Monday, July 30, 2007

Another Bad Review of The Edge of Evolution

PZ Myers draws attention to another review of Michael Behe's new book The Edge of Evolution [Behe gets another thumbs-down]. This review is published in the July issue Discover magazine [The Simplistic Manifesto]. The author is Cory S. Powell.

I disagree with PZ. This is not a good review. Actually, it's a very bad review. Like many of the published reviews of The Edge of Evolution the author seems to have been reading a far different book than the one I read. Powell says,
To reach this conclusion, Behe makes a number of invalid assumptions about how molecules evolve and interact. He alleges that, because many functional adaptations require multiple changes in proteins, two or more mutations must occur together at the same time in the same gene and only rarely can several mutations "sequentially add to each other to improve an organism’s chances of survival." But in fact natural selection does work on transitional forms, as molecules and traits evolve stepwise. Stepwise evolution has been well documented; one good instance of this is the emergence of color vision. Mutations add up little by little, leading to major changes to proteins over time.
The essence of Behe's argument is not that it's impossible to evolve a double mutation if each one is beneficial. The whole point of the book is that stepwise evolution requires that each step is beneficial. The evidence, according to Behe, shows that many cases involving double mutations involve intermediates that are disadvantagous. Thus, the double mutant had to arise in a single step and this is highly unlikely.

Behe isn't always as clear as he should be but he does make it perfectly clear that he accepts the mechanism that Powell describes. Thus, Powells' criticism is inappropriate and this makes it a bad review. Apparently Powell didn't read the section on the evolution of antifreeze proteins in fish (pp. 77-81) where Behe describes each of the many steps that lead to the modern antifreeze proteins.

Each step would have given the fish some protection against freezing water. Thus, Behe concludes,
Even though we haven't directly observed it, the scenario seems pretty convincing as an example of Darwinian evolution by natural selection. It's convincing because each of the steps is tiny&mdash'no bigger than the step that yielded the sickle cell mutation n humans—and each step is an improvement.
The Discovery review points out that complex combinations of mutations can arise in a stepwise manner by standard Darwinian mechanisms. It implies that Behe never thought of this in his book but that's total nonsense. Of course he did. Behe doesn't deny that phenotypes requiring multiple steps can be produced by random mutations, as long as each step is beneficial. The essence of his argument is that it's impossible to generate phenotypes that require multiple random mutations if the intermediates aren't beneficial.

I'm not arguing that Behe is correct. In fact, I'm preparing a series of postings that will challenge some of his ideas. What I'm objecting to is the mischaracterization of Behe's arguments in many of the published reviews. If you're going to criticize Behe then challenge the argument he makes in the book; namely, that most stepwise pathways are impossible because the intermediates are less fit than their parents.

Powell makes another common mistake in his review. He says,
Behe makes another big, related error in the way he interprets how proteins work together. He contends that for even three proteins to evolve in a cooperative association is wildly improbable, "beyond the edge of evolution." Within a protein, five or six amino acids (components of proteins) need to change simultaneously for it to bond with another protein, according to Behe. From this he concludes that it is impossible for proteins’ interaction to evolve, again requiring life to have been programmed for success from the start. Plenty of evidence contradicts this assertion, however. Many proteins within cells interact with other proteins in ways in which only two or three amino acids are critical for binding.
Behe admits that you may only need three or four selected changes in order to generate a new binding site (p. 114). He agrees that the evolution of a single binding site is within reach of evolution but the simultaneous generation of two binding sites is beyond the edge of evolution because the probabilities are so low. The point is that there are many complexes that require the interaction of several different proteins and the intermediates—where only two proteins interact—are not beneficial. Refuting Behe's real arguments requires a little more effort than the superficial criticism of arguments that Behe is not making.

Powell continues,
Such simple binding sites can arise frequently in proteins. And such interactions form the networks that regulate all sorts of physiological processes in cells and organisms. Cell biologists and biochemists are increasingly finding that, in truth, protein interactions and networks are easy to evolve. Behe should know this—but he has a long history of alleging evolutionary impossibilities and ignoring the scientific literature.
Powell is completely missing the point here. Behe does not deny that such complexes exist, nor does he deny that they evolved in the sense that they arose in organisms whose ancestors didn't have them. Once the mutations occurred, they became fixed in the population by natural selection. Furthermore, Behe does not deny that these networks are "easy to evolve." In fact, they are so "easy to evolve" that they cannot be explained by natural selection acting on random mutations as "Darwinism" requires. Thus, mutations cannot be random.

