|
Author
|
Topic: The Other Flagellum II
|
Pim van Meurs
Member
Member # 541
|
posted 20. September 2003 22:15
Mike: Very interesting. If such a finding extrapolates to biotic reality, this would strengthen the teleological interpretation of FLE, as a designer attempting to front-load multicellular life through unicellular life would need to rely on cooption. That is, cooption, in of itself, ceases to become evidence for a non-teleological origin.
Such remains of course a totally unsupported extrapolation as it fails to establish any relevance to FLE. Lacking any independent evidence of teleology, or front loading, such suggestions are meaningless. Co-option has interestingly enough become a way for ID to merge itself with evolutionary science by moving any issues of teleology to external of the inquiry.
Interesting.
Given that random mutations are ubiquitous as well, could we thus draw the conclusion that this strenghten the teleological interpretation of FLE, as a designer would need to rely on such random mutations? That is random mutations in of itself, ceases to become evidence for a non-teleological origin?
Mike's front loading seems to reject the DI approach of Dembski and instead seems to return to the appearance of design approach.
Murray states these issues quite well
quote:
And so we are compelled to admit that events which display the earmarks of design leave us
in the dark about whether or not the chain of events leading up to designed event came about by
intervention or purely nomically regular processes.
Any appeal to front loading however has to deal with issues such as what is the goal?, how do we establish independent evidence for such a goal?, how do non linearity, chaos theory, contingency fit in with this approach?
Issues of non teleology or teleology become fully separated from scientific inquiry here.
[ 20. September 2003, 23:05: Message edited by: Pim van Meurs ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 20. September 2003 23:10
Mike Gene wrote quote:
Very interesting. If such a finding extrapolates to biotic reality, this would strengthen the teleological interpretation of FLE, as a designer attempting to front-load multicellular life through unicellular life would need to rely on cooption. That is, cooption, in of itself, ceases to become evidence for a non-teleological origin.
The fact of cooption probably does not distinguish between DE and FLE, though as noted earlier, DE allows it while FLE depends on it. So far I see cooption less as a free-standing 'mechanism' than as a natural phenomenon that, as a by-product, often enables the evolution of increasingly complex structures and processes, particularly those that have no obvious "direct" evolutionary pathway. By that I mean that part of what the evolution of 'simpler' structures and processes does is provide 'processed' raw materials - organized subsequences of instructions (or genes) - that are available for recruitment to new functions. My point in the earlier posting about cooption in digital organisms is that it appears that cooption can occur at several levels, which complicates the estimation of some sort of cooption "frequency."
Regardless of whether one can produce a reliable estimate of cooption frequencies at the various levels, though, gedanken's point about missed distributions is strengthened if cooption occurs at several levels of analysis rather than just at the gene level.
RBH
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 21. September 2003 02:16
quote:
quote:
frequencies are established not by just looking at known expressions.
Which is irrelevant. What Ged was confusing was two concepts, a general co-option frequency for all evolution, and a specific co-option frequency for a particular gene.
Alonso, [changed in edit :
It does not matter in fact whether my presentation failed to distinguish a general cooption frequency over all evolution, and over a specific gene. My points are basically that Mike is still trying to take that general cooption frequency and apply it to a specific case (whether based in general frequency for all genes, or general frequency for specific gene).
To look at total cooption frequency indeed involves looking at all expressions, not just known expressions. Looking only at “known” expressions would establish only a lower bound on cooption frequency, which itself has little meaning as I will discus further.
End edit]
And to re-emphasize my point:
One has two problems when one tries to identify a specific “cooption frequency for a particular gene”.
First of all, to do so is to identify that “frequency” in a particular set of circumstances. That gene only exists in one (possibly very long list of) certain circumstances in the first place, and the cooption issue ratio is further constrained in trying to measure it. The measurement (for a “specific gene”) must be in terms of actual observed cases of cooption! This is different from the potential (in terms of probability) for that gene in a specific different circumstance.
And secondly, take the way the explanatory filter works. One has a “specification”, which is given independent of the event. (Key word “independent”—which means that any means of meeting the specification is relevant, not restricting to doing so with a specific gene!!!!!) So the “missed distributions” can include not only cases that actually involve the “specific gene” in question, but other ways of achieving a structure meeting the “specification” without involving changes in that “specific gene” [or its expressions].
Another area of that differentiation between “specific gene” frequencies [as a probability of involvement in cooption in a specific case], and overall cooption frequencies [whether of specific gene generally, or overall cooption frequencies], is that differing circumstances may introduce vastly differing probabilities as compared to a constraint that is wide open to all possible occurrence of cooption. (And Alonso is correctly considering those as differing.) But the problem is then transferring a measurement from one set of constraints discussing a “specific gene” to another condition. The two areas of “missing distributions” occur in that transfer. Not only do changed conditions give rise to differing constraints that change probability, but the possibility of meeting the “specification” in a completely differing manner is ignored. Both cases are “missing distributions” in the evaluation of the EF.
So far from confusing the concepts, I was addressing the relevance of distinguishing the concepts! Show my logic wrong; it does little good to say I was confused about the differentiation existing in the first place when that was my subject.
