|
Author
|
Topic: Convergence or Divergence?
|
Jay
Member
Member # 268
|
posted 03. July 2002 11:20
Hi Frances,
Frances: "On the other hand, one may argue that evolutionary science suffers from the same problem but as I have argued, it does not rely on elimination of alternatives, rather it proposes a hypothesis which can be tested and falsified.
So I repeat my question: What would ID contribute to science here? Jay suggests that ID could contribute, so I wonder how? And more importantly how would ID be a better explanation than its alternatives? What evidence do we have that indicates design in nature? Unless we can generate some positive hypotheses of ID, what does it contribute? What constrains an intelligent designer?"
Well, I can see that our discussion is rapidly spiraling out into new directions that I don't care to go. Not to be mean, but this is just turning into the same old 'I don't find ID convincing and you need to prove it convincing on my terms or else it's worthless' argument. I've gone over this one dozens of times, and don't care to again.
I wasn't even trying to say that ID never suffered from false positives - I was pointing that the homology inference very likely does. I don't know how we suddenly got hung up on ID and false positives again, but if this is your chief concern, then please start a new thread, as it really is not the point of this one.
Really, the first of two main points of this thread was to point out the very different level of judgement and skepticism between the homology inference and the ID inference. With the homology inference, there are almost never probabilities either for the favored scenario or for competing ones, yet it is still made without shame. On the other hand, the Id inference is routinely raked over the coals by the same 'homology believers' as not being robust enough, or for even having the possibility of making false positives.
Second, the point of this thread was to show that the homology inference could actually *benefit* from a design inference. With the impetus to look for common design, the design theorist would have no problem asking whether two proteins sharing lots of sequence similarity might still be independent designs, and not homologous.
Armed with this suspicion we will see more sensitive tools developed to detect possible similar, but still separate, design between similar proteins.
And this has practical value! Getting good sequence alignments is the headache of everyone in bioinformatics, as it is hard to say where one should cut off their aligned data, as sequences too 'diverse' will lead the data astray. I proposed that a big part of what *really* leads the data astray is not just too much diversity, but that we may be including instances of common design in our data set - a data set that presumes to include only instances of common descent, powered by RM&NS.
So, to develop an ID inference for single, similar genes would be a great boon. It would:
1.) Provide a clearer picture of when and where design happened. There are fewer large IC systems, and so fewer data points to let us determine at what level(s) design happened. But there are thousands of similar protein types, many of which may be from common design. This may let us come up with a more comprehensive Theory of Design, which includes a time chart of design in real history.
2.) It would help clean up sequence alignments. This would lead to better homology models, and so better understanding of the inner workings of protein folding, active sites, allosteric sites, etc... so the design inference could well lead to things like the discovery of new drugs.
__________________
So, the points of this thread are:
1.) that there is a very stark difference in *belief* of the homology inference and the ID inference, and it is not due to amount of evidence.
2.) that the *single-gene* design inference can help the homology inference, and thereby lead to practical benefits, as well as fleshing out ID theory by showing when and where design happened.
These are the points that I would like to stick to, if at all possible.
Thanks,
jay
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 03. July 2002 11:33
Frances writes, "If ICness were a reliable indicator of design, that is, no natural pathways could possibly exist, then identifying IC systems would be enough to infer design."
Perhaps it is just me, but it seems many are not only interpreting "reliable" to mean "infallible," but also interpreting "infer" to mean "prove." As I see it, if no natural pathways could possibly exist, then identifying IC systems would be enough to establish design (not merely infer). Frances claims that the moment false positives are allowed, " ICness or ID inference by relying on eliminative arguments fail." Well, they'd fail as a proof of design, sure. But an investigator is more interested in whether a method is more likelyto generate a false positive than a true positive, not in the possibility that a false positive may occur.
As for my hypothesis of front-loading, many of Frances questions were answered here . Needless to say, the hypothesis was/is not about front-loading "the myriad of species we now recognize."
The important point is whether or not I have answered Frances request for "a positive story, a mechanism, an identifiable designer" or whether we have to now modify that request such that it is "sufficiently detailed" or "sufficiently supported."
