|
Author
|
Topic: Evolving Inventions
|
RBH
Member
Member # 380
|
posted 08. February 2003 21:24
Frances wrote quote:
Perhaps it would be helpful to see if we can agree that GA's can indeed increase their hypervolume.
I'm a little perplexed by the apparent disagreement here. The "hypervolume" for a GA, as I understand the way that term is being used in this discussion, is the space defined by the possible combinations of alleles in the 'chromosome' - the gene string - of a critter in the population evolving in the GA. For example, in Avida one standard initial critter is a string of assembly language instructions that is 56 instructions long. There are 21 different instructions in the default assembly language, so there are 21^56 'locations' in the phase space - 21 possible 'values' on each of 56 genes/dimensions.
Now, among other kinds of mutations, Avida allows insertion mutations. On random occasions, a random instruction is injected into a random location in the gene string of a reproducing critter as it copies itself in preparation for dividing, increasing the offspring's genome length. When that occurs, the offspring's gene string (its list of assembly language instructions) increases from 56 instructions long to 57 instructions long. The space - the "hypervolume" - that this new mutated critter inhabits is thus larger than its parent's, with 21^57 locations versus the 21^56 locations its parent had, an order of magnitude increase. The new critter inhabits a "hypervolume" with one more dimension than its parent.
Many insertion mutations are lethal in the sense that they disrupt the ability of the offspring critter to subsequently reproduce itself. However, some insertion mutations are not lethal and may be neutral or selectively advantageous, and thus the increase in dimensionality can be preserved. When that occurs, for some time after the mutation some of the critters in the total population in the GA (the 56-instruction critters) are evolving in a phase space of one dimensionality while other critters in the population (descendants of the 57-instruction mutant) are evolving in a space of higher dimensionality. Over generations, if the mutant 57-instruction offspring is sufficiently selectively superior to its 56-instruction cousins it could take over the population. But while both varieties are present in the population, it doesn't even make sense to speak of the dimensionality of the hypervolume of the GA: there are two hypervolumes being evolved in.
This is not exotic stuff, folks. A GA can indeed change the dimensionality of the phase space it is evolving in.
RBH
[ 08. February 2003, 21:56: Message edited by: RBH ]
IP: Logged
|
|
John Bracht
Member
Member # 5
|
posted 13. February 2003 00:53
Hi everyone,
Wow. Life has been incredibly hectic, I apologize for being absent for so long. It's been frustrating, wanting to reply to so many points but finding absolutely no time to do so. Now that my Southern blots are mostly done (and they came out very well! ![[Smile]](smile.gif) ) I want to address a few points. Unfortunately, the sheer volume of response on this thread means that it will take some time and I will just have to start working through the pile. I'll go through, person by person, making responses. To some extent, I will probably have to pick and choose who to respond to, but some really good points have come up and I'll do my best to reply to those.
Keep in mind, it might take a week or two for me to get around to replying to all the important ideas! So I beg your patience. (lack of response doesn't mean I have no response, just that I'm too busy right now to give it).
John
IP: Logged
|
|
John Bracht
Member
Member # 5
|
posted 13. February 2003 01:37
Frances,
First off, I'm a bit annoyed by your attitude. You come across as the "battle warrior" who just wants everyone to basically agree that I'm wrong (about GA's and evolutionary processes increasing the dimensionality of the hypervolume) but you give me no indication that you really understand what I mean by the dimensionality of the hypervolume. Please stop making comments like "maybe now we can all agree that Bracht was incorrect" etc. These come not from a desire to gain knowledge or have a good discussion but from a desire to win the argument at all costs--and they don't fit the spirit of a board like Brainstorms. Thanks.
I feel like I've hammered this point home so many times and yet people are simply not getting it. To show what arguments I feel have not been understood and addressed, I will do a bit of cutting an pasting to repeat the arguments I've given that address France's constant refrain of how genetic duplication allows for expansion of the hypervolume.
quote:
Let's look at another good example, again from the paper: the Biomorphs program. Dawkins himself took a 9-gene program and made a 15-gene program from it. Did he do it just by gene duplication? No. He had to alter the previously existing structure so the new genes would have new functions. They have to "make sense" to the pre-existing network, and that requires the old network be re-wired to integrate the new variables. What happens if we just duplicate genes? We get duplicates of a function we already have--we don't get a whole new functionality.
Part of the problem for a GA is that oftentimes many genes or variables are tied together in the production of any given structure. This is certainly true in biology, and Dawkin's angel wings example demonstrates nicely why gene duplication alone doesn't give inventive change. In order to sprout angel wings, there are a host of new genes needed, and they have to do new things. Genes for new bone, muscle, tendon, blood vessels, etc., must be incorporated into the developmental program of the embryo to have an effect from the very beginning. Such a major change will involve altering a good many other developmental pathways such that this new one is properly integrated and regulated (controlled).
[from: http://www.iscid.org/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=6;t=000128 ]
Why don't gene duplications give you higher dimensional hypervolumes? Because:
quote:
As I've stated before, an inventive change in biology doesn't just involve adding new genes (though that is important) but also in integrating those genes into the existing network of genes (the regulatory network) in such a way that they mean something new; produce a new structure or feature. The intuition pump in this case came from Dawkins who talked about the difficulty of sprouting angel wings from a mammal's back. The process of embryonic development defines a hypervolume of possibilities which can be explored by genetic mutation. The ability to sprout angel wings is not in that hypervolume. Why? Because sprouting angel wings would require re-working the entire process of embryonic development, adding new genes and enmeshing them within the framework of other genes such that wings were formed. That re-working of the "rules" of gene expression and embryonic development, along with adding extra genes, is what I'm referring to as an inventive change in biology. I've never seen an example where that sort of change was observed to be the result of Darwinian processes. The reason, I argue, is that the process of embryonic development is a set of basic parameters which define the hypervolume in which the organism resides. Mutations to the genome serve to move about within that hypervolume (by their effect, "filtered" through development), but they do nothing to alter the fixed paramters of embryonic development in which the organism resides. (Notice I'm not saying that no mutations ever affect embryonic development; surely some mutations cause it to go badly awry or to terminate early, etc. But these do not alter the hypervolume that the organism exists in; they only mess up what is already established.)
[from: http://www.iscid.org/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=6;t=000128 ]
Basically, the bottom line is that inventive change involves altering the patterns of interactions of genes, how they tie together into an integrated whole. Again: think of the Biomorph example. Imagine that Dawkins had simply duplicated a gene in his program and allowed it to evolve under the rules pre-established for his program. Would he ever get the capacity for a "blue" biomorph? No. The reason is that the rules that interpret that new gene had to be changed to give new functionality. The reason this matters is that biological systems have "rules" which interpret genes. These rules consist of the ways that genes are wired to regulate each other. Having another gene doesn't give you another dimension of the hypervolume of possibility--that new gene has to be integrated into the network in such a way that it has new meaning!
Thus, for a GA to accurately model inventive changes, it has to have some sort of analog to embryological development. Sure, if you have a program where any new gene is automatically interpreted as a "dimension" in some vague sense, anything can evolve. You haven't constrained the search space in a biologically realistic way. That's why Dawkin's biomorphs example was so instructive--it accurately captures the constraints that biological systems have: they need to go through a developmental program that produces a viable organism.
The bottom line is this: you keep on pointing to computer programs that allow for gene duplications, and insisting that these changes are "inventive" and thus I've been falsified. But you haven't shown (1) that these programs accurately model biotic reality, and (2) that the programs produce anything truly inventive in the sense of altering the "rules" of development such that a new feature is produced. I'm sorry, but simply "literature bombing" with a few lines about "this program evolves new dimensions through gene duplication" are simply not enough. You've got to do the hard analytical work of showing how these changes were inventive and how the programs accurately model the biological reality we're interested in. Don't just provide the reference and a bald assertion; provide logic and thoughtful reasoning/evidence to support your claims (spell it out in your post). I'm not going to do that work for you, but I'd love to see you provide a convincing counter-example. It really would be fascinating and I'll gladly agree with you at that point (If you provide such an analysis, and you do so correctly) that I'm wrong and GA's can indeed increase the dimensionality of their hypervolume. Until then, you haven't even begun to address my arguments, much less falsify my argument about the dimensionality of evolutionary processes, so your continual refrain of "let's all agree that John's claim is incorrect" is simply hyperbole. Bottom line: you need to do the work of showing me wrong before you claim that I am wrong.
