

    ABxx
    xxCD

    There’s a global peak,

    ABCD

    But ABxD, ABCx, AxCD, xBCD are all deleterious. (That’s what we mean by a local peak, there’s nowhere to step but down.) Therefore, there’s no simple mutational pathway from our two strains, ABxx or xxCD, to the global peak, ABCD.

    Recombination of the two strains will result in many variants that are not available to point-mutation alone. One of these recombinations is the global peak that was otherwise unattainable, ABCD.

    gpuccio: There are no more similar sequences in the phage genome, only completely different sequences of different genes, with no homology to what has been lost.

    There is homologous recombination with other variants of the same evolving gene, and nonhomologous recombination. Recombination is found in everything from viruses to humans.

  55. 55
    gpuccio says:

    Zachriel:

    I am not denying that recombination exists. I was asking if you had any empirical evidence that it can help in a scenario like the one we were discussing. I don’t think you have given any.

    Remember, the sequence found in the experiment had nothing in common with the sequence in the wildtype. Your simple schemes with a few letters will not help explain long and complex functional sequences.

    Moreover, when you discuss recombination, you should be able to compute all possible recombinations, because again here it is a problem of target space against search space. Recombination is not magic, any more than random walks are.

    Random walks have clearly shown, experimentally, their huge limitations, even in optimization issues. Recombination, when tested experimentally, will show similar huge limitations.

    However, you will obviously keep your faith. If and when you have some real support for it, please let me know.

  56. 56
    Zachriel says:

    gpuccio: I was asking if you had any empirical evidence that it can help in a scenario like the one we were discussing.

    1. Population genetics shows the importance of recombination.
    2. Evolutionary algorithms show the importance of recombination.
    3. Empirical evidence shows that importance of recombination in organisms as varied as viruses and humans.
    4. A simple example was provided to show how recombination can overcome local peaks.

    gpuccio: Remember, the sequence found in the experiment had nothing in common with the sequence in the wildtype.

    So? Why is that so unexpected? There may be many sequences that accomplish the same goal, or even if there is a single global peak, there may be many pathways to that global peak.

    gpuccio: when you discuss recombination, you should be able to compute all possible recombinations

    There’s no way to determine the search space by “calculating it”. At this point, it can only be explored. However, we do know that recombination avoids the problem of becoming locked on a local peak.

  57. 57
    gpuccio says:

    Zachriel:

    IOWs, you have no empirical evidence that recombination can help in a scenario like the one we were discussing.

    There may be many sequences that accomplish the same goal, or even if there is a single global peak, there may be many pathways to that global peak.

    You seem to forget that:

    a) The wildtype sequence was much more efficient then the sequences they evolved (its infectivity is still about 2000 times greater than the infectivity of the best evolved sequence).

    b) A sequence like the one found in the experiment, which has nothing in common with the sequence in the wildtype, can scarcely be considered “a pathway” to the wildtype, which remains by far the optimal peak, and the one we find in the natural phage.

    There’s no way to determine the search space by “calculating it”. At this point, it can only be explored.

    There are certainly ways to calculate the search space, obviously starting from empirical data which explore it. That’s the way to see if a theory can be shown false. It’s a way to do good science, and to abandon the realm of verbal games, of just so stories, and of fairy tales.

    Finally, it can be of some interest to know that even in the rugged landscape experiment, which certainly, as I have always said, has the merit of exploring correctly a NS scenario, there is a design component which certainly reduces the search space and makes the “random” sequences less random. Here is how the random sequences were prepared (from “Solubility of artificial proteins with random sequences”, http://www.sciencedirect.com/s.....9396001238 ):

    “In this work, we have prepared a library of 141 amino acid residue proteins with random sequences. The random sequences include the 20 kinds of amino acids. The state of the random proteins in the cells of Escherichia coli as to their solubility was examined. Out of 25 proteins examined, 5 were soluble. Hence, about 20% of the random proteins with 141 residues are expected to be soluble.”

    And:

    “The schematic diagram for library construction is illustrated in Fig. IA. The mixture of 140-met single-stranded oligonucleotides (Rl40ss) was synthesized by Toagosei Co., Ltd. (Tokyo) according to our design. RI40ss contains a randomized portion composed of 6 repeated 16-mer random oligonucleotides flanked by fixed sequences which contain the primer sites for amplification and the restriction
    enzyme sites.”