Frances: So now we have two competing hypotheses, intelligent design and abiogenesis. What positive hypotheses does ID propose?
For starters, that the first life forms on this planet were exogenous and rather sophisticated entities rather than some simple, sloppy, quasi-life form that was spawned from geochemistry.
How can we detect design if not through elimination of natural pathways.
By looking for the better fit. Do the data better cohere around the notion that the first life forms were rather complex and sophisticated or around the notion they were extremely simple and sloppy?
As for OOL research, it's merely a judgment call for both of us. I don't demand THE answer, but the fact that you don't have THE answer means you don't have "the mechanism." When we look to the OOL, there is no evidence that indicates the Earth did indeed spawn life and no one has the mechanism that is required by the assumption life was indeed spawned from geochemistry. Thus, it's rather odd to hear demands for such things from teleologists, don't y'think?
In what way is the design inference constrained? According to my hypothesis, the data should better cohere around the notion that the first life forms were rather complex and sophisticated rather than the notion they were extremely simple and sloppy. We can also look to the engineers for ideas about constraint. Remember the argument about backward wired retinas?
I should also mention that I did not claim Elsberry's and Wilkins argument about Raelian claims was flawed. I was speaking about Hume's attempt to tease apart life from artifacts. "Much of what we call biology is really nanotechnology ," says Michael J. Heller, a professor of bioengineering (see my web page links page).
Why do I think teleological explanations are superior? Again, it's a judgment call, but I confess to assigning machines and codes to engineering-type causes rather than rock-forming causes. Something about effects of the same kind being assigned to the same causes. Not enough to establish design, but enough to build on.
As for biology's suffocating reliance on teleological concepts/terms, Frances asserts that we need to "realize that the usage of such terms hardly should be used as evidence of design." Why not? If something was indeed designed, we'd expect those trying to figure it out to rely on teleological concepts. That biology, while professing to be non-teleological to the core, dips so deeply from the teleological well to make sense of its object of study (life), follows from the design of life. It's all about Ontological Evidence (which I also discuss on my web page). Of course, I fully concede that such a move could be thoroughly mistaken. But thus far, there is no reason to think so.
Frances writes: Mike also objects to me suggesting that we need to identify the designer or that we need to identify the mechanisms.
First, as I said before, if Frances needs the designer or his/her methods, fine. But to insist that I must think and approach reality as Frances does makes no sense to me. My approach begins with the question of how one infers design without the things Frances needs.
Secondly, whenever someone laments that ID proponents need to "identify the mechanisms," we need to remember the "mechanisms" are protocols, blueprints, or recipes. So...we don't have the lab notes. Why is that significant? How is someone supposed to identify these things? Time travel? Mind reading?
What exactly does Frances suggest I do about identifying designers and mechanisms? It's one thing to naysay. How about some helpful suggestions? I'm not trying to be difficult or slippery. I'm just dealing with reality.
I think I'll let this reply be my last word, as my sidequest is getting further and further away from Jay's original topic and I can clearly see the same old impasses popping up everywhere. But in my mind, this whole business about identifying designers, their motives, and their methods is baseless. My main argument remains:
quote:
But what if life was designed? Does this entail that we should be able to find "unambiguous evidence of a designer." Nope. In fact, I cannot think of one good reason why one would argue, "Since life was designed, it follows we should be able to get our hands on the blueprints or recipes." None. Thus, Frances is elevating epistemology above ontology. That is, even if ID is behind the origin of life, we are to invent other accounts for lack of evidence that doesn't even follow from the design of life. This approach simply makes no sense to me.
Now, I'm happy to speculate about motives and methods, and even use those speculations as a guide to investigate. But like I said, if it is demanded that I establish such things, especially before the investigation begins, we'll simply part company.
[ 03 July 2002, 11:34: Message edited by: Mike Gene ]
IP: Logged
|
|
|
|
Frances
Member
Member # 169
|
posted 03. July 2002 22:57
Jay is worried that the discussion is spiraling into a direction in which the argument merely is that "I do not find ID convincing". But that was not my argument. My argument is that the ID inference is based upon an eliminative argument, thus without any mechanism, causal history, ID does not seem to provide us with much than regular science.