Let me explain a little more about what I mean about the "rules" of gene interpretation and how they tie into ontology. Let's consider a mouse vs. Drosophila Pax-6, the Homeobox gene that regulates eye development. You're undoubtedly familiar with studies which have shown that the mouse gene, when put into Drosophila, can drive the formation of normal-looking fly eyes (even ectopic expresson on the legs etc). So the very same gene, can "mean" two different things. In a fly, it "means" a fly eye. In a mouse, it "means" a mouse eye. It's this regulatory system of gene interactions (the genes upstream and downstream of the homeobox gene) that decide what the gene will "mean", much like Dawkin's computer program had code to determine what his biomorph's genes would "mean". It is the genetic regulatory network that does the work of interpreting a given gene and produce a given result. The network can be truncated, like Rex's Pod mutant example: early genes in this network were mutated, with the predictable result that the network didn't function properly or functioned poorly. But notice that the topology of the network didn't change. Or think of a Pax-6 mutation that simply produces no eye--you fail to activate the downstream genes that would ordinarily produce an eye. But those genes are still "wired" together in the same way, they just aren't activated.
So when I'm looking for inventive changes in biology, I'm looking for fundamental re-wiring of genetic regulatory networks. It's interesting, I'm taking a class right now on the evolution of the "non-coding genome" which is mostly about the evolution of gene regulatory networks, and there has been absolutely no mechanism proposed. Just a bunch of comments about how a new genetic network "arose", was "installed" or whatever. The reason, I think, is that there really is no way to gradually alter such a network. If you want to produce a new structure, like a new appendage, it's not enough to duplicate a gene, or even a bunch of genes. It (they) will simply be wired into the original position, in the old gene network. You've got to add new regulatory elements that cause the appendage to be turned "on" in the right tissues. You also have to have the right downstream differentiation gene batteries that form different parts of the appendage and add muscles, nerves, etc. You've got to add all this genetic programming to get the appendage in the first place--before it can evolve. All this genetic machinery is necessary to properly "interpret" the duplicated gene in a new way. Sure, once it's there, the new appendage can evolve non-inventively to get longer, or shorter, or whatever. But the inventive change is re-wiring the genetic regulatory system. More on this in my response to Yersinia (next).
John
IP: Logged
|
|
John Bracht
Member
Member # 5
|
posted 13. February 2003 02:16
Yersinia,
You gave an interesting scenario, point-by-point, for the addition of a new segment and appendages to an organism, so let's go through it and think a bit more deeply about it.
quote:
Regarding morphology/development, would the following set of stages, if evolved through, constitute an "inventive" or "routine" solution:
1) Start with segmented metazoan.
Simple enough.
quote:
2) Duplicate a segment (e.g. the corresponding Hox gene is serially duplicated). Critter now has an extra segment somewhere in the middle (say, 5 instead of 4) and corresponding pair of legs, etc., that go with the segment.
This is where things get a bit interesting. Simply duplicating a Hox gene is not going to buy you another body segment. It will just buy you another Hox gene that provides redundant function to its parent gene. How to you propose to get a new segment? It's interesting to think about how Drosophila specify segments, with pair-rule genes. I was at a seminar by Lewis Wolport, where he pointed out that every segment is specified totally independently of every other segment! This is rather remarkable. Each Hox gene has a set of enhancers that ensure that it gets expressed only in the correct stripe down the embryo. This set of enhancers is set up such that only in the target tissue does just the right combination of transcription factors activate it (these transcription factors are set up in gradients in the embryo). This actually makes it difficult to evolve a "new" stripe, because you can't just duplicate the gene. You have to evolve new enhancer elements which will bind to the unique combination of transcription factors (both enhancers and repressors) that exist in the target tissue. So just to add one stripe next to the parent stripe posterior of even-skipped 2, you'd have to make the following changes:
-activated by low levels of Bicoid instead of moderate levels
-repressed instead of activated by Hunchback
-activated, not repressed, by Krupel
-regulation by other transcription factors that define the posterior-most edge of the stripe (and possibly others at the anterior edge).
etc--and all these changes have to occur at the same time to get expression in a new stripe!
And you'd have to make sure the new gene has a new region of tissue in which to be activated (in other words, that it's not overlapping with the stripe that's already posterior to the parent stripe)--somehow the system has to have extra "room" and move everything down a bit so the new stripe can be intercalated.
And that's just to get one lousy Hox gene expressed in a new segment! Basically, you are having to re-wire the regulatory network such that the new Hox gene means something new. And we're not talking at all about the downstream genes that will be activated by this new Hox gene. Presumably, they will be seeing a new environment too (since the tissue has a different context than the parent tissue, different levels of various transcription factors in gradients, etc) they will also have to be "wired" into the system in such a way that they will work properly. I don't really know enough about how body plans are genetically wired to comment in more detail about this, though.
quote:
3) Repeat step 2 a number of times (e.g., selection for larger body size retains these duplications)
Ok...*grin*....maybe I should just make a bumper sticker: gene duplication doesn't give morphological novelty!
quote:
4) Once there are a fair number of segments, mutation and selection modify one or several of the more forward pairs, e.g. to improve prey capture or food-chewing or substrate/mate clasping (large number of possibilities here, lots of arthropods have these kinds of specializations).
Sure, I have no problem with mutation selection gradually adapting a given set of appendages that already exist. I have more of a problem of just getting more appendages (and segments, etc) in the first place.
This seems a good place to address Rex's concerns about the famous "mutant shrimp" study from the McGinnis lab (I know, I know, there was no mutant shrimp!). I actually got to hear Matt Ronshougan talk about his work last fall at the biology retreat, and it was very interesting. Basically, he argues that the basic wiring system between insects and shrmp, with regard to leg formation, is very similar, but shrimp have a sort of double-repression system in place that would "default" to a repression of legs if it weren't for a higher-level repressor of that repression. Insects lost that upper level of repression in their abdomens and thus don't produce legs there. His scenario really doesn't account for the production of any morphological novelty and actually begs the question of why a double-repression system would be present in shrimp (what's the selective advantage of that over a mere positive regulator!?). At any rate, it's clear that the genetic regulatory network is basically maintaining its wiring and nothing inventive is happening--you're just truncating the network at a given level in insects. Notice I'm not saying that there is no inventive change from insects to shrimp (I don't know enough about their mutual regulatory networks to comment) but just that in the loss of legs from shrimp to insects, nothing inventive appears to have happened. This is just intuitive, anyway--loss of legs hardly counts as the addition of morphological novelty. As an aside, I talked with Matt afterword (he's a really nice guy) and he agreed with me that his model (of loss of repression) can't really drive the sorts of morphological innovations that are common in biology--it would be a bit absurd to argue that the univeral ancestor had all possible body-plans encoded, and repressed, only to be unveiled by the loss of that repression over evolutionary time.
So I hope this helps. I think you're beginning to get a hint of what sorts of biological changes are inventive and which are not.
Ok, it's getting late and I'll have to quit for now. More responses later on, maybe this weekend (but no promises!)
John
[ 13. February 2003, 02:19: Message edited by: John Bracht ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 13. February 2003 04:50
Hey John,
OK, I probably oversimplified the situation with regard to segment duplication vs. hox gene duplication. But the fact is that segment number varies in a hereditary fashion in extant species:
quote:
Evol Dev 1999 Jul-Aug;1(1):62-9
Variable segment number in centipedes: population genetics meets evolutionary developmental biology.