You don't refute Behe by pointing to examples of evolution by common descent or natural selection; this includes evolution of protein complexes. That's not the point. The point is that "Darwinian" evolution, according to Behe, must require small steps where each step is beneficial and this cannot be demonstrated. Indeed, in many cases the intermediates will likely be detrimental. The conclusion is that multiple mutations have to occur simultaneously as in some drug resistance. For most populations the probability of this happening by random mutation is very small. The fact that it happened is evidence of directed mutation, or so Behe thinks.

In order to show that Behe is wrong you have to demonstrate that his understanding of evolution (i.e., "Darwinism") is wrong and this has led him to false conclusions about probabilities. Many reviewers have failed to do this, possibly because they accept Behe's version of Darwinism.

You can read Michael Behe's responses to his reviewers on the Amazon.com site [Michael Behe's Amazon Blog]. I think it's fair to say that Behe makes some good points (and many bad ones) when he accuses his reviewers of misrepresentation.

18 comments:

  1. Larry,

    The problem is that Behe's argument has two steps:

    1. Binding sites require multiple simultaneous mutations (no beneficial intermediates)

    2. If #1 is the case, then they are too improbable for natural processes and must have been guided by ID.

    But almost all of Behe argumentation is about #2, and #1 is basically just assumed rather than argued for. This kind of thing is endemic throughout ID writing.

    So it is perfectly legit to point out that Behe can't just assume #1. If Behe were serious he would spend his whole book just arguing #1, which is the controversial premise. But instead it's mostly #2.

    ReplyDelete
  2. Behe doesn't deny that phenotypes requiring multiple steps can be produced by random mutations, as long as each step is beneficial. The essence of his argument is that it's impossible to generate phenotypes that require multiple random mutations if the intermediates aren't beneficial.

    Yeah, Behe has retreated from his original conception of irreducible complexity to his second, "evolutionary" definition phrased in terms of unselected steps (see here for the definitions).

    Not too long ago I did a re-analysis of the supplementary data published by Lenski, et al., in support of their 2003 Nature paper using Avida that showed that IC structures (in Behe's first 'structural' definition) can evolve easily with only random mutations and selection operating. I traded back the history of every component of the final IC structure to determine whether it was selectively advantageous, neutral, or deleterious when it first occurred via a mutation. Of the structural IC components of the 'final' structure, 28% were either selectively neutral or in one case actually deleterious when they first occurred via an insertion or substitution mutation. That is, structures that are IC by Behe's first definition can easily evolve via pathways that are IC by Behe's second definition because neutral pathways are available in the space of potential pathways. There's a summary of that re-analysis here.

    RBH

    ReplyDelete
  3. Erm, make that "traced back the history of every component" up there.

    ReplyDelete
  4. Nick says,

    So it is perfectly legit to point out that Behe can't just assume #1. If Behe were serious he would spend his whole book just arguing #1, which is the controversial premise.

    Behe spends eight pages arguing that you need to change several amino acids in order to get a decent binding site (pp. 127-134) and he quotes experimental results that appear to support his argument. It's also soundly supported by theoretical arguments since the binding energy of a single amino acid side chain is far too small to bring two proteins together.

    The experimental results show that you have to examine about ten million different binding sites in order to get one that works. Since each amino acid position in a potential binding site patch can be occupied by 20 different amino acids, this means that you need to change five or six amino acids in order to generate this many binding sites (20^5 is about 3 million).

    It is NOT perfectly legit to dismiss his reasoning out of hand just because you don't like it. You have to show why his reasoning is wrong and why the experimental results don't mean what he thinks they mean. Do you know of any review that has done this?

    rbh (above) is on the right track. The real problem is Behe's flawed understanding of evolution and it's a flaw that's shared by many evolutionists. Behe thinks that a mutation has to be beneficial in order to be fixed in a population. He doesn't allow for (significant) neutral mutations and that's what screws up his calculations (but see 114).

    ReplyDelete
  5. "If you're going to criticize Behe then challenge the argument he makes in the book; namely, that most stepwise pathways are impossible because the intermediates are less fit than their parents."

    I'm just a lay person, and fairly new at this, to boot, but isn't a problem with his argument that he assumes that the environment against which "fitness" is measured is a constant?

    IOW, "less fit" for this generation might just as well be "more fit" for a generation or two down the line. And vice versa.

    Or am I 'way off here?

    ReplyDelete
  6. Behe spends eight pages arguing that you need to change several amino acids in order to get a decent binding site (pp. 127-134) and he quotes experimental results that appear to support his argument.

    No, most of what Behe does is argue the premise "most binding sites consist of several key amino acids." This does not establish the claim "you have to change several amino acids at once to get a binding site". E.g., it leaves out preadaptation, where several of the right amino acids for binding are already sitting around in the wild-type proteins, and where all it takes is one new change, plus the chance presence of useful neighboring members, to get a weak-but-selectable binding site.