--
Then to Mike’s point:
quote:
Why not? Nic himself wrote, “You can take almost anything and sooner or later, if you trace it back far enough, you will end up at a change-in-function "event".” If you can do this, and there is a “sooner or later,” there’s something to measure, right?
I have already agreed in principle that one could do such a measurement. I discussed potential methods that included trial genetic engineering to exhaustively discover all gene combinations from potential precursors to see if those gene changes produce a structure that meets the “specification”. Doing so, using all potential precursors, could produce such a “fan-in” probability, when evaluated in a proper Bayesean and/or Markovian perspective.
So is this difficult to do? If so, my point is made.
Do you have another way to verify that one has not missed a relevant “probability distribution” to the explanatory filter natural process pathway?
How are specific “cooption” frequencies limited to specific genes in specific circumstances going to capture this information? Are you defining your problem in a meaningful way?
--
I just realized what RBH (possibly) meant about differing levels of “process” for cooption. Genes could potentially change in different locations and have a very similar affect in some protein or other level of structure. (Remember I’m not a biologist—I shall leave to others any details here including correcting my suggestion.) Differing gene changes might be “expressed” in different combinations of effects, for example one gene might produce particular effect plus other effects, while other gene might produce same first effect but different set of other effects. Hundreds or thousands of minute genetic changes may affect these “different level” changes slowly over long time periods. I assume for the most part that most structure “construction” events are relating to some genetic information at the lowest level for the most part, so the events to lead to the cooption are still genetic changes even if taken at “differing levels”, though certainly not necessarily of specified genes. This gets to the issue above of “missing distributions” and precisely how they could occur through considerations differing from a “specific gene”.
--
Lastly, I highly recommend putting these issues of comparing DE and FLE in terms considered in my thread Does intelligence imply “motive”?. How is any Bayesean comparison being used to support the FLE case? Or is it simply an eliminative argument on the part of DE, having not “eliminated” the case in terms of FLE? That alone does not establish a posterior probability of a designer action—a necessary precursor to FLE as described.
If the “inference” is not the “explanatory filter”, then what is the logic of the inference? Use some mathematics to back it up. At least Dembski describes mathematics and considerable documentation of his method of inference—though it may not hold to logical analysis.
The whole issue of “specification” itself is challenged and analyzed in that thread. We all know that a specific series of 500 coin flips will have (as a whole) a probability of less than 10^-150. Yet that particular series of 500 coin flips can easily be generated if we simply do 500 flips, write down that result as our “specification”, and then claim that we just met that specification. The issue of “independent of the event” and of classifying the event as a “specification” that is independent is the subject of that thread.
What is the “specification” here? If the specification is the particular flagellum, along with some hundreds of parts—say “specific genes” involving sequences more than hundreds long in context—then how is this any more than a “fabrication” by specifying after the event? Sure it is incredibly unlikely, but “run the tape” again and one will get a quite different set of unlikely events.
[ 21. September 2003, 14:22: Message edited by: gedanken ]
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 21. September 2003 04:02
For what it is worth, once multiple genomes of closely and distantly related species come online (e.g. human, chimp, gorilla, mouse, rat, dog, wolf,coyote, etc.), we'll be in a much better position to compute something that looks like a "cooption frequency". In particular, we should get pretty good histories for a large number of genes.
I'm not nearly so pessimistic as some people about the measurement and use of cooption frequencies once we get enough genomic data. In particular, we will have a broad enough perspective to have sampled many species (and therefore many environments) and many genes, which will allow us to estimate average behavior, plus notice if there are classes of genes that seem to get coopted more than others.
If evolutionists have been uninterested in measuring these frequencies so far, it is probably because they couldn't imagine how it could be done. It still can't be, but it will be feasible soon.
However, I really don't see that DE or FLE say anything specific about cooption. Chemistry is sticky. Things cross-react. If being build out of carbon, oxygen, etc., counts as "front loading", okay, but that's not particularly informative about the outcome. You couldn't prevent cooption in DE. Likewise, you could have things front-loaded without cooption; just maintain what you have loaded via error-checking until it becomes ready to use (at which point selection will keep it in). So although extremes of cooption would cast doubt on traditional Darwinian mechanisms, there is a broad range over which both DE and FLE would account for the data. So even though I think we'll get pretty good estimates for cooption in the next decade or so, it'd be rather surprising to me if simply based on frequency we could distinguish DE and FLE hypotheses.
[ 21. September 2003, 04:03: Message edited by: Rex Kerr ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 21. September 2003 10:02
Rex, even Mike says he is not so interested in "general" cooption frequencies. (Yea, there was a little dispute about what "ubiquitous" meant numerically.) Rather he wants to talk about cooption frequencies in very specific constrained situations, and then draw implications from that circumstance differentiating DE and FLE.
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 21. September 2003 11:53
Since I’m up against my self-imposed posting/thread limit, let me offer the final set of replies. I think I’ll start up a cooption thread next month to pick up on some of these issues.
Nelson,
I agree that the concept of FLE can start helping us put some meat on the co-option scenarios. As it stands, co-option is largely treated as a brute given and as something that “just happens.” We think of it largely in anecdotal terms, perhaps because DE gives us no reason to expect a pattern. FLE entails that we take a much closer look at such things. Estimating a gene cooption frequency (gcf) is a place to start. Then, at some point, the “why?” question has to come into play. I.e, why does gene X have a cf 100X higher than gene Y? Etc.