    And:

    “The genes encoding the artificial random proteins were designed with the following criteria: (1) all the 20 kinds of amino acids are included; (2) the length of the randomized portion is about 100 amino acid residues; (3) the amino acid sequence is highly random; and (4) the mean value of the net charge of the random proteins is about +2. The above criteria were met by the synthesized randomized portion of RI40ss and the strategy of constructing the gene (Fig. l). It should be pointed out that no stop codons appear in all the six frames of the randomized portion even if frame shifts occur during the synthesis and construction, and that the mean value of the G + C content of Rl40ss is set to be 53.5%, as high G + C content interferes with PCR reactions.”

    Emphasis mine.

    And, from the first Hayashi paper (Can an arbitrary sequence evolve towards acquiring a biological function?):

    “we replaced the D2 domain of the fd-tet phage genome with the soluble random polypeptide RP3-42.”

    Emphasis mine.

  58. 58
    EugeneS says:

    The toy examples suggested by Zachriel assume that since we can walk, we can walk to the Moon, given enough time. What about scenarios where there is NO landscape at all around a peak? I.e. where no recombination or mutation selectably leads from one group of peaks isolated by chaos to another.

  59. 59
    gpuccio says:

    EugeneS:

    You are right. Zachriel’s “examples” are abstract toys, completely out of context. He is probably a good expert of algorithms, especially with words, but I am afraid that he is less familiar with biological contexts. Not his fault, however.

    The simple fact is that neo darwinism (or its personal variants) badly needs vague explanations. So, random walk and genetic drift along imaginary pathways is a neo darwinist star, until experiments show that it cannot even really optimize a damaged domain in an existing, still functional protein. Then when numbers like 10^70 come to the attention (for once, not because of IDist plots), some new magic is needed, something that has not yet been really tested in a biological context, and can therefore be vague enough to satisfy the true believers. So, recombination is ready to fill the void.

    OK, we are here to patiently witness all these games. Luckily, in the meantime true biology makes ever new discoveries of unending functional complexity (see epigenetics and cell differentiation), and this ugly nightmare of biased cognition will be over, sooner or later.

  60. 60
    EugeneS says:

    GP,

    Absolutely. Evolutionism as any other type of reductionism suffers from its ancestral diseases, so to speak. They claim that biology must be reducible to chemistry, chemistry to physics. But in the real world, there are huge problems with reducibility: combinatorial problems are irreducible to polynomial time problems, semiotic phenomena (biology included) – to physicality, information – to mass/energy, consciousness – to matter.

  61. 61
    Virgil Cain says:

    1. Population genetics shows the importance of recombination.

    Yes, recombination is an important design feature.

    2. Evolutionary algorithms show the importance of recombination.

    Evolutionary algorithms exemplify evolution by DESIGN.

    Zachriel is too dim to grasp any of that.

  62. 62
    Mung says:

    Zachriel: 1. Population genetics shows the importance of recombination.

    Does population genetics show the importance of lateral gene transfer and symbiosis?

    Zachriel: 2. Evolutionary algorithms show the importance of recombination.

    Not every EA uses recombination. Are you claiming an EA always performs better if you introduce recombination?

  63. 63
    Zachriel says:

    gpuccio: you have no empirical evidence that recombination can help in a scenario like the one we were discussing.

    We provided four lines of evidence concerning the effectiveness of recombination in scenarios like the one we were discussing. We did not provide evidence of the effectiveness of recombination in the exact scenario because the researchers did not include that in their study.

    gpuccio: You seem to forget that: a) The wildtype sequence was much more efficient then the sequences they evolved

    Didn’t forget it. In fact, we addressed it directly. Lack of recombination leaves evolution stuck on local peaks.

    gpuccio: b) A sequence like the one found in the experiment, which has nothing in common with the sequence in the wildtype, can scarcely be considered “a pathway” to the wildtype

    That can’t be known based on the research. Recombination is such that offspring are usually unique.

    gpuccio: “we replaced the D2 domain of the fd-tet phage genome with the soluble random polypeptide RP3-42.”

    Solubility doesn’t significantly change the denominator, so it doesn’t change the overall findings. Keep in mind that, in nature, the starting point is probably a duplicate or fragment of an existing sequence, so the denominator is much lower than in a random sequence, even if we assume solubility.

    EugeneS: The toy examples suggested by Zachriel assume that since we can walk, we can walk to the Moon, given enough time.

    No. The example is much more limited. It shows 1) how simple mutation can be stuck on local peaks; 2) how recombination can overcome local peaks to find the global peak.

    EugeneS: What about scenarios where there is NO landscape at all around a peak? I.e. where no recombination or mutation selectably leads from one group of peaks isolated by chaos to another.