(links to related articles, there are lots of articles on this)
Arthur W.
The case studies of population genetics focus on intraspecific variation, but most cases--at least where the variation is polymorphic--deal with characters that are not directly linked to organismic structure or ontogeny. Conversely, the case studies of evolutionary developmental biology focus directly on structure/ontogeny, but usually involve only interspecific comparisons. To integrate these complementary approaches, it is desirable to have a model system that permits study of intraspecific variation in development, using a character whose genetic basis either is already known or can be elucidated. Segment number in geophilomorph centipedes is proposed as a possible model system of this kind. Segment number is variable in natural populations of geophilomorphs, while in the other centipede orders it is fixed, either completely (scutigeromorphs, lithobiomorphs), or at least within species (scolopendromorphs). Statistical analysis of data on the extent of variation in different geophilomorph species suggests that segment number may be of selective importance, rather than the variation being merely an inevitable consequence of the difficulty of achieving a high degree of repeatability when there is a large number of segments.
...it doesn't look nearly as hard as you seem to think it is. And note that things like centipedes, which basically have two types of segments, head and trunk, resemble the proposed ancestors of crustaceans and insects in that respect:
quote:
Development 2000 Jun;127(11):2239-49
Crustacean (malacostracan) Hox genes and the evolution of the arthropod trunk.
Abzhanov A, Kaufman TC.
Representatives of the Insecta and the Malacostraca (higher crustaceans) have highly derived body plans subdivided into several tagma, groups of segments united by a common function and/or morphology. The tagmatization of segments in the trunk, the part of the body between head and telson, in both lineages is thought to have evolved independently from ancestors with a distinct head but a homonomous, undifferentiated trunk. In the branchiopod crustacean, Artemia franciscana, the trunk Hox genes are expressed in broad overlapping domains suggesting a conserved ancestral state (Averof, M. and Akam, M. (1995) Nature 376, 420-423). In comparison, in insects, the Antennapedia-class genes of the homeotic clusters are more regionally deployed into distinct domains where they serve to control the morphology of the different trunk segments. Thus an originally Artemia-like pattern of homeotic gene expression has apparently been modified in the insect lineage associated with and perhaps facilitating the observed pattern of tagmatization. Since insects are the only arthropods with a derived trunk tagmosis tested to date, we examined the expression patterns of the Hox genes Antp, Ubx and abd-A in the malacostracan crustacean Porcellio scaber (Oniscidae, Isopoda). We found that, unlike the pattern seen in Artemia, these genes are expressed in well-defined discrete domains coinciding with tagmatic boundaries which are distinct from those of the insects. Our observations suggest that, during the independent tagmatization in insects and malacostracan crustaceans, the homologous 'trunk' genes evolved to perform different developmental functions. We also propose that, in each lineage, the changes in Hox gene expression pattern may have been important in trunk tagmatization.
There are innumerable cases of within-species variation in repeated structures, e.g. snake vertebrae, human ribs, etc.
Anyway, the basic idea I proposed, duplication and specialization of segments, is not something I came up with myself. It appears to be a very well-accepted idea, so you've got to argue with the literature rather than little old me:
quote:
Nature 1995 Aug 3;376(6539):420-3
Hox genes and the diversification of insect and crustacean body plans.
Averof M, Akam M.
Wellcome/CRC Institute, Cambridge, UK.
Crustaceans and insects share a common origin of segmentation, but the specialization of trunk segments appears to have arisen independently in insects and various crustacean subgroups. Such macroevolutionary changes in body architecture may be investigated by comparative studies of conserved genetic markers. The Hox genes are well suited for this purpose, as they determine positional identity along the body axis in a wide range of animals. Here we examine the expression of four Hox genes in the branchiopod crustacean Artemia franciscana, and compare this with Hox expression patterns from insects. In Artemia the three 'trunk' genes Antp, Ubx and abdA are expressed in largely overlapping domains in the uniform thoracic region, whereas in insects they specify distinct segment types within the thorax and abdomen. Our comparisons suggest a multistep process for the diversification of these Hox gene functions, involving early differences in tissue specificity and the later acquisition of a role in defining segmental differences within the trunk. We propose that the branchiopod thorax may be homologous to the entire pregenital (thoracic and abdominal) region of the insect trunk.
Nature 1997 Aug 14;388(6643):682-6 Related Articles, Links
Crustacean appendage evolution associated with changes in Hox gene expression.
Averof M, Patel NH.
Wellcome/CRC Institute, Cambridge, UK.
Homeotic (Hox) genes specify the differential identity of segments along the body axis of insects. Changes in the segmental organization of arthropod bodies may therefore be driven by changes in the function of Hox genes, but so far this has been difficult to demonstrate. We show here that changes in the expression pattern of the Hox genes Ubx and AbdA in different crustaceans correlate well with the modification of their anterior thoracic limbs into feeding appendages (maxillipeds). Our observations provide direct evidence that major morphological changes in arthropod body plans are associated with changes in Hox gene regulation. They suggest that homeotic changes may play a role in the normal process of adaptive evolutionary change.
quote:
So I hope this helps. I think you're beginning to get a hint of what sorts of biological changes are inventive and which are not.
I'm getting the sense that it is entirely subjective and that the goalposts are very subject to motion. You haven't specified whether any of the innovations I cited (yucca moths) or postulated (segment duplication & modification) would, or would not, constitute "true inventiveness".
Until you give us a rigorous definition of "inventive", or at least lots of biological examples of what would count and what wouldn't, this is just another meaningless buzzword that supposedly constitutes a block to evolution but is actually a moving target that moves just beyond whatever we are able to document with overwhelming thoroughness that evolution can accomplish.
(other examples include IC, SC, and "information")
Here's another case: novel "invention" or not?
quote:
Curr Biol 2002 Oct 1;12(19):1711-6
Diverse adaptations of an ancestral gill: a common evolutionary origin for wings, breathing organs, and spinnerets.
Damen WG, Saridaki T, Averof M.
Institute for Genetics, University of Cologne, Weyertal 121, D-50931, Koln, Germany.
Changing conditions of life impose new requirements on the morphology and physiology of an organism. One of these changes is the evolutionary transition from aquatic to terrestrial life, leading to adaptations in locomotion, breathing, reproduction, and mechanisms for food capture. We have shown previously that insects' wings most likely originated from one of the gills of ancestral aquatic arthropods during their transition to life on land. Here we investigate the fate of these ancestral gills during the evolution of another major arthropod group, the chelicerates. We examine the expression of two developmental genes, pdm/nubbin and apterous, that participate in the specification of insects' wings and are expressed in particular crustacean epipods/gills. In the horseshoe crab, a primitively aquatic chelicerate, pdm/nubbin is specifically expressed in opisthosomal appendages that give rise to respiratory organs called book gills. In spiders (terrestrial chelicerates), pdm/nubbin and apterous are expressed in successive segmental primordia that give rise to book lungs, lateral tubular tracheae, and spinnerets, novel structures that are used by spiders to breathe on land and to spin their webs. Combined with morphological and palaeontological evidence, these observations suggest that fundamentally different new organs (wings, air-breathing organs, and spinnerets) evolved from the same ancestral structure (gills) in parallel instances of terrestrialization.
If this kind of thing wouldn't constitute an invention in TRIZ terms, then I don't see how the term would have any applicability to biology.
Rigourous definition, please.
[added in edit] Err, rigorous.
[ 13. February 2003, 07:29: Message edited by: yersinia ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 13. February 2003 19:32
Ged:
Your quote from Genetics is about a “majority” of duplicated genes. It says nothing about the capabilities that can occur in lower probability. Also that the presence of the new function “already exist in the ancestral genes” is not a pointer to a “more complex ancestor” (or simpler descendent), rather it is a pointer to an ancestor of the same complexity. So what about the minority? What happens there?