RBH,
Yes, cooption could occur at several different levels and this could complicate efforts to measure a cooption frequency. So let’s say that we have a bacterial cell, which is the natural place to start considering my thesis of FLE. Other than the cooption of genes (and thus gene products), what other levels are we talking about? We can pick it up from there.
Gedanken,
I am getting to feeling that we were talking about different things, so I need to clarify something:
quote:
Rather he wants to talk about cooption frequencies in very specific constrained situations, and then draw implications from that circumstance differentiating DE and FLE.
I am talking about the cooption frequency of any given gene across deep time. This would not entail implications regarding the overall picture with regard to DE and FLE, but it would help to assess the strength of the cooption explanation regarding that gene. Yes, I cannot absolutely rule out cooption as one can always argue, “Sure it is incredibly unlikely, but “run the tape” again and one will get a quite different set of unlikely events.” But I consider this a non-teleological escape hatch. Keep in mind I am not out to convince the non-teleologist of design. I’m simply pointing out that the non-teleological explanation is unsupported and fails to convince me. And this is not because of any bias against the non-teleological explanation, as I explained. This whole issue was raised in relation to the role cooption may have played in the evolution of the IFT machinery described in the OP. And I explained this above:
quote:
The cool thing about all of this is that I can approach this subject is a substantially objective fashion. I’ve already tried to make it clear to people that ID does not impact on my metaphysical views (I have no vested interest in this regard). But since this may be hard for many to swallow, consider this. In this situation, I’m interested in weighing a potential front-loading scenario vs. one that traces back to intelligent intervention. Critics have long wanted to see ID proponents argue with each other. Well, it’s happening in my head.
[and]
I originally took a second look at the eukaryotic flagellum because of the front-loading perspective. I have long used this as a heuristic example to illustrate how FLE could work. And I plan on fleshing this out in more detail fairly soon.
In doing all this, it became clear that the IC challenge posed by the eukaryotic flagellum is much stronger than someone like Miller would have you believe. The most common and powerful method cited to evolve IC entails cooption. But as I have been explaining on ARN, I’m starting to see cooption as something that is more at home with front-loaded evolution than standard neo-Darwinism. So the question then arises in my mind – was the flagellum itself front-loaded or the product of intelligent intervention? In a fuzzy sense, testing for the strength of the cooption explanation provides guidance.
What I am saying is that thus far, the interventionist explanation appears stronger. For many, it’s easy to pigeonhole this with standard “anti-evolution” accusations. But I am also pointing out that this is not simply a matter of my metaphysical or religious background beliefs imposing the answer on me.
Or consider it from another angle. In the original essay on the flagellum, Nic predicted that all the crucial pieces of the IFT machinery would have homologs among the non-ciliary tubulin-based systems. Yet this is not really a prediction of Darwinian evolution. DE would simply predict that “something” from a pre-flagellar cellular state was used that “worked” in “some way” to confer some type of selective benefit. DE is simply not capable of coming up with such a specific prediction. Of course, we can cite other examples of evolution and use these to make the prediction, but that’s not DE doing the predictive work – that’s simply using the way evolution has unfolded to spark analogous thinking. On the other hand, it occurred to me that the specific type prediction Nic made is quite at home with FLE. Here, a designer would not have to endow the original eukarya with flagellar, but simply front-load its existence such that basic cytoskeletal machinery would be likely to spawn such an entity. Thus far, the evidence for such front-loading does not appear to exist. And with this, DE appears to suffer the same fate.
Keep in mind also that I am not trying to force all the data into one explanatory mold. I’m the guy who views evolutionary history as a plausible mosaic, involving intelligent intervention, front-loading, and traditional evolution (in fact, all three are starting to connect up in my mind). That puts me in a fairly unique position to perceive biotic reality.
Rex,
Thanks for expressing your opinions, as you are saying something close to what I have been trying to say. Cooption frequencies (cf) could be estimated with genomic data. It would thus be interesting to see if certain types of genes have a higher cf than others and if there is any other pattern associated with this. But I’m not sure why we need to wait, as there are dozens of closely and distantly related species of bacteria out there. This is also why I don’t think the silence on this issue is simply about an inability to imagine how it could be done. Certainly, it is odd that no evolutionary theorist over the last century has raised the issue and speculated about it in the literature. And we’ve had the bacterial genomes for about a decade now and the idea doesn’t seem to have occurred to anyone. As far as I know, I am the only one who has raised the idea (and as we can see here, the idea is met with tremendous incredulity on the part of several Darwinian evolutionists). And the idea was raised because of the concept of IC and FLE. My point, once again, is to demonstrate that an ID approach does have novel insights to contribute and this raises the question of how much of mainstream science is a function of its working assumptions and ground rules.
Thanks for all the replies, everyone. But I really would like it if someone could address a question I have asked 3-4 times by now – what do you think of the following working assumption –
“Cooption events involving genes (and their products) are most often preceded by an earlier gene duplication.”
Anyone?