    Then that peak will probably never be found by evolutionary search (keeping in mind that actual evolution includes a lot more flexibility than the toy example).

    Turns out, though, that the natural landscape is not chaotic in that sense, but highly ordered.

    gpuccio: Zachriel’s “examples” are abstract toys, completely out of context.

    The toy example shows 1) how simple mutation can be stuck on local peaks; 2) how recombination can overcome local peaks to find the global peak.

    Mung: Does population genetics show the importance of lateral gene transfer and symbiosis?

    Classical population genetics didn’t, but modern population genetics has incorporated these mechanisms. Computer simulations, in conjunction with empirical evidence, are now the primary means of exploring population genetics.

    Mung: Not every EA uses recombination. Are you claiming an EA always performs better if you introduce recombination?

    Most evolutionary algorithms use some sort of crossover, because most complex spaces require recombination for an extensive search.
    https://en.wikipedia.org/wiki/Crossover_(genetic_algorithm)

  64. 64
    Virgil Cain says:

    Evolutionary algorithms still exemplify evolution by DESIGN.

  65. 65
    EugeneS says:

    Zachriel,

    “Then that peak will probably never be found by evolutionary search…”

    You are right. It won’t be found in that case.

    “…keeping in mind that actual evolution includes a lot more flexibility than the toy example”

    I can’t agree with this though. Flexibility means control. There is minimum control in blind evolutionary search. The control is effectively binary: survive or die. Flexibility, IMO, assumes a lot more than that. You seem to always conflate the capabilities of artificial selection with those of natural selection.

  66. 66
    Zachriel says:

    Eugene: Flexibility means control.

    Flexible means capable of bending. Many mutations to genomes cause little or no selective change, hence genomes are considered flexible.

    Eugene: The control is effectively binary: survive or die.

    That is incorrect. Most natural selection is due to small differences in reproductive potential.

  67. 67
    Mung says:

    Zachriel: Most natural selection is due to small differences in reproductive potential.

    So small that “Mother Nature” can’t tell the difference.

  68. 68
    Zachriel says:

    Mung: So small that “Mother Nature” can’t tell the difference.

    Some changes are below the effect of natural selection, depending on population size. Some changes are subject to natural selection, most of which cause small differences in reproductive potential.

  69. 69
    gpuccio says:

    Zachriel:

    However, thank your for your contributions. They are appreciated.

    I think I have said more or less all that I had to say on the points you offered. I hate repetition, so that’s it.

    If you have any biological papers regarding the role and powers of recombination, I would be interested.

  70. 70
    Dionisio says:

    gpuccio @59

    […] in the meantime true biology makes ever new discoveries of unending functional complexity (see epigenetics and cell differentiation) […]

    🙂

  71. 71
    Dionisio says:

    gpuccio @69

    If you have any biological papers regarding the role and powers of recombination, I would be interested.

    I agree. Perhaps that should be a requirement for any serious discussion. It could be named “the gpuccio rule”: biology-related arguments should be supported with references to specific papers that can stand thorough review by anyone.

  72. 72
    Zachriel says:

    gpuccio: If you have any biological papers regarding the role and powers of recombination, I would be interested.

    How about an entire issue of the journal Viruses dedicated to recombination: “Recombination is an important source of genetic variability in viruses”

  73. 73
    gpuccio says:

    Zachriel:

    OK, thank you. I will look into it.

  74. 74
    gpuccio says:

    Zachriel:

    OK. I have read, briefly, the introduction to that issue and the 6 papers in the issue. Here are my comments.

    1) Recombination does occur in viruses, although for many types of them it seems to be an uncommon event. In other types, it is very frequent.

    2) Recombination usually occurs between different strains or subtypes of the same virus, and even between different, but similar viruses. The most common scenario for recombination events is the infection by two different types or subtypes of viruses.

    3) Recombination can be homologous (between homologous genes), or non homologous.

    4) In many cases, recombinants are non vital, or show less fitness than the parental strains. In other cases, fitness is not affected.

    5) The most important effect of recombination is to increase genetic heterogeneity and diversity. In that sense, it acts like mutations. Recombination generates different mosaics of viral genes in different strains.

    6) Obviously, that has important consequences regarding vaccine sensitivity, virulence against specific hosts, resistance to therapies, and phylogenetic studies of viruses.

    7) What I have not found is any indication, in any of the papers, that gene recombination may have any helping power in the generation of new functional genes. Its role seems to be simply to redistribute existing genes, and mix them somewhat. While the consequences of that can certainly be of great relevance, there seems to be absolutely no relevance of these phenomena to the generation of complex functional information. All of them seem to be examples of molecular microevolution, and none of them seems to be implied in scenarios like the one we have discussed previously.