You point out that homology might suffer from false positives and indeed I have no problem with that but homology is not based on elimination of all other alternatives. The reason why ID inferences is 'raked over the coals' is because 1) it is purely eliminative 2) its main proponents of the inference seem to claim that it is without false positives. Once you allow for false positives in the design inference, it has lost its glamor.
I am still wondering why a design inference would benefit science in case of homology. How would design lead to conclusions which would not be following from the usual scientific inquiries?
Indeed, proteins sharing a lot of sequence similarity may not be homologous, but that's a testable hypothesis. How would ID provide us with additional tools?
How would ID be different from scientific inquiry in providing us with a clearer picture how and when design happened? How would it clean up sequence alignments better than scientific approaches?
And most importantly of all, what is the foundation for assuming design? Why is it a better explanation than non-design? How are we going to determine probabilities of designed 'homologies' unless we understand their causal history? But if we understand their causal history what would design add to scientific inquiry? Jay proposes new drugs, perhaps but why would such drugs not be found by scientific inquiry?
As far as the points of the thread, yes there are difference between the homology inference and the design inference, one is based on positive evidence, hypotheses, elimination of such hypotheses, the other one depends intimately on elimination of all competitive hypotheses, does not provide for a mechanism, seems to not be able to help us determine probabilities.
Eventually one may find evidence of design, whether it be internal design or external design. The Raelians surely seem to have placed their hopes on such but so far the evidence of such seems to be lacking.
As Mike points out a design assumption can be helpful to understand the nature of the system and as such it has been used in science, so science itself does not seem to have problems with a design analogy. Of course we need more than analogies to conclude design, the design inference proposed so far are truely eliminative, IC claims that no natural process or chance could have generated the system and thus design can be infered.
Mike's suggestion that life started from first life forms generated by a human like intelligence using bioengineering approaches. How would such an assumption lead to different scientific approaches? How could such an assumption be falsified for instance?
Somehow perhaps the problem in our communication lies in the meaning of the term teleology. Does teleology presume a forward looking goal or does teleology merely identifies an after the fact goal? Depending on its definition I believe that teleology could certainly be merged with science in a fruitful manner. Indeed approaching problems from a system based approach with for instance feedback loops, etc can be a helpful analogy in understanding the systems but they are a tool not necessarily a justification for infering design. We like to think in analogies, we like to look at things as if they have some 'purpose' but that may just be our personal bias.
Mike states that me asking for a designer or methods is not how he goes about infering design. But how is design infered then? Ala Dembski with all its problems or does Mike propose a different approach which deals with the objections raised by for instance Elsberry. Why would design have a privileged position of being the default bucket once chance and regularity have been eliminated? Why not propose probability estimates for the design hypotheses. Of course in order to do this one requires some understanding of causal history, motives, capabilities of the designer(s).
So is there a scientific foundation to intelligent design which differs from current scientific inquiry? Or is the foundation of ID eliminative in nature with all its pitfalls.
As I said before, I am not denying that design assumptions may be helpful in understanding systems by relating them to known systems for which we have built some understanding but analogy is quite limited in that way. People also may be great scientists by relying on their trust in an intelligent design(er) encouraging them to explore the designed world and understand it better.
But my issue is with the issue an intelligent design assumption contribute something new that could not be achieved with regular scientific inquiry methods? And surely I would be the last one from discouraging Mike or others to let their enthusiasm with the design inference lead them to new discoveries. Perhaps we may even be able to determine if there is support for an actual intelligent design and designer(s).
IP: Logged
|
|
Jay
Member
Member # 268
|
posted 04. July 2002 01:09
Hi Frances,
Frances, I appreciate your time in critiquing my ideas, but we clearly are not getting through to each other. I don't want to drag out this arugment again with another reply, and will let you have the last word if you wish, but I feel that we aren't getting anywhere.