[ 21. September 2003, 12:00: Message edited by: Mike Gene ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 21. September 2003 13:27
Mike asks "Cooption events involving genes (and their products) are most often preceded by an earlier gene duplication"
Interesting question to which I do not know the answer but co-option does require that the previous function is somehow released from its constraints and gene duplication would be a suitable mechanism here.
In "Gene co-option in physiological and morphological evolution" True and Caroll argue that this often involves gene duplication.
Annu Rev Cell Dev Biol. 2002;18:53-80. Epub 2002 Apr 02
quote:
Genes can acquire new functions by changes in amino acid coding sequences (including alternative splicing), changes in spatiotemporal patterns of gene regulation, or both. The necessity to maintain the gene’s original function is expected to constitute a strong constraint on that gene’s evolution in most cases. There are three ways by which genes may evolve new functions, even in the presence of
selection to maintain their current function (Figure 1): (a) Changes may occur in parts of the amino acid sequence not required for the current function; i.e., new or alternative protein domains may evolve (Figure 1A); (b) changes may occur in gene regulation, expressing a protein in novel tissues and/or developmental stages (Figure 1B); and (c) gene duplication events may occur, followed by divergence in the amino acid sequences and/or regulatory DNA sequences of the descendant paralogs (Figure 1C,D).
This third mechanism was thought to be required for the evolution of most novel gene functions (Li 1980). However, there is now much evidence that the evolution of novel gene functions via co option can take place with or without gene duplication and that the order in which duplication and adoption of novel functions takes place can vary in different evolutionary trajectories. A fourth mechanism involves cobbling novel genes together from genomic fragments. This phenomenon, of which there is an increasing number of known examples, can involve
either exon shuffling (Long 2001; Gilbert 1978, 1985) or, in some cases, the fortuitous switching of DNA fragments from coding to non-coding and vice versa (e.g., Nurminsky et al. 1998, Long&Langley 1993). In this review we concentrate
on co-option involving extant genes because this appears to be a major mechanism through which developmental diversity in multicellular organisms has arisen.
I am reading Ruse's thought provoking "Darwin and design: does evolution have a purpose". Ruse make the following statement which may be worth pursuing
quote:
“Darwin seems to have expelled design from biology, and yet we still go on using and seemingly needing this way of thinking. We still talk in terms appropriate to conscious intention, whether or not we believe in God. In biology, we still use forward-looking language of a kind that would not be deemed appropriate in physics or chemistry? Why is this? And what does it say about the way humans think? – Michael Ruse
and
quote:
There is no reason to think that biology calls for special life forces over and above the usual processes of physics and chemistry. Nor is there reason to think that biology is little more than complicated physics and chemistry... There does seem to be something distinctive about biological understanding--something having to do with purposes and ends in evolution... We treat organisms--the parts at least--as if they were manufactured, as if they were designed, and then we try to work out their functions. End-directed thinking--teleological thinking--is appropriate in biology because, and only because, organisms seem as if they were manufactured, as if they had been created by an intelligence and put to work... Darwinism does not have design built in as a premise, but the design emerges as Darwinism does its work and some organisms get naturally selected over others. (p. 268-9)
source
The works by Pascal Boyer may also be helpful establishing they why
Cognitive science & neuroscience of religious thought and behaviour Pascal Boyer
Cognitive templates for religious concepts: cross-cultural evidence for recall of counter-intuitive representations
[ 21. September 2003, 13:59: Message edited by: Pim van Meurs ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 21. September 2003 13:52
Mike said:
quote:
quote:
Rather he wants to talk about cooption frequencies in very specific constrained situations, and then draw implications from that circumstance differentiating DE and FLE.
I am talking about the cooption frequency of any given gene across deep time. This would not entail implications regarding the overall picture with regard to DE and FLE, but it would help to assess the strength of the cooption explanation regarding that gene. Yes, I cannot absolutely rule out cooption as one can always argue, “Sure it is incredibly unlikely, but “run the tape” again and one will get a quite different set of unlikely events.” But I consider this a non-teleological escape hatch. Keep in mind I am not out to convince the non-teleologist of design. I’m simply pointing out that the non-teleological explanation is unsupported and fails to convince me. ...
(Quote for several points below)
To clarify, a “cooption explanation regarding that gene” is necessarily a particular circumstance involving the gene. It still has its unique structure, perhaps this is the combination of other genes so that its expression actually changes something relevant, while all other gene combinations did not couple so as to produce that expression result. As Pim points out, it could also mean a “releasing of constraints” in a “duplication” scenario, once again providing a new situational probability in a specific circumstance.
I think that Mike still introduces a possibly intentional confusion by at some points talking as though the subject was to include an objective comparison of FLE and DE, then at other times suggesting only an interest in the subject of DE limited to correctness of very specific aspects (I will discuss further) in a context of a possible future FLE presentation. I think that making this distinction clear (as Mike just did here) is very important. Consider statements like the following:
quote:
If such a finding extrapolates to biotic reality, this would strengthen the teleological interpretation of FLE, as a designer attempting to front-load multicellular life through unicellular life would need to rely on cooption. That is, cooption, in of itself, ceases to become evidence for a non-teleological origin.