Nelson:
Wrong again Ged. I actually specifically responded to this point as well. Here is what I wrote:
quote:
If we stick to the DDC model, that section of logic that would be duplicated contains instructions that are, say binary. 1 for true, 0 for false. Since a duplication event partitions these functions, we have one piece which encodes for true(1), the other encoding false(0). Since this is most likely the common mechanism of gene duplication, what does this say about inventiveness according to TRIZ? This further supports the claim that trial-and-error processes are restrained within the hypervolume but never jump to a new "higher-order" hypervolume. This is not of course, to say that a rare event in some time in the past has never happened that comes close, as rare unlikely events do happen.
What the DDC model states is that what we commonly see in gene duplications is subfunctionalization. My focus was both that what we get from gene duplication is less complexity and that this is the most common mechanism of gene duplication. This pretty much pulls the rug from your post.
1. The examples of gene duplication that we have are not the inventive solutions that John Bracth is asking for.
2. The most common form of gene duplication results in subfunctionalization and not a gene of equal complexity. I did discuss this in my original post. To get a little specific with respect to the paper, they cite the gene eng1a and gene eng1b from Zebrafish, so it goes like this:
Eng1(expressed in both)
/ \
/ \
/ \
eng1b(bud) eng1a (hindbrain)
That didn't come out quite as well as I wanted it to, but you get the gist of it.
Ged:
Alonso, you continue to focus on the aspect that provides no new function while ignoring the statements in your own quotes that do provide new functions.
Nelson:
As I show above, that is not what my focus was on, although it comes into play. After all, we are talking about quite a few gene duplication events,especially when it comes to multi-part systems. Are you sure you don't think this point is that important?
Ged:
Once again, that a majority or the higher frequency or probability of duplication events not producing a new dimension of variability says nothing about the possibility, the capacity, or the capability of those duplicated genes.
Nelson:
The argument here is not one of logical possibility, as we see some duplication events leading to new functions. The problem here is one of capability. If all gene duplication does (commonly, with the expection of a few quite simple examples) is partition the function then we don't really have that much empirical evidence to point to and say that "this is an example of evolutionary inventiveness", when in fact all it is trial and error processes.
Ged:
Once gain, if the low to mid probability event of a new function occurring (“whereas some gene duplicates clearly evolve new functions”) still occurs regularly, then a new dimensionality of change has been enabled by the gene duplication. In other words, the hypervolume has increased. (And if it did not, then the hypervolume was semi-infinite in the first place.)
Nelson:
The paper states that it does not occur regularly. Furthermore, you're just exhausting the problem with problems, since now you are pointing to a less complex gene that has a specialized function, and therefore less dimension.
Besides, you're just moving the problem forward in time when you say that duplicate gene offers the opportunity for new dimenionality. What do we really want? We want to know if natural processes can act outside the box, after all, even with new opportunities we still need to know if it can invent and not continue the trial and error processes that lead to technical contradictions.
Ged:
Sorry about the “[SIC]”, I was feeling obnoxious. As to John Bracht not responding, my first related posts were in December 2001 on the ARN thread he links to, and he didn’t answer those either.
Nelson:
I posted twice how Bracht has already answered the gene duplication issue and I'm finding your new material (FPGA), to be quite trivial to respond to (notice you never responded to my characterization of it).
Ged:
So as Frances says “Now the goal posts seem to have shifted to: Gene duplication is not enough in evolution, but that was not the point of objection.”
Nelson:
That was always the point, the goal posts have never been shifted. One of the mechanisms of evolution is gene duplication, but that is exactly the kind of trial and error process spoken about in the paper.
[ 13. February 2003, 22:54: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 14. February 2003 02:24
quote:
Nelson:
Wrong again Ged. I actually specifically responded to this point as well. Here is what I wrote:
…
quote:
What the DDC model states is that what we commonly see in gene duplications is subfunctionalization. My focus was both that what we get from gene duplication is less complexity and that this is the most common mechanism of gene duplication. This pretty much pulls the rug from your post.
Since you did not contradict what I said, it cannot logically “pull the rug” out of my point.
Alonso Nelson still argues that the most common mechanism does not produce new functionality. So what? I really don’t care if you had another point or wanted to express another point, my point is the one I made. You have not supported your implication that “new functionality” is exceedingly rare. And if you didn’t mean to imply that “new functionality” was exceedingly rare, then you are not arguing against my point. Lots of words about another subject are irrelevant.
If sometimes new functionality develops from gene duplication, then new functionality is a capability or potential result of gene duplication. Period.
IP: Logged
|
|
Aardvark
Member
Member # 141
|
posted 14. February 2003 03:59
Frances, in his first post, claims that computer simulations can produce high levels of information. The problem is that the information produced by computer simulations is really a function of the initial CSI that is input into the code, intelligently, by the programmer. Thus, CSI in = CSI out, and all the CSI in is the product of intelligent design. Frances then cites "gene duplication" as a mechanism for getting new CSI, as a way to get new hypervolumes. This of course is the typical evolutionary answer to how new information is generated. It is invoked everywhere from the literature down to undergraduate level discussions (see Evolution The World of Darwin
Sources of Variation, http://www.brooklyn.cuny.edu/bc/ahp/LAD/C20/C20_Variation.html ).
Once Rex joined in, I think I hit the nail on the head when he said, "real roadblock is the topology of the fitness function." I think that Rex is right on--the real question is, do the fitness peaks connect through some ridge or otherwise without going through some nonfunctional stage?
Perhaps I haven't read this thread close enough, but I'd like to see more discussion about whether or not adaptive landscapes really can connect a gene duplicate to its final form. To me, it seems like they would have to go through some kind of a valley/ocean-bottom of low/non-functionality before getting you anywhere close to the new function which must be obtained.
However, I'd like to address one point that Frances raised, namely that the evolution of the bacterial flagellum is beginning to look less improbable. Frances cited an article by Ken Miller. Even by Miller's own logic, the flagellum is still staggeringly improbable (see below).
Remember, any theory can be “saved” by speculative miraculous ad-hoc reasoning.
Miller Claim 1 (paraphrased): IC is an argument from ignorance.
Response: I think that Co-optation may block IC, but Miller must first justify co-optation as a viable mechanism before he forces IC and the design inference into a negative-design-of-the-gaps position.
Miller Claim 2 (paraphrased): The Type III Secretory System is an intermediate.
Consider these excerpts from Mike Gene, previously mentioned:
"Independent evidence suggests the type III system is recent. It is not only restricted to gram-negative bacteria, but to animal and plant pathogens. In fact, the function of the system depends on intimate contact with these multicellular organisms. This all indicates this system arose after plants and animals appeared. In fact, the type III genes of plant pathogens are more similar to their own flagellar genes than the type III genes of animal pathogens. This has led some to propose that the type III system arose in plant pathogens and then spread to animal pathogens by horizontal transfer."
…
"When we look at the type III system its genes are commonly clustered and found on large virulence plasmids. When they are in the chromosome, their GC content is typically lower than the GC content of the surrounding genome. In other words, there is good reason to invoke horizontal transfer to explain type III distribution. In contrast, flagellar genes are usually split into three or more operons, they are not found on plasmids, and their GC content is the same as the surrounding genome. There is no evidence that the flagellum has been spread about by horizontal transfer."
(“Mike Gene” (http://www.idthink.net/biot/flag1/))
and also conside:
"In line with the previous concern, Van Till offers the type III secretory system as a possible precursor to the bacterial flagellum. This ignores that the current evidence points to the type III system as evolving from the flagellum and not vice versa (cf. Milt Saier's recent work at UCSD). But beyond that, finding a component of a functional system that performs some other function is hardly an argument for the original system evolving from that other system. One might just as well say that because the motor in a motorcycle can be used as a blender, therefore the motor evolved into the motorcycle. Perhaps, but not without intelligent design. Even if it could be shown that the type III system predated the flagellum (contrary to Milt Saier's work), it could at best represent one possible step in the indirect Darwinian evolution of the bacterial flagellum. But that still wouldn't constitute a solution to the evolution of the bacterial flagellum. What's needed is a complete evolutionary path and not merely a possible oasis along the way. To claim otherwise is like saying we can travel by foot from Los Angeles to Tokyo because we've discovered the Hawaiian Islands. Evolutionary biology needs to do better than that.”