    The problem with Behe's argument is very similar to the misdirection of Behe's original IC argument, where he conflates "multiple required parts" with "no gradual pathway."

    Knocking amino acids out of a functional binding site does not simulate evolution in reverse, any more than does knocking proteins out of a modern flagellum.

    It's also soundly supported by theoretical arguments since the binding energy of a single amino acid side chain is far too small to bring two proteins together.

    Again, this assumes a bogus model of reality, as if binding sites evolved by one amino acid sticking out of a glass sphere, then adding another, etc. In reality any new binding site will involve multiple amino acids, but there is no reason to think that all of those amino acids are new mutations. Some could well just be native residues that happen to contribute to the binding -- by themselves, not enough to constitute a binding site, but in the presence of the new key point mutation, they help out.

    If you think Ian Musgrave knows what he's talking about, and I do, all these proteins in their native, unmutated states are already sticking to each other to varying degrees without having what we call "binding sites". So Behe mistakenly thinks that mutation has to go from zero binding to a perfect, strong binding site, whereas in reality evolution just has to go from the default interactions and stickiness to an interaction somewhat stronger that is then selected and refined. Are you seriously going to argue that this latter situation is always beyond the reach of a single point mutation?

    (I am not saying that this scenario would always produce a binding site in any evolutionary situation -- all that has to happen is that this works occasionally amongst the zillions of combinations of species/proteins/targets that are available.)

    (Another fallacy is to assume that "binding" is a binary state. We all know it is quantitative and that mutations can move it up or down.)

    The experimental results show that you have to examine about ten million different binding sites in order to get one that works. Since each amino acid position in a potential binding site patch can be occupied by 20 different amino acids, this means that you need to change five or six amino acids in order to generate this many binding sites (20^5 is about 3 million).


    These studies involve drastic multi-site mutations of sequences, many of which destroy the native shape/structure/solubility of the protein. Plus there are just few rounds of selection used (like 3) and so mostly they are trying to hit a binding site by brute force of numbers.

    The binding site that results is just the strongest "winner". This does not provide an estimate of the number of weak-but-selectable binding sites or how often they occur.

    It is NOT perfectly legit to dismiss his reasoning out of hand just because you don't like it. You have to show why his reasoning is wrong and why the experimental results don't mean what he thinks they mean. Do you know of any review that has done this?


    Have you read the one I submitted to TREE, which I emailed around awhile back? Gimme a break, I have spent as much time rebutting Behe on his own terms as anybody...

    rbh (above) is on the right track. The real problem is Behe's flawed understanding of evolution and it's a flaw that's shared by many evolutionists. Behe thinks that a mutation has to be beneficial in order to be fixed in a population. He doesn't allow for (significant) neutral mutations and that's what screws up his calculations (but see 114).


    I don't disagree with this, but Behe thinks (a) the AVIDA-type computer simulations are irrelevant, and (b) that Behe-Snoke 2004 dealt with the accumulation of neutral mutations possibility. Search PT for the critiques of both Behe-ish arguments: http://www.pandasthumb.org/archives/2004/10/theory_is_as_th.html

    One of the key points was that the 3 "key" amino acids Behe/Snoke thought were required for DPG binding were in actuality not always required.

    ReplyDelete
  7. Thanks for this post Larry, I too have the feeling that some people may be hyping particular reviews simply on the basis that these reviews slam (rightly or wrongly) Behe.

    I've been looking at one particular aspect of Behe's response to Sean Carroll and wonder if Behe might have a point. Specifically, this passage:

    In his enthusiasm Carroll seems not to have noticed that, as I discuss at great length in my book, no protein binding sites — neither short linear peptide motifs nor any other — developed in a hundred billion billion (1020) malarial cells. Or in HIV. Or E. coli. Or in human defenses against malaria, save that of sickle hemoglobin. Like Coyne, Carroll simply overlooks observational evidence that goes against Darwinian views. In fact, Carroll seems unable to separate Darwinian theory from data. He writes that “what [Behe] alleges to be beyond the limits of Darwinian evolution falls well within its demonstrated [my emphasis] powers”, and “Indeed, it has been demonstrated [my emphasis] that new protein interactions (10) and protein networks (11) can evolve fairly rapidly and are thus well within the limits of evolution.”