This is missing so many aspects of a comparison by any means, eliminative or Bayesean or other comparative, that there is really no “strengthening” of the FLE assumption. The issue of cooption as evidence for DE is one of consistency with the specific pattern seen in DE of sequential branching of the tree according to principles of random processes that are relatively well understood (much to the protestation of ID enthusiasts).
Cooption never was evidence against FLE, rather the lack of any reasonable prior probability assumptions that can be supported for reasons other than philosophical bias is the reason against the FLE assumption.
Furthermore the following issue seems to be a favorite one in ID circles, and is implicit in this explanation of what the thread is about: It may well be that a very specific line can be ruled out as existing in an event. For example if a certain gene mutation cannot be part of any cooption event wherein the end state is the one in question, then clearly that particular gene’s cooption event could not be part of the process steps leading to the end state. That is not an argument against the end state arising by natural processes of cooption, however! For one as RBH has suggested, there are events at other levels that could provide means for that particular gene to participate in some cooption event, that eventually leads to the final state, even though the pathway does not include the directly studied changes. Other circuitous routes are expected. This is back to the IC issue, and in fact many presentations have shown that Miller’s description of cooption in general are still very well supported as argument against IC.
Once again, studying a claim of a particular jump being impossible (IC concept) has little to do with other circuitous routes being possible—nothing has been demonstrated in that regard. That is why it was so important that Mike give the IC definition he was using, so that we could see that by Miller’s own argument the IC had little to do with any claim of difficulty for DE.
In fact, studying a single gene’s cooption rate over long periods of time has very little to do with the probability that it was involved in a cooption event at a particular point in time and circumstance, because those circumstances may have been just the circumstances to raise the probability of that gene’s cooption probability.
Now I see some progress here, however, and Mike may wish to comment on this progress further in his next thread. That progress is that issues of “motive” have been considered at least in passing. This is an indication of not sticking to a strict eliminative approach, but starting to consider aspects of the property of the “designer” behind any FLE approach. This will be useful direction because to produce any sort of non-eliminative comparison, such considerations are required of the “designer” action to establish that the comparison methodology is not severely flawed with false positives. But this is only a start on that direction—one must provide much stronger construction that simply that some particular “motive” was possibly present—I just mention that as a step in the right direction among many such steps that may be useful.
As long as Mike makes no claims that he is actually talking about any evidence for FLE over DE, but is rather just speculating on how FLE could in one part operate (in complete absence of any other part of FLE such as presence of the “front loader” Himself, himself, herself, or itself.) I have no dispute with Mike’s suggestions.
[ 21. September 2003, 14:30: Message edited by: gedanken ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 21. September 2003 14:59
Mike Gene asked quote:
Yes, cooption could occur at several different levels and this could complicate efforts to measure a cooption frequency. So let's say that we have a bacterial cell, which is the natural place to start considering my thesis of FLE. Other than the cooption of genes (and thus gene products), what other levels are we talking about? We can pick it up from there.
I'm not a molecular biologist (and I'll be darned if I'll play one on TV!) but the four "mechanisms" in Pim's quotation from True and Carroll above suggest (at least) two levels: individual whole genes, and gene fragments. The "gene fragments" notion is what I was getting at in my posting of 20. September 2003 11:42: quote:
Rather, it appears that substrings of previously evolved (sub)programs are coopted - a part of the code that evolved to perform a simpler function may be recruited by evolution to contribute to the performance of a more complicated function. Sometimes the part is a single instruction, sometimes a string of several instructions.
I'm wary of direct analogies between specific aspects of the simulation and specific biological structures (for example, I'm not about to identify individual instructions in an evolved Avida program as being directly analogous to "genes"); the analogy (model?) is at a slightly higher level of abstraction: in an evolutionary simulation there appears to be cooption at several levels of analysis, and one can then immediately ask whether that's the case in biology. Fortunately, Pim's True and Carroll quotation provides an answer: Yup, there appears to be cooption at multiple levels of analysis in biology. A next question would be something like are there duplication events that involve multiple genes that are duplicated in one fell swoop? I don't know, but I wouldn't be amazed, particularly in light of Herbert Simon's concept of nearly decomposable systems. Simon's remarks on the relative rate and complexity of evolution in nearly decomposable hierarchical systems (in The Sciences of the Artificial) begin to seem more and more relevant here.
True and Carroll's mention of "the fortuitous switching of DNA fragments from coding to non-coding and vice versa" may find an analog in the Avida simulation, too. Again from incomplete pilot work, it appears that enabling 'junk' instructions to accumulate by not selectively penalizing total critter genome length is (if not absolutely necessary) at least more conducive to the evolution of lineages that perform complex functions. Avida allows setting whether to not penalize total genome length or to not penalize just 'coding' (executed) genome length (the latter therefore penalizes the occurrence of non-coding instructions). From very preliminary results of pilot runs, it appears that the former (which allows 'junk' instructions without selective penalty) leads to more rapid evolution of complex functions. Be aware that since I can only devote one machine to this work at the moment (we're still having trouble with the Beowulf installation), I have only a few runs on which to base that tentative conclusion. Nevertheless, it suggests yet another resource - another source of raw materials, so to speak - for evolutionary processes to draw on.