    There is no doubt that viruses are probably the best scenario for RV + NS: their rate of variation is astounding, and they are probably engineered to change as much as possible and to gain as much as possible from those changes. That is probably implicit in their basic program.

    However, even in that incredibly favorable scenario, the limitations of RV + NS are obvious, and nothing goes beyond simple microevolutionary events, whose functional complexity is very low.

  75. 75
    gpuccio says:

    Zachriel:

    Of course, if what you mean is that the defective phage in the Hayashi paper could easily retrieve the wildtype efficiency by recombination if it recombines with a normal phage, which has retained the wildtype sequence, well, that is obvious! Recombination certainly can do that! It can shuffle existing information, and that’s more or less what it can do.

    But I don’t think that that is what you meant.

    Another comment: in your toy examples, you always reason about recombinations of small sequences, like a couple of letters. But that’s not what usually happens in biology. As you can see in the quoted papers, recombination usually implies long sequences of DNA. The so much invoked exon-shuffling, for example, implies more or less whole exons.

    There is an important consequence of that: recombination of long sequences is bound to be recognizable because of the implicit homology: we can recognize what was recombined, and how. That’s exactly how recombination is studied and detected.

    Small variations of a couple of nucleotides are more easily explained by simple mutations, including indels.

  76. 76
    Zachriel says:

    gpuccio: 5) The most important effect of recombination is to increase genetic heterogeneity and diversity. In that sense, it acts like mutations. Recombination generates different mosaics of viral genes in different strains.

    Recombination allows for variants not available to simple mutation, including the evolution of new strains. A schematic example was provided above that should be clear enough.

    gpuccio: 7) What I have not found is any indication, in any of the papers, that gene recombination may have any helping power in the generation of new functional genes.

    The discussion wasn’t of new genes, but of how genes are optimized in a complex landscape. With simple point mutation, genes tend to become fixed on local fitness peaks. With recombination, much more of the landscape can be explored. Consequently, recombination is an important mechanism of genetic diversity.

    As for new genes, we know that even random sequences can have function, so duplicates and fused fragments of old genes clearly can have function. There are a number of known mechanisms for the creation of new genes. See this review article for a discussion; Long et al., The origin of new genes: glimpses from the young and old, Nature Reviews 2003.

  77. 77
    gpuccio says:

    Zachriel:

    The wildtype of our discussion is an isolated sequence, which has not been found and cannot be reasonably found by mutation and NS. You think that recombination with unknown genes, which have nothing to do at sequence level with the wildtype itself, should help, but you have given no reason to believe that.

    The simple reason why the wildtype, the true optimal peak, cannot be found is because it is too small and isolated, and the random search is not powerful enough to find it. Recombination is a form of random variation too, unless part of the functional sequence to be found is already present in the genes which recombine.

    Therefore, unless you recombine with phages which still have the wildtype gene, I can’t see how recombination can help in that scenario, and you have given no reasons to believe that.

    The simple truth is that the highly functional wildtype is not an optimization of the random sequences that were used in the experiment: there is no convergence towards the wildtype at the sequence level, as the authors state clearly.

    You seem to forget too often that all the variation happens at sequence level in the genome. The search is a search for a sequence, the sequence which bears the optimal function. If that sequence is small enough and isolated enough, that peak will simply not be found by random variation, of whatever kind, including recombination.

    And you have shown no experimental data where recombination really helps to find a functional result which eludes simple mutations. That would be some support to your ideas. Not certainly the papers about the role of recombination in virus diversity.

    Nobody denies that “recombination is an important mechanism of genetic diversity.”. That is simply obvious.

    The point is, and always has been: is recombination an important mechanism in the random search for complex functional sequences?

    I believe that the answer is definitely: no.

  78. 78
    Dionisio says:

    gpuccio @74, 75, 77

    Very insightful comments.

  79. 79
    Zachriel says:

    gpuccio: The wildtype of our discussion is an isolated sequence, which has not been found and cannot be reasonably found by mutation and NS.

    That’s your claim.

    gpuccio: You think that recombination with unknown genes, which have nothing to do at sequence level with the wildtype itself, should help,

    That is incorrect. In the particular case, homologous recombination would probably be sufficient. What the study showed was that the landscape has a huge number of local peaks, so that simple mutation starting from random sequences can only explore a small portion of the landscape.

    gpuccio: but you have given no reason to believe that.