But, for the record, I'll respond once more to a few things you have raised. Your main objections "1) it is purely eliminative 2) its main proponents of the inference seem to claim that it is without false positives." are, in my opinion, not very relevant to this thread but are a separate topic. If you feel that you have something here, by all means, post it in your own thread! This is not the place to argue about what others think about ID false positives, or even about ID being purely negative. Frankly, I'm sick of going over those again and again and getting nowhere, which is why I wanted to post some new stuff here at ISCID where the routine bickering is usually not brought up. I'm looking for new insight, not rehashing old, long standing wars.
Frances: "How would design lead to conclusions which would not be following from the usual scientific inquiries?
Indeed, proteins sharing a lot of sequence similarity may not be homologous, but that's a testable hypothesis. How would ID provide us with additional tools?
How would ID be different from scientific inquiry in providing us with a clearer picture how and when design happened? How would it clean up sequence alignments better than scientific approaches?"
I believe that I have addressed these questions more than once already. One of the main points of my thread was explaning why I think that an ID perspective would give us a different viewpoint which, in my opinion, *is* a new tool when looking at things like determining homology. I will leave it to the reader to judge whether my responses were appropriate (and feel free to interject questions!). If someone has a *specific* problem with my reasons why an ID perspective is a useful tool, or why the ID inference is really not relevant to homolgy, and wants to address the reasons that I already gave (or give new ones!), please do so! I'd like to learn something and maybe offer something back!
But anyway, thanks for taking the time to reply, and I hope that we can continue a productive discussion. I'm going to do my best not to respond to old or tangential arguments anymore in this thread, but again, feel free to start your own! Your objections are certainly worth considering!
Thanks,
jay
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 04. July 2002 02:27
Jay, I apologize if I made you feel uncomfortable with my discussion. I believe that looking for an ID filter which would not be purely eliminative would certainly be an improvement.
You mentioned that ID was a new tool and I was curious how this tool would contribute to our knowledge. You made some suggestions and I was interested in the details.
Perhaps we could continue the discussion with Mike Gene's paper? Tubulin versus ftsZ.
IP: Logged
|
|
Jay
Member
Member # 268
|
posted 04. July 2002 13:43
Hi Frances,
I appreciate the kind reply. Thankyou for helping keep this discussion productive! You, as the critic, are a vital part of that.
Yes! Let's discuss Mike's paper! This is exactly the sort of productive discussion I was hoping for. Also, did you get a chance to look at that link to my post about the DNA sliding clamps?
http://www.arn.org/cgi-bin/ubb/ultimatebb.cgi?ubb=get_topic;f=1;t=001728
It seems to kinda touch on some similar stuff, just with a different system that I'm a bit more familiar with. You may get a better discussion out of me with this one, unless Mike wants to help me out with his (and he is more than welcome to do so!)
Both of these examples highlight, to me, some potential hotspots for intelligent common design, rather than common ancestry. In each case, there are large, profound differences (both in construction and in their history) that widely separate the different types, yet there are still some clear similarities. What really piques my interest is that many of the similarities between the systems seems to be more *conceptual* rather than simply hardware linked (which we would expect from common descent). What I mean by this is that they seem to share common ideas, even though their histories suggest that they were probably always separate.
For example, with the DNA sliding rings (PCNA and beta clamps), they give me the impression of being more of a *conceptually* related idea. They really do not share any sequence similarity above normal chance and their history gives us no reason to suspect that they were related (the thread and its replies goes into this) other than that they share a strong conceptual similarity in what they do, and look nearly alike in their final state of folding. They have been around for billions(?) of years, and appear to have always been separate, distinct types. To me, this is an excellent place to start looking at possible common design. The same general argument appears to be true for Mike's excellent example as well.
What do you think? Do you have any specific criticism of either system that leads you to better suspect common ancestry in them? When you look for common ancestry, what signs are necessary and sufficient to suggest common descent, and do these systems meet those criteria?
On the other hand, what about these data might suggest common design to you? How might we spot and interpret any possible relationships by *idea* rather by descent and modification of the actual hardware?
And in the spirit of this thread, how do you feel our inference of common design in these cases compares to the inference of homology? Is one clearly more solid than another? Let's keep in mind when we explore these systems how our inference of common design compares with the corresponding inference of common descent.