RBH
[ 21. September 2003, 15:06: Message edited by: RBH ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 21. September 2003 15:08
RBH, you may also check out "Selective Pressures on Genomes in Molecular Evolution" by Ofria, Colier and Adami.
quote:
We describe the evolution of macromolecules as an information transmission process and apply tools from Shannon information theory to it. This allows us to isolate three independent, competing selective pressures that we term compression, transmission, and neutrality selection. The [b]first two affect genome length[/b[: the pressure to conserve resources by compressing the code, and the pressure to acquire additional information that improves the channel, increasing the rate of information transmission into each offspring. Noisy transmission channels (replication with mutations) give rise to a third pressure that acts on the actual encoding of information; it maximizes the fraction of mutations that are neutral with respect to the phenotype. This neutrality selection has important implications for the evolution of evolvability. We demonstrate each selective pressure in experiments with digital organisms.
Neutrality selection, evolvability and robustness... Some fascinating research.
I see neutral evolution as a very relevant force in evolvability, robustness
Another interesting paper
quote:
We have shown that adaptive evolution can take place in the complete absence of what is ordinarily understood as advantageous genotypes. Even if the fitnesses of all viable genotypes are completely identical does selection favor particular regions of genotype space over others. What makes the difference is the density of neutral sequences. In a region of genotype space with a higher density of neutral sequences, chances are higher that a mutated offspring is neutral rather than deleterious. Therefore, neutral sequences in such a region have a higher robustness against mutation, and hence a higher reproductive success. This gives them sufficient selective advantage to outcompete sequences from a less densely connected region. The transitions between different such regions will often occur
in sudden jumps, followed by relatively long periods of stasis. Evolution on neutral networks alone can thus lead to epochal dynamics observed in so many natural and artificial evolving systems.
Adaptive evolution on neutral networks
Claus O. Wilke
[ 21. September 2003, 16:31: Message edited by: Pim van Meurs ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 21. September 2003 17:16
Ged wrote:
quote:
Alonso, [changed in edit :
It does not matter in fact whether my presentation failed to distinguish a general cooption frequency over all evolution, and over a specific gene. My points are basically that Mike is still trying to take that general cooption frequency and apply it to a specific case (whether based in general frequency for all genes, or general frequency for specific gene).
No Mike is not attempting to take any general co-option frequency. Again, I really don't know where you are getting all of this from. Mike has twice now attempted to clarify this to you but you seem to keep comming back to it.
I'll also note that Ged did not reply to my brachyury gene example. He cannot say that
co-option frequencies is impossible/difficult to measure when I gave such a clear example of how one can measure it. So let me get into this a little further. However, I think it is only a matter of time before the moderator zaps this thread for getting way off-topic.
From the phylogeny of deuterostomes we can easily see co-option frequency for the brachyury gene.
As you can see it looks like the 5-star shape of echinoderms is new compared to most of the others. The others are bilateral, even the echinoderm sister group is bilateral. Thus we can
infer that the ancestor of deuterostomes was also bilateral. (this is all of course tentative)
When we look closer at the genetic level we can get a closer look at the co-option frequency. As I already mentioned, the ancestral function seems to be expression in posterior mesoderm and hindgut. In both sea urchin and hemichordate embryos for blastoporal/oral ends of the gut. However, it can get very specific. For example, for sea urchins, the ancestral function seems to be expression in the posterior coelom. Expression
in the middle coelom seems to be an echinoderm + hemichordate co-option. And expression in the anterior coelom seems to be a co-option for this specific clade. This is true of every single one of the clades.
We can even get more precise. What I find interesting about co-option is it's own need to be precise in order to even work. For example, one part of the network (say upstream part) needs to be conserved while another part of the network (say downstream) can change. So co-option, is like myosin. One part of the network stays where it was last put and has a function so that it confers a selective advantage. While the other part moves and if it finds a function, all the network parts may become different from the ancestral state, and we get examples such as with the deuterostomes. You can get a lot more information on this from Davidson's book (my source) "Genomic Regulatory Systems".
Ged writes:
quote:
To look at total cooption frequency indeed involves looking at all expressions, not just
known expressions. Looking only at “known” expressions would establish only a lower bound on cooption frequency, which itself has little meaning as I will discus further.
There is no way to measure a co-option frequency when one cannot see it. Of course if one cannot see it, it may also mean that it did not occur.
Ged writes:
quote:
First of all, to do so is to identify that “frequency” in a particular set of circumstances. That gene only exists in one (possibly very long list of) certain circumstances in the first place, and the cooption issue ratio is further constrained in trying to measure it. The measurement (for a “specific gene”) must be in terms of actual observed cases of cooption! This is different from the potential (in terms of probability) for that gene in a specific different circumstance.
As I stated, there is absolutely no such thing as an observed case of co-option. One can only infer it. This is not a problem however, as most things in Biology are inferred.
Ged writes:
quote:
And secondly, take the way the explanatory filter works. One has a “specification”, which is given independent of the event. (Key word “independent”—which means that any means of meeting the specification is relevant, not restricting to doing so with a specific gene!!!!!) So the “missed distributions” can include not only cases that actually involve the “specific gene” in question, but other ways of achieving a structure meeting the “specification” without involving changes in that “specific gene” [or its expressions].