    Yet if one looks up the papers he cites, one finds no “demonstration” at all. Those papers show, respectively, that: A) different species have different protein binding sites (but, although the authors assume Darwinian processes, they demonstrate nothing about how the sites arose); or B) different species have different protein networks (but, again, the authors demonstrate nothing about how the networks arose). Like Jerry Coyne, Sean Carroll simply begs the question. Like Coyne, Carroll assumes whatever exists in biology arose by Darwinian processes. Apparently Darwinism has eroded Coyne’s and Carroll’s ability to separate data from theory.


    I checked out references 10 and 11 a while back. I only briefly scanned them and still haven't gotten round to a full review. They can be found here, with 10 being:

    http://www.pnas.org/cgi/content/full/102/39/13933

    and 11 being

    http://compbiol.plosjournals.org/perlserv/?request=get-document&doi=10.1371%2Fjournal.pcbi.0030025

    My initial feeling is that Behe is right about the PNAS article. There doesn't appear to be anything about how protein interactions arose (though my limited understanding could be at fault here). I'm less sure about Behe's dismissal of the PLOS article. They do provide evidence for rates of change in protein networks, but not much in the way of mechanism (though there is some material at the end).

    Has anyone else taken a look at these?

    ReplyDelete
  8. Nick,

    I know that you understand the flaws in Behe's argument and I agree with most of what you say. Behe doesn't understand how evolution actually works.

    My point is that if you're going to criticize Behe then that's the point you have to make. You need to explain in a review that Behe doesn't understand neutral mutations and pre-adaptation.

    Did you see anything in Powell's review to suggest that he (Powell) understands the fatal flaw? I didn't. In fact, it seems to me that Powell accepts "Darwinian" theory since he mentions it without adding a disclaimer.

    ReplyDelete
  9. stevef,

    I agree that Behe has a point. Behe doesn't disagree with the view that complex things have evolved so it's a waste of time for reviewers to bring that up. Whenever they do so they demonstrate that they haven't read Behe's book carefully.

    Behe agrees that proteins with several binding sites have evolved. His argument is that they can't have evolved by selection of random mutations because the probabilities are astronomical. Therefore, the fact that they have evolved is evidence of nonrandom mutations. It's proof that God exists.

    In order to refute Behe you have to demonstrate that complex things can arise from random mutations. In other words, you have to show that evolutionary theory ("Darwinism" in Behe's terminology) is capable of explaining evolution. That's not as easy as most people think.

    I'm embarrassed by most of the reviews I've seen. They make Behe look good when he says that evolutionists have a dogmatic approach to their science and don't take the time to think about the implications of their theory.

    ReplyDelete
  10. The problem is not with Behe, it's with the evolutionists that believe that evolution is a science.

    Many people, when they can't provide evidence for their theory, adopt the strategy of falsehood. Such is the case with many of those who have fallen victim to the propaganda of renowned evolutionists.

    If evolutionists want to end the arguments all they have to do is, get their brilliant heads together and assemble a 'simple' living cell. This should be possible, since they certainly have a very great amount of knowledge about what is inside the 'simple' cell.

    After all, shouldn't all the combined Intelligence of all the worlds scientist be able the do what chance encounters with random chemicals, without a set of instructions, accomplished about 4 billion years ago,according to the evolutionists, having no intelligence at all available to help them along in their quest to become a living entity. Surely then the evolutionists scientists today should be able to make us a 'simple' cell.

    If it weren't so pitiful it would be humorous, that intelligent people have swallowed the evolution mythology.

    Beyond doubt, the main reason people believe in evolution is that sources they admire, say it is so. It would pay for these people to do a thorough examination of all the evidence CONTRARY to evolution that is readily available: Try answersingenesis.org. The evolutionists should honestly examine the SUPPOSED evidence 'FOR' evolution for THEMSELVES.

    Build us a cell, from scratch, with the required raw material, that is with NO cell material, just the 'raw' stuff, and the argument is over. But if the scientists are unsuccessful, perhaps they should try Mother Earth's recipe, you know, the one they claim worked the first time about 4 billion years ago, so they say. All they need to do is to gather all the chemicals that we know are essential for life, pour them into a large clay pot and stir vigorously for a few billion years, and Walla, LIFE!

    Oh, you don't believe the 'original' Mother Earth recipe will work? You are NOT alone, Neither do I, and MILLIONS of others!

    ReplyDelete
  11. Ah, the troll is back. And I'm SURE he'll be back again to eat crow in a decade or so when that very feat is successfully accomplished. Not. (Of course it has nothing to do with Behe or darwinian evolution, either, but who expects understanding from trolls?)

    Back on topic. I think the real problem is commissioning a brief dismissal from a science writer instead of a careful review from an evolutionary biologist. But hey, that's the Discover we know and "love".