    1. Population genetics shows the importance of recombination.
    2. Evolutionary algorithms show the importance of recombination.
    3. Empirical evidence shows that importance of recombination in organisms as varied as viruses and humans.
    4. A simple example was provided to show how recombination can overcome local peaks.

    gpuccio: The simple reason why the wildtype, the true optimal peak, cannot be found is because it is too small and isolated, and the random search is not powerful enough to find it.

    It’s probably not isolated in the multi-dimensional recombination space. We don’t know, because the study didn’t include recombination, but we do know that recombination works like this in other cases.

    gpuccio: The point is, and always has been: is recombination an important mechanism in the random search for complex functional sequences?

    Even point mutation resulted in a complex functional sequence.

  80. 80
    gpuccio says:

    Zachriel:

    Again, I will not repeat what has already been said.

    I want to comment on your last statement, because I believe you are definitely wrong (or at least vague).

    You say:

    “Even point mutation resulted in a complex functional sequence.”

    OK, if we define retrieval of infectivity, then certainly the use of random libraries of growing number + RV and NS did result in one or more functional sequences.

    But the point is: how complex?

    What the study shows, very clearly IMO, is that low levels of function retrival have some complexity, but not a very high complexity. Indeed, they are easily found by the system exactly for that reason: because the target space is big enough that it can be found by the system.

    IOWs, gross function, at a very low level, has less functional complexity.

    But the level of function provided by the wildtype sequence has much higher complexity, so much so that huge combinatorial resources are needed to find it (a 10^70 sequence library).

    You should know that in ID we quantify functional (specified) information. That a sequence with low dFSCI can be found by a highly efficient RV + NS system like the phage system is simply what we can expect.

    What cannot be found by such a system is a sequence with very high dFSCI, beyong a threshold which makes the probabilistic resources of a natural system completely powerless.

    The wildtype sequence seems to be in that range, if the authors are right in their conclusions. After all, they anticipate about 35 AA substitutions to get to it, and, if true, that would be a functional complexity of the order of magnitude of 150 bits, enough to frustrate any natural system.

    And yet, that would still be much less than what we observe in a lot of functional proteins, like the many times quoted alpha and beta subunits of ATP synthase and histone H3.And many others. Those examples are definitely beyond any cosmic natural system.

    So, remember, ID is a quantitative theory. Affirming that “a complex functional sequence” has been found is vague. You have to try to evaluate the functional complexity.

  81. 81
    Alicia Cartelli says:

    Pucci, you’re still confusing “conserved amino acids” with “amino acids that are required for function,” though.
    =)

  82. 82
    Dionisio says:

    gpuccio @80

    Apparently your interlocutors don’t understand the basic concept of complex complexity.

    🙂

  83. 83
    Mung says:

    Pucci, you’re still confusing “conserved amino acids” with “amino acids that are required for function,” though.

    I’ll start doubting him when he denies the existence of the peptidyl transferase center.

  84. 84
    Alicia Cartelli says:

    Still going on about that Mungy?

    And still finding ways to be wrong.
    I doubted the existence of a peptidyl transferase enzyme that “adds amino acids to a growing strand,” and which isn’t called “the ribosome.”
    And I still doubt it, because you are yet to come up with an enzyme that fits these requirements.
    There is only one thing that fits these requirements and it is “the ribosome,” which is technically not an enzyme, it is a ribozyme.
    When will you learn?

    Let me know when someone (like EA) comes up with something about my response to “EA’s challenge.”

  85. 85
    Zachriel says:

    gpuccio: But the point is: how complex?

    A complex three-dimensional structure, but not as highly specified as the native form.

    gpuccio: But the level of function provided by the wildtype sequence has much higher complexity, so much so that huge combinatorial resources are needed to find it (a 10^70 sequence library).

    They didn’t test recombination, and there are reasons to expect — as the author’s pointed out — that recombination would result in much higher specificity.

  86. 86
    gpuccio says:

    Zachriel:

    OK, so we agree that “how much complex a structure is” is an important point. Which is a central concept in ID.

    And let me know if and when somebody tests recombination in a similar context.

  87. 87
    Zachriel says:

    gpuccio: And let me know if and when somebody tests recombination in a similar context.

    There are a plethora of papers on recombination. For a review, see Long et al., The origin of new genes: glimpses from the young and old, Nature Reviews 2003. For something more specific, try Gomez, Creating New Genes by Plasmid Recombination in Escherichia coli and Bacillus subtilis, Applied and Environmental Microbiology 2005.

    Of course, the importance of recombination for exploring rugged landscapes is easily demonstrable with evolutionary algorithms, or even with simple abstractions.

Leave a Reply