When we look at specific cases such as the brachyura gene, we can definitely see it performing differently in each clade. So this is not a problem either. For example, with sea urchins, we can see that the radial symmetry arises from an imaginal structure. This is the result of the vestibule and hydrocoel "uniting". The Brachyury gene is expressed in both, and in seperate clades, it seems to have been specifically co-opted for different tasks.
Ged writes:
quote:
Another area of that differentiation between “specific gene” frequencies [as a probability of involvement in cooption in a specific case], and overall cooption frequencies [whether of specific gene generally, or overall cooption frequencies], is that differing circumstances may introduce vastly differing probabilities as compared to a constraint that is wide open to all possible occurrence of cooption. (And Alonso is correctly considering those as
differing.) But the problem is then transferring a measurement from one set of constraints discussing a “specific gene” to another condition. The two areas of “missing distributions” occur in that transfer. Not only do changed conditions give rise to differing constraints that change probability, but the possibility of meeting the “specification” in a completely differing manner is ignored. Both cases are “missing distributions” in the evaluation of the EF.
How can we be missing distributions when the gene is obviously performing differently then it did in the ancestor? Why does measuring co-option frequency require certainty? When we show co-option frequencies over a vast sea of phylogenies, we can then determine whether co-option is truly ubiquitous.
Ged writes:
quote:
So far from confusing the concepts, I was addressing the relevance of distinguishing the concepts! Show my logic wrong; it does little good to say I was confused about the differentiation existing in the first place when that was my subject.
Simple. You were simply saying that one cannot rule out other possibilities when it comes to inferring co-option (and then you oddly bring up motive). That simply means that one can not be certain about his conclusion because there lies the possibility that the there was no co-option event. But this is not about certainty.
[ 21. September 2003, 17:27: Message edited by: Nelson-Alonso ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 21. September 2003 17:49
Nelson hits the nail on the head when he says "There is no way to measure a co-option frequency when one cannot see it. Of course if one cannot see it, it may also mean that it did not occur. "
This shows how accurately measuring co-option frequency is far from trivial. Thus when Ged refers to 'observed' it should be clear in what context this argument is been made.
Nelson wonders: Why does measuring co-option frequency require certainty?
Without some form of certainty, co-option frequency loses much of its meaning and relevance.
Nelson: When we show co-option frequencies over a vast sea of phylogenies, we can then determine whether co-option is truly ubiquitous.
But if these frequencies are inherently unreliable how can we establish the accuracy of out conclusion?
Gedanken puts it quite well " Looking only at “known” expressions would establish only a lower bound on cooption frequency, which itself has little meaning as I will discus further."
If Nelson thinks that calcualtion of co-option frequencies are easy ("He cannot say that co-option frequencies is impossible/difficult to measure when I gave such a clear example of how one can measure it.) then let him provide the relevant calculations based on his brachyury gene example.
Gedanken's remarks are quite relevant and should not be rejected without a supporting argument. Simple doesn't do it...
Motive however is very relevant to Mike's arguments of front loading. Without addressing motive or details of the 'designer' FLE and DE are virtually equivalent. Mike is merely using metaphorical language in his explorations. The issues of motive may be very relevant to the working of the design inference as argued elsewhere by Gedanken.
Or as Ged puts it "This is an indication of not sticking to a strict eliminative approach, but starting to consider aspects of the property of the “designer” behind any FLE approach."
One may use the argument that 'science is always tentative' but without any attempt to define how to accurately measure co-option frequency, we have to but conclude that 'we don't know sofar'. So far there is no real attempt to show that good data for co-option frequency exists, although Nelson does allude to its existence he seems to also consider that there may be many pitfalls and that his original stance that it may be "easy to see co-option frequency' may be more of a qualitative than a quantitative measure.
Nelson points out exactly why motive is important when he states that "DE and FLE can be distinguished in that FLE predicts a pattern" But this pattern can only be identified through motives of the designers since there is nothing inherent in FLE which predicts this pattern. Certainly since FLE co-opts exactly the naturalistic pathways of DE, motive may be the only way to distinguish FLE from DE. Whether or not metaphorical language is used depends on what Mike means with design and teleology. The existence of (metaphorical) design in nature seems be something we can all accept, however the step from this to identifying the designer is what differentiates the two scenarios. If Mike's usage of teleology is merely metaphorical then I have to agree with Gedanken. If Mike's usage of teleology goes beyond the metaphorical then motives do play a significant role. Notice that we are not talking about Mike's motives here btw.
Motives are not relevant to calculating co-option frequencies, but they are relevant to what to do with such data especially once one wants to invoke the design inference.
[ 21. September 2003, 18:38: Message edited by: Pim van Meurs ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 21. September 2003 18:08
Pim writes:
quote:
This shows how accurately measuring co-option frequency is far from trivial. Thus when Ged refers to 'observed' it should be clear in what context this argument is been made.
In some cases measuring co-option may not be trivial. In other cases, it is striking (as with the deuterostomes). Of course these things are always tentative.
Pim writes:
quote:
Without some form of certainty, co-option frequency loses much of its meaning and relevance.
Science is always tentative. So although we can rely on some good data to infer co-option, 100% certainty is not always possible. This is true of just about anything when inferring from history.