    ReplyDelete
  12. Larry Moran said...

    Nick,

    I know that you understand the flaws in Behe's argument and I agree with most of what you say. Behe doesn't understand how evolution actually works.

    My point is that if you're going to criticize Behe then that's the point you have to make. You need to explain in a review that Behe doesn't understand neutral mutations and pre-adaptation.

    Did you see anything in Powell's review to suggest that he (Powell) understands the fatal flaw? I didn't. In fact, it seems to me that Powell accepts "Darwinian" theory since he mentions it without adding a disclaimer.

    Tuesday, July 31, 2007 8:54:00 AM


    I think the difference between us may be that I take Behe's "concessions", e.g. about antifreeze, much less seriously than you do. Behe accepts the evolution of an antifreeze gene because he accepts that there is a stepwise pathway. But he then goes and argues that features XYZ (binding sites, double binding sites, IC complexes, whatever) cannot have evolved *because*, he *assumes* (with minimal or nonexistent argument) that there is no stepwise pathway. Then he spends most of the book arguing that without a stepwise pathway, natural evolution is wildly improbable. But this is all misdirection because the key argument, that there is no stepwise pathway, has not been made.

    Basically, we become beholden to Behe's arbitrary instinct about whether or not there is a stepwise pathway. If he didn't think there was a stepwise pathway for the antifreeze gene -- and again and again he has denied the existence of such pathways in other cases where the evidence is just as good -- he could "demonstrate" with his math that "random evolution" couldn't produce the antifreeze gene by "pure chance" in the lifetime of the universe -- which it definitely couldn't.

    So in this context, the Discover magazine review doesn't seem bad to me. It points out that Behe mostly ignores the possibility of stepwise pathways to the features he is interested in, and points to a few examples where such pathways have been demonstrated for systems that Behe would, if his instinctive "design detector" were consistent, call wildly improbable under random evolution and therefore designed. This is probably the most important point to make.

    In a short, popular-level review I'm not sure you can expect much more. I'm sure Discover would nix any detailed talk about amino acid substitutions or binding site strengths etc as being too much detail for the general reader.

    ReplyDelete
  13. "In a short, popular-level review I'm not sure you can expect much more. I'm sure Discover would nix any detailed talk about amino acid substitutions or binding site strengths etc as being too much detail for the general reader."

    I can see your point, Nick, but I will add this. I do think it would be beneficial for even a popular science mag like Discover to assign a reviewer who has the knowledge that Larry has in order to explain the flaws. This post taught me more than the original review did, and I am a lay person whose visits to this blog have taught me a great deal. Larry has the skill to explain the details of microbiology and evolution in an understandable way.

    Assigning a reviewer who gets it wrong does more harm than good, and leads to justifiable criticism of "darwinists" by creationists. I look forward to the reviews that Larry has promised.

    ReplyDelete
  14. "Behe thinks that a mutation has to be beneficial in order to be fixed in a population. He doesn't allow for (significant) neutral mutations ..."

    Behe discusses the sickle-cell mutation, though, which has remained fixed to a degree outside of areas where malaria is a concern. But Behe's point still stands with neutral mutations, the argument is about many mutations being required, without being selected for the aspects they will have in the final function.

    ReplyDelete
  15. Larry Moran: "I'm not arguing that Behe is correct. In fact, I'm preparing a series of postings that will challenge some of his ideas. What I'm objecting to is the mischaracterization of Behe's arguments in many of the published reviews."

    Larry Moran: "In order to show that Behe is wrong you have to demonstrate that his understanding of evolution (i.e., "Darwinism") is wrong and this has led him to false conclusions about probabilities. Many reviewers have failed to do this, possibly because they accept Behe's version of Darwinism."

    Larry, do you have a review of Behe's book or any posts that challenge his ideas or any posts showing that his understanding of evolution is wrong?

    ReplyDelete
  16. Hello Larry. We are still waiting for your review of Behe's latest book. Just reminding...

    ReplyDelete
  17. Hello, Larry

    In your post you say:
    "I'm not arguing that Behe is correct. In fact, I'm preparing a series of postings that will challenge some of his ideas. What I'm objecting to is the mischaracterization of Behe's arguments in many of the published reviews. If you're going to criticize Behe then challenge the argument he makes in the book; namely, that most stepwise pathways are impossible because the intermediates are less fit than their parents."

    I've just read Edge of Evolution and I am sure there are weaknesses in Behe's argument, but I don't have the scientific background to pinpoint them. Was that series of posts ever written? They would be a big help to me.

    ReplyDelete
    Replies
    1. Sorry, I never published those posts even though I have draft versions.

      Delete