Ok... I'm done peppering you with questions. Feel free to only explore the interesting ones. But really, I'd like to know what the critic thinks (that goes for anyone critical here!)
These examples may be excellent lead-ins that can teach us how to spot common design. The specific way that I see these two particular systems (sliding clamps and cytoskeleton components) teaching us is in learning to spot relationship by *idea*, or learning to spot common engineering ideas even when the hardware doesn't suggest a relational history. Other examples may help us spot common design in different ways, but I feel that learning to spot relationship by idea is a very useful concept that is sorely ignored.
Thanks,
jay
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 04. July 2002 20:38
quote:
Both of these examples highlight, to me, some potential hotspots for intelligent common design, rather than common ancestry. In each case, there are large, profound differences (both in construction and in their history) that widely separate the different types, yet there are still some clear similarities. What really piques my interest is that many of the similarities between the systems seems to be more *conceptual* rather than simply hardware linked (which we would expect from common descent). What I mean by this is that they seem to share common ideas, even though their histories suggest that they were probably always separate.
Let me start out by explaining that I do not have any formal training in biology thus I am relying heavily on 'authority' and online papers for my arguments. Learning to understand the evidence and the hypotheses will be a process of trial and error. Nevertheless I like to expand my knowledge.
The problem I have with this argument is that we are assuming how common descent would be evidenced in nature. For instance, we need to understand why the 'common designer' used two very different DNA sequences to come up with very similar structures. Furthermore while the data may not support divergence, it does not mean that the data disproves common descent. Is evidence of convergence, evidence of 'common design'? How do we establish that 'common design' did not have a natural designer?
So lets look at some possibilities
1. Common descent
2. Convergence
3. Common design
I am limiting the possibilities to the three which seem to be most relevant to our discussion. Let's assume that we can eliminate common descent as a possibility although I do believe that there is evidence that would support common descent for these structures. Now we have two different possibilities. How do we establish if what you see as a common conceptual idea resulted from convergence or from common design?
What if we propose that there was not a single common ancestor but perhaps three such ancestors? Would this solve the problem? In this case the common designer might have been natural although we do not address the issue of origins of life here yet. In both your example as well as Mike's example the argument is that there is limited sequence overlap but that the overall three dimensional structure of the proteins is very similar.
quote:
The formation of the FtsZ ring is dependent on GTP and in many ways FtsZ polymerization is similar to tubulin polymerization. From this evidence (and even more convincing evidence discussed later) it is very likely that FtsZ and tubulin share a common evolutionary ancestry - i.e., they are homologous.
How was ancestry detected?
By recognition of a tubulin motif in FTsz
quote:
A relatively new and interesting finding is that FtsZ and tubulin appear to be homologues (3, 17, 20). Tubulin, the key component of eukaryotic microtubules, are nearly universal among eukaryotes and are involved in various cellular functions such as intracellular trafficking, motility, and cell division (1, 20). This idea was first suggested following the initial observation that FtsZ contained the tubulin signature GTP-binding sequence motif, GGGTG(T/S)G (5, 17). This prompted biochemical analysis of FtsZ, which revealed that it too, like tubulin, had a GTPase activity (17, 19). Subsequently, it was also found that it could assemble into protofilaments, two-dimensional sheets, and protofilament rings in in vitrostudies, which was consistent with FtsZ having a cytoskeletal-like function (9). Early conclusions from sequence alignments <../present/alignment.htm> between FtsZ and tubulin lead to a growing notion that these two proteins were homologous, at least over their GTP-binding domains. It was found that they shared substantial sequence identity over their N-terminal GTP-binding domains, but that there was a complete lack of sequence identity over their C-terminal domains. Overall, sequence identity between FtsZs and tubulins is less than 20% (21). Convincing evidence supporting the evolutionary relatedness