Pim writes:
quote:
If Nelson thinks that calcualtion of co-option frequencies are easy (He cannot say that
co-option frequencies is impossible/difficult to measure when I gave such a clear example of how one can measure it.) then let him provide the relevant calculations based on his brachyury gene example.
Inferring it from evidence is not so difficult as to not be doable. I don't really think it's trivial, the difficulty comes in when there is controversy over the cladistic information or when deciding exactly what is the ancestral function of, say, the brachyury gene (but then again, thats why research is fun ![[Wink]](wink.gif) ). I'm not quite sure how one would go about calculating the frequency, as I am simply applying some of Mike's ideas based on some observation of some comparitive analysis. Things are not clear-cut, but thats true of just about anything.
Pim writes:
quote:
Motive however is very relevant to Mike's arguments of front loading. Without addressing motive or details of the 'designer' FLE and DE are virtually equivalent. Mike is merely using metaphorical language in his explorations.
I don't really see how motive is relevant wrt calculating co-option frequencies. That is a seperate philosophical discussion. DE and FLE can be distinguished in that FLE predicts a pattern, as Mike has noted, and has nothing to do with being able to independantly identify the designer. Mike isn't using metaphorical language.
[ 21. September 2003, 18:21: Message edited by: Nelson-Alonso ]
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 21. September 2003 18:49
Mike suggests looking for co-option in bacteria. Unfortunately, the bacterial strains that have been sequenced are not all that closely related (thus yielding lots of "gosh, there's a lot of different stuff here!" conclusions). Most sequenced bacteria are pathogenic, and thus are under extreme selective pressure, share virulence plasmids, etc. etc.; all things considered, not where you would want to start (unless you are studying pathogenic bacteria).
Just be patient. The data will be in much better shape in another decade.
[ 21. September 2003, 23:48: Message edited by: Rex Kerr ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 21. September 2003 22:30
quote:
No Mike is not attempting to take any general co-option frequency. Again, I really don't know where you are getting all of this from. Mike has twice now attempted to clarify this to you but you seem to keep comming back to it.
My apologies to all for any lack in clarity there, which gave Alonso a chance for optimally distant presentation of my statements from their intended meaning.
I was of course speaking of the original comparison of X and a specific case of a specific gene in a specific situation, wherein X was one of these two values:
1) a general cooption frequency over all evolution
2) a “general” (as opposed to specific situation) cooption frequency restricted to all occurrence of any activity of a specific gene, wherein the cases being considered are all cases known of activity (or some sort of prsence) of that gene. (Alonso, “general” has to be taken as a relative word, the rest of the sentence in which it appeared did indeed make the restriction as per Mike’s usage. I was indeed pointing out the “generality” relative to the very specific situation of a very specific occurrence of a specific event involving that gene—using general vs. specific in the normal way those terms are used.)
My intention had been to indicate that in the edit I intended case 2 to be the wording. This was a correction, which did not materially affect my overall argument, but which targeted my response more directly and correctly to Mike’s points. (Since mike indeed did not mean to discuss overall general cooption frequencies, but rather the general frequencies of cooption with the only restriction being the involvement or presence of a specific gene. Once again, my argument is intended to be comparing this case of restricted but otherwise “general” frequency, to the probability of an event in an extremely specific individual case.)
Mike explained “I am talking about the cooption frequency of any given gene across deep time.” This is what I am referring to in “2”, that is what I meant to compare to very specific instances of an event and the probability in such an individual and specific set of circumstances.
Unfortunately my sentence structure allowed Alonso to construe the two cases as though they were the items to be compared, rather than a clarification of the first item. So this allowed Alonso to extract from context and make my statements seem to refer to the overall general measurement of cooption relative to some time or other units (a subject which at some point I had also discussed). By extracting from context which would generally make that clear, Alonso was able to use considerable space debating a point that I was not intending to discuss. So for giving Alonso that opportunity, I once again apologize.
I also apologize to Alonso for any lack of clarity causing him to waste his time debating other points that I did not make. For example my discussion of “missing distributions” is to be taken in the context of the thread that I referred to about that subject, called Does intelligence imply “motive”, which has several discussions of the effects of “missing distributions” in versions of the “Explanatory Filter”. This is specifically referred to by Dr. Dembski on page 67-68 of No Free Lunch, in the sentence “To be sure, this cannot be done with absolute finality since there is always the possibility that some crucial probability distribution was missed.” (Emphasis mine.) Some references to use of “motive” as a consideration of the “designer” that is not simply eliminative is also meant to refer to that thread for context. It is not my intention to debate the subject of that entire thread here, but rather just to make use of some of the results.
At one point Alonso says “But this is not about certainty.” The thread mentioned last paragraph discusses how the eliminative “design inferences” are not simply not about certainty, but how they are also not reliable. The significance factor used in the statistical tests of various versions of the elimination argument is basically irrelevant, as other factors determine the reliability of the explanatory filter completely independent of the significance factor “alpha” for example—making all the statistical significance arguments basically irrelevant.
Thanks to others who seem to have effectively responded to the points. Oh, that appears to be Pim, but others also seem to have “pre” discussed some relevant points, so look back through the posts.
[ 21. September 2003, 23:24: Message edited by: gedanken ]
IP: Logged
|
|
|
Printer-friendly view of this topic