(William Dembski, response to Howard Van Till)
I think Dembski's analogy at the end hits the nail on the head: "What's needed is a complete evolutionary path and not merely a possible oasis along the way." Essentially, I want my bus ticket to Japan. Show me the continuous morphospace!
Miller Claim 3 (very paraphrased): Ok, so we still don’t know where all 30 proteins come from, even given the Type III Secretory System. But, inferring design assumes the truth of its own argument. Therefore Dembski’s 1/10^-1170.
(“Stated more bluntly, what this really means is that the "method” [of detecting design] first involves assuming the absence of an evolutionary pathway leading to the object, followed by a calculation "proving" the impossibility of spontaneous assembly. Incredibly, this a priori reasoning is exactly the sort of logic upon which the new "science of design" has been constructed.” Kenneth Miller, The Flagellum Unspun)
Response:
Miller has it backwards. Design detecting works by ruling out the “natural cause” first. If Dembski can do the calculation and rule it out, he’s done his job. Dembski did the calculation first, and then inferred design, and Miller seems to be wrongly reversing the process of the design inference. Besides, problems with the TPSS co-optation argument aside, Miller admits that 2/3 of the 30 proteins are still unaccounted for. Dembski’s numbers may not completely satisfy Miller, but if we follow Miller's recommendation, the odds are still 10^-780. Those are bad odds for Miller. That calculation must be refuted and Miller has made no such attempt. Miller either does not understand Dembski’s methods, or has constructed a straw man.
Miller Claim 4 (paraphrased): You’ve over come the “irreducible complexity” hurdle, but now you must overcome the “co-optation” hurdle. Otherwise, your argument becomes merely a negative “Design-of-the-gaps,” you just don’t know how it got co-opted yet argument. We are working on the co-optation argument, and will figure it out.
• Miller Argument 4 (quoted):
“[Design proponents claim] … we can look at a complex biological object and determine with absolute certainty that none of its component parts could have been first selected to perform other functions.” (Kenneth Miller, The Flagellum Unspun)
Response:
Do design proponents require “absolute certainty” or is Miller again constructing a straw-man? Perhaps in these epistemological games, “absolute certainty [that the natural mechanism doesn't work]” isn’t needed to infer design--just high probability [that the natural mechanism doesn't work].
Perhaps co-optation can explain away irreducible complexity, but we can always appeal to miracle mutations to stop design arguments. Does that mean that it is the design-proponent who employs “gap” type arguments? Why isn’t this just “co-optation” of the gaps? At some point , we must recognize that there is a gap that all must deal with. Statistically-based arguments, never “absolute certainty,” are the only possible epistemological way to address such questions. In other words, when playing poker, it’s always possible that someone else can beat your flush, but that doesn’t mean you fold your hand. At the end of the day, you’ve got to play the game by the numbers. Science is fundamentally a numbers and statistics game, and concrete facts of “absolute certainty” simply do not exist. It’s no surprise that Miller has found an ad hoc explanation--there is always an ad hoc explanation. The question is, “is it a likely or reasonable one?”
I think that John Bracht actually puts it quite well in his response to Goodenough:
"The problem is that the proteins which are to become the flagellum are coming from systems that are distinctly non-flagellar in nature (after all, we are discussing the origin of that very system) and being co-modified from their original molecular interactions into an entirely new set of molecularinteractions. Old interfaces and binding sites must be removed and new ones must be created. But given the sheer number of flagellar proteins that must co-evolve and the fact that the entire change must happen in one step, co-generating all the proteins required for flagellar function (again, this is true at some point in the flagellum's evolutionary past even if there were earlier steps that were not so tightly constrained), the Darwinian explanation is really no different from appealing to a miracle.”
(John Bracht, The Bacterial Flagellum: A Response to Goodenough)
• Miller Claim 5 (paraphrased):
• “We have intermediates for the once-alleged irreducibly complex Krebs Cycle, why can’t we assume we may end up finding one here?”
Response:
Fine that's great! But let’s all show our cards for this round of Poker. Where’s the explanation for the flagellum? What relevant reason do we have to believe co-optation works here?
• Miller Claim 6 (paraphrased):
Dolphins are missing one protein for blood clotting that we have, thus proving blood clotting is not irreducibly complex.
Response:
Irreducible complexity does not imply there is only one way of doing something, it just implies that one particular system will cease functioning if changed. What if dolphins just have a different pathway for blood clotting, which is also irreducibly complex?
• Miller again: “The very existence of the Type III Secretory System shows that the bacterial flagellum is not irreducibly complex.”
Response:
Those are pretty low standards you have for irreducible complexity. The TPSS is not anything like a flagellum, it just uses some of the same parts to perform a different function. Irreducible complexity, as Miller quotes Behe, says, “cannot be produced directly by numerous, successive, slight modifications of a precursor system, because any precursor to an irreducibly complex system that is missing a part is by definition nonfunctional.” Has Miller demonstrated that “a flagellum that is missing some part is still functional?” Has he shown it isn't IC in Behe's definition. I think not.
• “As an icon of anti-evolution, the flagellum has fallen.”
I don't think there was anything new in this argument that hasn’t already been addressed by design proponents before. The flagellum does not fall until one can come up with a plausible naturalistic mechanism for its existence. As of yet, the co-optation argument is too unlikely to stop the design inference. Until it is demonstrated that co-optation is a viable mechanism, it seems it is still a “miracle mutation” or “co-optation-of-the-gaps,” and epistemologically equal to an argument from ignorance. Miller is making grand and unbacked claims that are not true to the nature of the controversy, in what seems like an effort to silence opposition, not address arguments. I honestly am disappointed by this paper, and had hoped for better from “Irreducible Complexity’s Worst Nightmare,” Kenneth Miller (see BioScience, 10-1-00, No. 10, 50: 926 where botanist Barry Palevitz refers to Miller as such in a critical review of Miller’s book, Finding Darwin’s God).
Irreducible complexity will have sweet dreams tonight of Ken Miller provides no plausible explanation for the co-optation of the flagellum other than the TPSS, and still can't account for 2/3 of the proteins, much less how the TPSS proteins found their way to flagellumville, re the objections of John Bracht. Just some thoughts!
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 14. February 2003 05:22
Aardvark wrote:
quote:
Irreducible complexity does not imply there is only one way of doing something, it just implies that one particular system will cease functioning if changed.
This version of irreducible complexity only has consequences for evolution if "indirect pathways" are ruled out, and if gradual co-dependence is ruled out. If this point is unclear (Behe admits at least the indirect pathway problem in DBB), I will explain in more detail.
Dolphin blood clotting--which is missing one protein (with the other seven or so are very similar)--shows a clear example of a co-dependency in some organisms that need not exist.
The Type III Secretory System shows a functional complex that indirectly gets you about halfway to the flagellum (in terms of information content).
Examples like these do not say that these systems aren't irreducibly complex (or don't have an irreducibly complex core consisting of many proteins). Rather they cast doubt on the assumptions needed to make irreducible complexity, as defined, relevant to evolution.
quote:
Miller Claim 5 (paraphrased):
"We have intermediates for the once-alleged irreducibly complex Krebs Cycle, why can't we assume we may end up finding one here"
Fine that's great! But let's all show our cards for this round of Poker. Where's the explanation for the flagellum? What relevant reason do we have to believe co-optation works here?
Well, we have the Type III Secretory System. If you co-opt that, instead of building it again from scratch, you've gotten halfway there.
But the real question is what we should expect to find in nature if evolution vs. some other process is responsible for a particular protein complex.
First, note that we have not solved either the protein folding problem or the structure-function problem. Thus it is impossible for us to do any reasonable reverse-evolution simulation. So we are, for the most part, limited to looking around at what we can find in nature.