of FtsZ and tubulin came to light when atomic structures of both proteins recently became available (5, 20). Subsequently, it was discovered through structural alignments that not only were the N-terminal domains almost identical in structure, but there was also significant structural identity over the core C-terminal domains (3). This was an unexpected finding because of the absence of sequence similarity over these regions. Furthermore, it was found that the GTP-binding domain of FtsZ/tubulin are clearly homologous to the GTP-binding domains of other typical GTPase proteins (like p21Ras), which contain Rossman fold structures for GTP-binding (3). Again, this was a surprise because of the lack of sequence homology between the GTP-binding domains of FtsZ/tubulin and other GTPase proteins (3). Therefore, although the sequences involved in nucleotide binding in FtsZ/tubulin are different than those in most GTPase proteins, the similarities between their structures suggest that they are related. Erickson (3) suggested a possible explanation which would account for the different nucleotide binding motifs found in FtsZ/tubulin and most GTPase proteins and the evolutionary relatedness of FtsZ and tubulin. It was speculated that an ancestral protein containing a Rossman fold-like GTP-binding domain diverged in two directions. In one direction it evolved into the typical GTPase family of proteins and in the other, into the atypical GTPase proteins FtsZ and tubulins (3). Alternatively, the similarity in structure between FtsZ and tubulin could be coincidental and there may not exist true homology between these proteins (21). Nevertheless, it is intriguing to think about the evolutionary history that most likely exists between FtsZ and tubulin and especially to wonder, Could FtsZ be the evolutionary precursor to tubulin and microtubules (1, 16)? (see diagram <../present/evolution.htm>)
I found the following references to be very useful
http://www.cellbio.duke.edu/Faculty/Erickson/HPE_pubs.html
http://jcs.biologists.org/cgi/reprint/112/14/2301.pdf
http://www.the-scientist.com/yr2001/nov/palevitz_p18_011126.html
Bacterial ancestry of actin and tubulin
quote:
The low sequence identity between FtsZ and tubulin (10%–18% on the amino acid level) may be a reason to argue that both proteins are the result of convergence rather than true homology. However, their three-dimensional structures are remarkably similar and both proteins exhibit a similar mechanism in their GTP-dependent polymerisation [12]. In both tubulin and FtsZ, a loop (T7) from the neighbouring subunit in a protofilament inserts into the active site and activates GTPase activity, ensuring that hydrolysis only occurs in the polymeric form. The structural and functional properties combined make it unlikely that this evolved twice.
Tubulin-like protofilaments in Ca2F-induced FtsZ sheets
quote:
These examples may be excellent lead-ins that can teach us how to spot common design. The specific way that I see these two particular systems (sliding clamps and cytoskeleton components) teaching us is in learning to spot relationship by *idea*, or learning to spot common engineering ideas even when the hardware doesn't suggest a relational history. Other examples may help us spot common design in different ways, but I feel that learning to spot relationship by idea is a very useful concept that is sorely ignored.
Are you suggesting that convergent evolution is evidence of design in all cases, some cases? Perhaps we can understand this 'common design' as nature finding similar solutions for the same problem?
How do we establish homology? Perhaps by looking at other important proteins in prokayotes and eukaryotes?
quote:
It was thought until recently that bacteria lack the actin or tubulin Ælament networks that organize eukaryotic cytoplasm. However,
we show here that the bacterial MreB protein assembles into Ælaments with a subunit repeat similar to that of F-actin–the
physiological polymer of eukaryotic actin. By elucidating the MreB crystal structure we demonstrate that MreB and actin are very
similar in three dimensions. Moreover, the crystals contain protoÆlaments, allowing visualization of actin-like strands at atomic
resolution. The structure of the MreB protoÆlament is in remarkably good agreement with the model for F-actin, showing that the
proteins assemble in identical orientations. The actin-like properties of MreB explain the Ænding that MreB forms large Æbrous
spirals under the cell membrane of rod-shaped cells, where they are involved in cell-shape determination. Thus, prokaryotes are
now known to possess homologues both of tubulin, namely FtsZ, and of actin.
But on the other hand Hartman et al state that there distinguishing convergence from divergence in this case is going to be hard and that "the best one can surmise" is common ancestry.