Unfortunately, there are no "protein fossils". We can only find proteins that exist now. From fossil evidence, we see that something like 99% of all species have gone extinct, and many major phylogenetic groupings have also gone extinct (e.g. dinosaurs). The situation is less clear with protein complexes, but likewise, we might expect that the vast majority of proteins no longer exist anywhere. This is a real barrier to our being able to find a complete step-by-step set of proteins for every process.
Even more unfortunately, the structure and function is known for only a tiny fraction of the proteins in extant organisms. So even if there are intermediate complexes that we could find, in very very many cases we would expect to not have found such complexes yet.
So it simply isn't fair to demand of evolutionary theory that any protein complex that happens to be known now be explicible in step-by-step detail, using examples from existing protein systems. In fact, the nature of evolution (e.g. most everything dies and is lost) specifically predicts that you won't be able to do that in many cases.
Now, does this mean that molecular evolution is untestable and fits all hypotheses? Absolutely not. Because it makes other strong predictions about the nature of the data that we should find. Specifically, each protein complex ought to be at least one of
- Non-irreducibly complex
- Gradually co-dependent
- Obtainable using indirect functional intermediates
If this is the case, we ought to find many examples of each. And in fact, we do find many cases of each. Some subunits of, say, voltage-gated calcium channels (or DNA polymerase II, or the nuclear pore, etc. etc. etc.) can be lost while still maintaining useful functionality. There are all types of derived co-dependence observed, such as the dolphin blood clotting example, the Eng1 vs. Eng1A/Eng1B example in previous posts, and so on. And there are all types of complexes that could serve as functional intermediates, such as the TPSS for the flagellum.
But we can devise more stringent tests than that. Specifically, based on which species a complex is found in (among other things) we can make estimates of how old a given protein complex is. The more recent the complex is, the better chance we have of finding intermediates. So we would expect that recently evolved complexes could have the steps filled in much more easily than ones that evolved a long time ago. An ancient complex has had much more time for any key intermediates to be lost.
This is why it is exactly backwards to focus attention on ancient complexes like the bacterial flagellum and nuclear pore and so on. You don't expect to be able to fill in every piece in an evolutionary pathway of that age (without vigorous handwaving), and gee, we can't. What would be particularly telling is if we looked at recent complexes and couldn't explain those! That would be a much more serious blow to evolution, especially if you could come up with quantitative predictions for how often you ought to be able to fill in which types of gap given the age of the complex and the number of extant species and the times of their divergence.
Anyway, the bottom line is that we have many examples of evolved co-dependence and surprising functional intermediates. Therefore, these absolutely cannot be ignored as "improbable" when irreducible complexity is applied to evolution.
Edited for typos.
[ 14. February 2003, 12:21: Message edited by: Rex Kerr ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 14. February 2003 10:49
Aarvark wrote quote:
Once Rex joined in, I think I hit the nail on the head when he said, "real roadblock is the topology of the fitness function." I think that Rex is right on--the real question is, do the fitness peaks connect through some ridge or otherwise without going through some nonfunctional stage?
Perhaps I haven't read this thread close enough, but I'd like to see more discussion about whether or not adaptive landscapes really can connect a gene duplicate to its final form. To me, it seems like they would have to go through some kind of a valley/ocean-bottom of low/non-functionality before getting you anywhere close to the new function which must be obtained.
Rex Kerr has addressed some of the specific issues associated with the flagellum and Miller's and others' arguments. Here I want to focus on a more general cautionary point.
The metaphor of a fitness "landscape," with hills and valleys and ridges as we know them as physical features of the 3-dimensional world, is intuitively appealing and even seductive. It is also more than slightly misleading. A GA employed in an applied search context evolves on high-dimensioned fitness landscapes. There is one dimension of the "surface" of the landscape for every selectively-relevant gene (assuming one-to-one gene-to-phene mapping in the application), with the values on that dimension being defined by the gene's alleles. Thus, for example, the GAs my company builds and deploys have fitness landscapes in 72 dimensions.
For an evolutionary algorithm to traverse such a landscape, it is only necessary that the landscape be locally correlated in at least one dimension. That is, for the population to 'move' about the landscape it need only be able to find slightly more fit adaptations on as few as one of the 72 dimensions. The immediate implication is that, contrary to the expectation induced by our intuitions, it is hard to "trap" the population. In order to trap one of our populations on a local maximum, unable to find a path to distant parts of the landscape, it is necessary that the current local maximum be isolated by "cliffs" in every one of those 72 dimensions. Since our GAs model real-world phenomena, and since many variables in the world are continuous, that turns out to be a relatively low frequency event rather than being common.
There is a further consideration. Fitness landscapes in nature are not static, but are plastic. Long-term changes in the physical environment alter the fitness landscape on which a biological population evolves. More important, the fitness function, and therefore the topology of the landscape induced by the evolutionary operators, is partly determined by the existence and evolution of other populations with which our population interacts. Co-evolution changes fitness landscapes. As a predator becomes more efficient at hunting, or acquires a new technique for hunting, the fitness landscape of the prey is deformed and changes.
Finally, I must express some reservations about Rex's remark that quote:
The real roadblock is the topology of the fitness function on that hypervolume. Specifically, whether an "inventive" solution can be found depends on the ability of the genetic search algorithm to escape the current distribution about the old local fitness maximum and begin to climb the fitness peak that contains the "inventive" solution. If the topology has many ridges, the genetic algorithm will have an easier time of exploring than if the topology consists of gigantic peaks.
"Fitness function" and "fitness landscape" are not synonyms. A given fitness function (an equation) in combination with an evolutionary operator induces a fitness landscape with a particular topology.
A fitness function absent an evolutionary operator specifies no particular topology. Assuming a fixed-for-the-moment fitness function, each evolutionary operator (point mutation, recombination, insertion mutation, duplication mutation, deletion mutation, etc.) induces a different fitness landscape, in the sense that "nearest neighbors" on one landscape are almost certainly not nearest neighbors on another landscape, and thus their topologies are different. The nearest neighbors of a given point on the fitness landscape induced by point mutations are different from its nearest neighbors on the fitness landscape induced by recombination. For a population to be "trapped" it must be trapped on all dimensions of all the fitness landscapes induced by the array of evolutionary operators. (I developed this point in more detail elsewhere, but I'l be darned if I can find that thread now.) Some evolutionary operators directly alter the dimensionality of fitness landscapes. For example, as I remarked earlier in this thread, insertion mutations, like duplication mutations, increase the dimensionality of the "hypervolume" in which an evolutionary process swims.
The conclusion to be kept in mind is that implications drawn from a metaphor that encourages seeing the topology of fitness landscapes as single, static, low-dimensioned structures can be very misleading. It is necessary to analyze the specific case in order to make statements about the possibility, or lack thereof, of evolution from "here" to "there."
John is right in that for insertions or duplications to qualify as generating TRIZ-like "inventive" change they must go on to be incorporated into the "network" of existing genetic and developmental interactions. We know (inferentially) of instances where something like that had to have happened. For example, the migration of bones from reptilian jaws to the sound transmission apparatus of mammals is well documented in the fossil record and clearly required substantial developmental changes. Is that "inventive" in TRIZ terms? It sure looks like it to me.
Finally, I will say again that Dawkins' BioMorph and WEASEL, along with ISCID's MESA, are not good bases for illustrations, inferences, or conclusions about actual evolutionary processes either in nature or in fully-developed GAs used in actual applications. They are at best illustrative of very limited points; at worst they are actively and lethally (for arguments based on them) misleading.
RBH
IP: Logged
|
|
Aardvark
Member
Member # 141
|
posted 14. February 2003 13:02
RBH,
Thanks for clarifying about the difference between a fitness function and an adaptive landscape, that made a lot of sense. I completely understand that an "adaptive landscape" can only represent 3 dimensions, whereas the realworld / GA's can deal with many more variables.
I also understand that landscapes will shift according to the environment. There's an excellent example of this online at "http://www.sigmaxi.org/amsci/articles/02articles/grantcap8.html" where the famous finchbeak researchers, the Grants, discuss how adaptive peaks shift for Darwin's finches in response to environmental changes.
One way I perceive these problems (tell me if I am way off or not) is that we can have adaptive landscapes which are more static, governed by static "enviromental rules" dictated by the laws of physics and chemistry. For example, couldn't an adaptive landscape be constructed to show the specificity of a flagellum, in that it always has to have certain parameters just right in order to function? I just don't see those parameters changing, and wouldn't they create a static adaptive landscape? Am I way off here? Appreciated your thoughts!
Also, RBH, please help me to be clear in my understanding of what you wrote: you said trapping a GA on a local maximum is rare because one of the variables is often continuous and can usually help the GA get from peak A to peak b. Is that correct? Does this simply imply that there is often more than one way of doing something? Maybe I missed your point completely.
Whoops, I just saw Rex replied to me too, I'll have to look at that later today when I return.
By the way, RBH, the concept that the migration of reptilian jawbones --> mammalian earbones is documentd in the fossil record is somewhat of an "urban legend" among evolutionists, that I think needs to die. There are a few skulls that close the gap on either end of this transition, but the most important part, the actual transition itself, remains a gap:
"It is not yet certain when the malleus and incus became incorporated into the middle ear, but the grooves on the medial surface of the dentary that indicate their position of attachment in early Jurassic mammals are missing in Upper Jurassic genera.”
(Carroll, Robert L. 1988. Vertebrate Paleontology and Evolution. W. H. Freeman. New York.)
Here are a couple of other interesting notes:
"Since the Theria and the Atheria separated from each other before the changes in the middle ear had taken place, these two major groups must have evolved mammalian auditory ossicles independently. This is a most surprising fact”
(Kermack, D. M. and K. A. Kermack. 1984. The Evolution of Mammalian Characters Kapitan Szabo Publishers. Washington, DC.)
"The vertebrate lower jaw, for example, is composed either of several separate parts and joined to the skull by the auricular, as in reptiles, or -- as in mammals -- consists of a single bony element, the dentary, which takes on the function of articulation with the skull. Slow, smooth transitions between these qualitatively opposing structures taking place during postembryonic developmental stages, when the jaw mechanism must be able to function, are inconceivable. To be sure, we recognize in the reptilian lineages that lead to mammals a gradual, quantitative reduction of the articular and of the other individual bones of the lower jaw, paving the way for the transformation and bringing the two types closer together. However, the fundamentally decisive, final step -- the complete disappearance of these bones or their transformation into elements of the auditory area -- must have taken place discontinuously, suddenly, between one individual and the next, during an embryonic developmental stage.”
(Schindewolf's O. (emphasis in original), Basic Questions in Paleontology (1993), pages 211-212, University of Chicago Press, Chicago. (Original German edition 1950).
I think re the reptile-mammal jawbone-earbone transition, we are really dealing with a transition that is inferred based upon some fossils that close the gap to a limited extent, not because of actual fossils documenting the jawbone-earbone transition itself. I think it's an urban legend that needs to die.
Also, this last quote makes me wonder if John's analogy might work with this transition. It seems that some pretty radical renovation of the bone structure is necessary for this to happen, perhaps this is a good example of John's hypervolume at work.
~AA
[ 14. February 2003, 13:09: Message edited by: Aarvark ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 14. February 2003 17:50
Gedankin,
The only examples that we see where gene duplication evoke a new function were very simple examples. That "sometimes new functionality develops from gene duplication, then new functionality is a capability or potential result of gene duplication" is irrelevant, as we are discussing whether gene duplication would likely result in the type of inventiveness discussed in John Bracht's paper. Lets stay focused. If gene duplication truly results in an increase in dimensionality, you are doing a good job of hiding it.
As paper states:
quote:
As we noted earlier, even the most basic premise of the classical model of duplicate gene evolution--that gene duplicates are preserved only by the evolution of new functions--has never been tested.
Now as Bracht states:
quote:
That re-working of the "rules" of gene expression and embryonic development, along with adding extra genes, is what I'm referring to as an inventive change in biology. I've never seen an example where that sort of change was observed to be the result of Darwinian processes.
Do you have any biological counter examples that escapes the three points I make above?
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 14. February 2003 19:36
AA,
This is going to be partly free association: It's Friday night and I'm tired.
A caution here: I did not mention "adaptive" landscapes. I referred to fitness landscapes, which are abstract surfaces in a high-dimensioned space. The topology of the surface depends on the interaction of the fitness function (the equation that assigns fitness values to points) with an evolutionary operator. "Adaptive" adds some connotational baggage that may not be appropriate. In this domain of inquiry, one must soon abandon the geological metaphors and move to the mathematics of topologies. Three-dimensional visualizations can be misleading and deceptive, as I noted earlier.
AA wrote quote:
One way I perceive these problems (tell me if I am way off or not) is that we can have adaptive landscapes which are more static, governed by static "enviromental rules" dictated by the laws of physics and chemistry. For example, couldn't an adaptive landscape be constructed to show the specificity of a flagellum, in that it always has to have certain parameters just right in order to function? I just don't see those parameters changing, and wouldn't they create a static adaptive landscape? Am I way off here? Appreciated your thoughts!
The question of the flagellum is whether on any of the fitness landscapes induced by the various evolutionary operators there is an incremental pathway to the flagellum from precursors that have some selectable function (not necessarily the final function of the flagellum). (Of course, this leaves out consideration of such phenomena as scaffolding and spandrels as potential contributors, as well as any potential role of neutral mutations later recruited to a selectable function. I speak here only to the selective routes on fitness landscapes; that does not exhaust the possibilities.) The properties of the "solution" (in the flawed search metaphor) do not determine the topology of the fitness landscapes; it is a point (or more properly, a set of neighborhoods) on those landscapes.
In this context, a "fitness function" includes terms describing the allelic variability (and consequent phenotypes exposed to selection) appropriate to all the variables that are selectively relevant, from (relatively) static physical geography to the constantly varying co-evolutionary pressures of, e.g., parasites and predators and prey. That single massive equation determines the fitness of a given phenotype. The fitness values for all possible combinations of alleles (realized as selection-vulnerable phenotypes) is what amounts to an unordered list. When those values are mapped onto the surface induced by an evolutionary operator like point mutations or recombination, where the operator defines nearest neighbors reachable in one operator step, the list takes on a topology. Nearest neighbors of a given point on that surface are accessible to the algorithm in a single step of the evolutionary operator. The nearest neighbors accessible by, say, point mutations are different from the nearest neighbors on the surface induced by recombination. Thus the several fitness landscapes are those induced by the various evolutionary operators; they do not represent different subsets of selective variables or subsets of alleles. In nature there are no "static" fitness landscapes.
In order to demonstrate that the flagellum cannot evolve by incremental steps, then, one must show that there are no smooth gradients in any of the dimensions of the several fitness landscapes that can lead from a functional precursor (not necessarily having the same function as the flagellum) to the flagellum. (Of course, this leaves out consideration of such phenomena as scaffolding and spandrels as potential contributors, as well as ignoring any potential role for neutral mutations later recruited to a selectable function. I speak here only to the selective routes on fitness landscapes; that does not exhaust the possibilities.)
quote:
Also, RBH, please help me to be clear in my understanding of what you wrote: you said trapping a GA on a local maximum is rare because one of the variables is often continuous and can usually help the GA get from peak A to peak b. Is that correct? Does this simply imply that there is often more than one way of doing something? Maybe I missed your point completely.