[ 04 July 2002, 22:12: Message edited by: Frances ]
IP: Logged
|
|
Jay
Member
Member # 268
|
posted 05. July 2002 13:35
Hi Frances,
Thanks for the reply. I appreciate the info.
Frances: "Furthermore while the data may not support divergence, it does not mean that the data disproves common descent. Is evidence of convergence, evidence of 'common design'? How do we establish that 'common design' did not have a natural designer?"
Frances: "What if we propose that there was not a single common ancestor but perhaps three such ancestors? Would this solve the problem?"
Frances: "Are you suggesting that convergent evolution is evidence of design in all cases, some cases? Perhaps we can understand this 'common design' as nature finding similar solutions for the same problem?"
These statements/questions are excellent, and I'll try to address them here. Please let me know if this reply helps!
Before we even begin, we need to clearly keep a rule of thumb in mind as we go along. I continually pointed out the obscure way that we infer common descent in many proteins to make the point that in making these sorts of origins inferences, the data is not going to be crystal clear. As Mike Gene always points out, we're going to have to be able to tolerate lots of abiguity, especially at first, if we are to work on finding new data patterns.
What I'm interested in finding is data that fits into a sort of paradox from a naturalistic standpoint as my *starting point* in making the design inference. Again, not proof or conlusive evidence of design, but data that makes me start wondering.
Here's the paradox, summarized:
There are certain protein machines that, while appearing to be very similar, clearly strain a common descent relationship. The examples of types that Mike and I posted give the historical appearance of either always being separate, or of somehow having undergone a mysterious rapid and very large divergence once in ancient history. My original post, as well as Mike's, touched on why a simple look at history gives one good reason to think that the types were probably always separate. Their natural history suggests it (and as I pointed out in this thread, it would be even more robust if we were able to determine actual physical constraints within the protein types that would keep them from being related by RM&NS.) Why ancient, crazy amounts of divergence are unlikely is something that I'll touch on more later and that can be found in our original posts.
But, the other side of the paradox is that these systems, while appearing to have always been separate, bear an uncanny resemblance. For the sliding rings, what we have is two ring types (bacterial and archaeal/eukaryotic) that are almost superimposable on each other, even though they share no sequence similarity and are built from different numbers of different looking subunits. Not only that, but both fulfill a very similar design idea, that of being a tether that straps the polymerase to the DNA and slides down the DNA during replication while holding the polymerase, and then snaps open and leaves when replication is complete.
But, as you point out, at this point the paradox still leaves us with two options:
1.) convergence
2.) common design
So how are we to rule out convergence in favor of common design? Well, in truth, we can't ever completely rule it out! It is the same problem that those who assign homolgy face - they can't ever really know that convergence is ruled out. But... the does not stop us from making reasonable inferences still.
The reason that I favor the common design inference in the DNA sliding clamps is due to several different data points that, overall, seem to add up in my mind:
1.) The two different ring types are found between the domains (bacteria, archaea, and eukaryotes). I suspect that the domains are independently designed already for other reasons, the first being that at this level, phylogenetic relationships break down. We no longer get a pattern of nested hierarchies at this level, but rather, we get a pattern of mix-and-match of genes and gene systems. Not to say that this proves common design, but it gets me questioning.
2.) We see the introduction of some novel IC systems in certain domains, and not in others. I suspect that IC systems were intoduced via design, and so finding classes of organisms with a novel IC system not found in other classes leads me to *suspect* possible separate design for that class. This further makes me suspect that these organismal types are indeed separate.
3.) For the protein machines themselves, the pattern of relationship between them is more analogy then homology, and has a certain 'feel' of common design to it. A rather complex, elegant design problem was solved in the same way - amazingly enough, by two differently built machines that look superimposable! We had a design need to keep our DNA polymerase on the DNA while it worked, and we solved it with two very similar ideas.