Not exactly. It means that there is typically more than one way to escape what seems to be a cul de sac or local optimum on that high-dimensioned fitness landscape. Remember, we're not talking about a GA moving. What is evolving is a population spread out on some abstract high-dimensioned surface that is constantly deforming, and subpopulations are scattered around on it with various "paths" available to the various subpopulations. Some of those paths are almost inevitably smooth gradients. If none that lead to "safety" are shallow-enough smooth gradients the population is stuck, and if the selective environment changes too rapidly it is doomed. And thousands and tens of thousands of populations go extinct. Billions and billions of organisms die. Evolution is an enormously wasteful process. And remember, the population is not "searching" for a peak; it's just executing the evolutionary operators, which in biology are biochemical and physical events, and the GA's tendency to find local optima is a by-product of those processes.
Of course, there are some paths that cannot be traveled - they may descend too far into lethal valleys or may be too long for the physical process to make it from here to there in computable time. You can't always get there from here. (See above: Lots go extinct.) There are a slew of contraints on where a population might go depending on everything from the local topology of the fitness landscapes to developmental constraints and the like. Contrary to the (often implicit) belief of some, evolution cannot create anything - it is not infinitely creative - but neither is it as severely limited as others claim.
Finally, AA wrote quote:
By the way, RBH, the concept that the migration of reptilian jawbones --> mammalian earbones is documentd (sic) in the fossil record is somewhat of an "urban legend" among evolutionists, that I think needs to die. There are a few skulls that close the gap on either end of this transition, but the most important part, the actual transition itself, remains a gap:
I'll let the paleontologists and comparative anatomists fight this one out. Even given your references (the most recent 15 years old, the oldest more than 50 years old), I think "urban legend" is a pretty strong overstatement.
RBH
[ 14. February 2003, 19:40: Message edited by: RBH ]
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 14. February 2003 21:54
Rex: Well, we have the Type III Secretory System. If you co-opt that, instead of building it again from scratch, you've gotten halfway there.
No, you don't have the type III secretory system since it most likely evolved from the flagellum. The type III secretory system does not shed any light on the origin of the flagellum.
Unfortunately, there are no "protein fossils". We can only find proteins that exist now. From fossil evidence, we see that something like 99% of all species have gone extinct, and many major phylogenetic groupings have also gone extinct (e.g. dinosaurs). The situation is less clear with protein complexes, but likewise, we might expect that the vast majority of proteins no longer exist anywhere. This is a real barrier to our being able to find a complete step-by-step set of proteins for every process.
Yet diversity is only skin-deep. Functional molecules prone to cooption in a variety of different contexts should be rather immune to disappearing. Take the flagellum. If you want to propose the existence of some transporter that foreshadowed the flagellum, well, bacteria have plenty of uses for transporters (and many different types) that would exist to this day. Yet we are to believe that in a striking feat of coincidence, this once functional transporter completely disappeared at about the time, oh, when the flagellum appeared.
Even more unfortunately, the structure and function is known for only a tiny fraction of the proteins in extant organisms. So even if there are intermediate complexes that we could find, in very very many cases we would expect to not have found such complexes yet.
So it simply isn't fair to demand of evolutionary theory that any protein complex that happens to be known now be explicible in step-by-step detail, using examples from existing protein systems. In fact, the nature of evolution (e.g. most everything dies and is lost) specifically predicts that you won't be able to do that in many cases.
You seem to be trying to have it both ways. If evolution predicts we won't be able to find the precursors for the flagellum, then why the talk about knowing only a tiny fraction of proteins in extant organisms? If evolution is going to make a true specific prediction, let it predict whether or not the precursors exist.
This is why it is exactly backwards to focus attention on ancient complexes like the bacterial flagellum and nuclear pore and so on. You don't expect to be able to fill in every piece in an evolutionary pathway of that age (without vigorous handwaving), and gee, we can't. What would be particularly telling is if we looked at recent complexes and couldn't explain those! That would be a much more serious blow to evolution, especially if you could come up with quantitative predictions for how often you ought to be able to fill in which types of gap given the age of the complex and the number of extant species and the times of their divergence.
This makes sense from a perspective of epistemological evidence (explained on my web page). But if the hypothesis is a follow-up of Crick/Orgel's hypothesis of directed panspermy, the nature of the hypothesis obliges us to focus on things like flagella and nuclear pores. Again, the distinction between EE and OE is essential (from my web page).
Besides, if there is a time limit in which evolution predicts we will not find evidence of its happening, lay it on the table. It doesn't make much sense to complain the flagellum is too old when Miller et al. claim to have evidence for the evolution of the Krebs cycle.
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 15. February 2003 01:50
(Do you mind if I address you as “Alonso”, as there is another “Nelson” posting in these threads?)
Alonso, in the posts I have made in this thread, there have been two (vaguely related) issues.
One issue (which I will return to) is the issue of duplication of structure (in some domain, whether FPGA logic or genetic material), followed by further modification of the duplicated section.
The other issue is the logical form of the argument that Alonso made, and ensuing quotations purported to be in support of the argument. Principle support for Alonso’s argument was the quote (with my emphasis marked), and this quote or related quotes have been repeated several times:
quote:
However, empirical evidence from genome duplication events suggests that gene duplicates are preserved in genomes far more commonly and for periods far in excess of the expectations under [the classical model], and whereas some gene duplicates clearly evolve new functions, there is little evidence that this is the most common mechanism of duplicate-gene preservation. An alternative hypothesis is that gene duplicates are frequently preserved by subfunctionalization, whereby both members of a pair experience degenerative mutations that reduce their joint levels and patterns of activity to that of the single ancestral gene.
Alonso, you argued that this quote is support for a lack of capacity for new functions (in essence) from gene duplication. Please read your own quote again. Do you read the point that you quoted, that “some gene duplicates clearly evolve new functions”?
You will probably try to imply again that this is incredibly rare -- but I have seen no evidence presented from the literature. If you were really going to be successful in supporting the point that such “new functions” are incredibly rare, you would quote statements from the primary literature saying how rare they were, and how that compared to the degree of change that would be needed in evolution to match the rates seen in the fossil evidence. But you repeatedly cite the same argument, discussing a more “common” event, and saying nothing about the (possibly) less common event of my interest. Repeating again the attempt to change the subject to a slightly different subject as though that supported the point will only create greater embarrassment, in my opinion.
I will not match wits with you on “biological examples”, as this is not an area of my expertise. Others more experienced may chime in with examples from biology. My point about your argument was a logical one of whether any information was presented that actually supported your point. My lack of experience in finding biological counter examples is irrelevant, I am quoting your very own quote that you gave in support of your position! Do your quotes actually support your position? I think not.
One aspect of the process of descent with modification, followed by selection is that there is an potential of amplification of an event, even if it occurs with low probability. If a “new function” occurs with low probability (very far from happening as the “most common mechanism” or in a “majority” of cases) it can be amplified because it provides selective advantage. It is the amplification of the lower probability event that is an important aspect of natural selection.
It is because of this amplification provided by selection that allows for creativity from the random event. I understand that the vast majority of changes and genetic drifts, etc., (descent with modification) are not changes that result in “creativity” or “new functionality” or apparent “inventiveness”. (And this is what Alonso’s quotes support -- not a different argument about capacity in the less common case.) For example the evolving focus of the fish eye to hone it to precisely the focal length needed for optimal food gathering and escape from predator -- this would be the type of change that is seen in a very large number of “descent with modification” operations. The lack of apparent “inventiveness” in this vast majority of cases is irrelevant to the question of whether it happens with some low but significant frequency (or probability).
Other quotes by Alonso seem to relate to the paper discussing the issue of what is the “preservation” of “gene duplicates”. The paper, as the quotes seem to show, is discussing the prevalence of that “preservation”. The paper is not directed to the question of whether there are events of somewhat lower probability wherein “some gene duplicates clearly evolve new functions”, but that is the focus of my statements, and the paper quoted clearly seems to support that point Here Alonso quotes again:.
quote:
As we noted earlier, even the most basic premise of the classical model of duplicate gene evolution--that gene duplicates are preserved only by the evolution of new functions--has never been tested.
| | |