If these rings were known to be from human engineering, we would certainly assume that a similar *idea* was shared between these two ring machines. In fact, the design idea and even the shape of the machines is so similar that if we had invented one of the ring types, we would likely want to sue whoever it was that built a ring look-alike. And we would probably win. They are not built from the same material - they are built quite differently - but nontheless they share the same concept and even shape. The sharing of concepts is something that designers do all the time. An example is the operating systems for Macs and PCs. Not the same on an actual nuts-and-bolts level (the actual coding is different), but clearly they share the same ideas (the interfaces and operations are very similar). It would be rather unlikely that they both are that similar by chance, but also unlikely that they both shared an actual common ancestor (i.e. the actual code was the same, but subsequent versions slowly diverged away fom each other).
So, we have a previous suspicion that the organism types containing our machines are separate designs, and we then find that the nature of the relationship between the machines themselves reflects design-type sharing of concepts and not actual hardware.
Taken together, these data points lead me to infer common design. Not to say that it disproves common descent or convergence, but to me, it appears to be the *best* explanation. It simply lines up with the data in such a way that I find most useful and likely! Much as the homology inference often does.
And again, compare the strength of this inference to the inference of common descent (keeping in mind how little the common descent inference often really knows). Then ask yourself if such common design inferences ought to be allowed to be explored at least in theoretical journals! And ask whether, perhaps, the common design inference really isn't so far out in left field after all.
And again, this is worth pursuing. Finding good instances of common design can lead us to question and develop models for determining more of what may be common design. This would give us a competing explanation for homology that was robust against high sequence similarity (convergence is not robust against this, as far as we can tell). Having this would actually sharpen our inference of homology, as we would now have to rigorously rule out a competing explanation, making us think through our homology inference a little better. And once again, being able to rationally rule out some high sequence similarity homology inferences could be a great boon for those in bioinformatics, as it would clean out lots of misleading data from alignments. This is a rather practical exercise, but it is still in its infancy, and so will have to be hammered out quite a bit more.
Thanks,
jay
[ 05 July 2002, 13:42: Message edited by: Jay ]
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 05. July 2002 13:43
The abstracts and links provided by Frances are informative for many reasons.
First, they illustrate the problem of citing scientific opinion/conclusion with regard to the debates we have here. Frances correctly cites the three possibilities that are relevant to our discussion: 1. Common descent
2. Convergence and 3. Common design. But we need to remember that when we read the scientific opinions/conclusions, they were not arrived at by considering all three possibilities. In every case, its a binary decision between 1 and 2. Thus, any opinion/conclusion that comes down in favor of 1 (for example) is simply playing against 2, not 2 and 3. In other words, the scientific arguments provide no basis for choosing 1 (or 2) over 3, as that dispute is not even on the table.
Because science excludes possibility 3 from the start, appealing to science in our discussions is simple question begging. Consider the claim, "Thus, prokaryotes are now known to possess homologues both of tubulin, namely FtsZ, and of actin." This is not known. This is an interpretation of the structural similarity. Yet for our discussions, why should we conclude that structural similarity establishes a homologous relationship? The evidence for the homology of the two proteins is the structural similarity. Yet possibility 3 can also explain this (on my web page article, I suggest the mechanism of rational protein design).
Hopefully people can begin to appreciate the confusion that can arise from using scientific authority in our discussions, given that science doesn't even address the issues we raise in our discussions.
Secondly, even by considering the scientific argument divorced from our discussion, it should be clear how fuzzy this all is. For example, Frances quotes, " Furthermore, it was found that the GTP-binding domain of FtsZ/tubulin are clearly homologous to the GTP-binding domains of other typical GTPase proteins (like p21Ras), which contain Rossman fold structures for GTP-binding." But this clearly homologous relationship seems to be person-dependent, as Kull and Fletterick argue: “the conclusion that the ftsZ fold is related to that of the G proteins is misleading.....the two protein families should not be thought of as being related, at least in a direct evolutionary manner.” (Trends Cell Bio 8:306-307). These homology arguments must be fuzzy if what is clear to one person is misleading to another.
In the same quote that Frances provides, we find "A relatively new and interesting finding is that FtsZ and tubulin appear to be homologues" and "Alternatively, the similarity in structure between FtsZ and tubulin could be coincidental and there may not exist true homology between these proteins." In other words, no one has ruled out coincidental relationship. Yet many now proceed with the conclusion that ftsZ and tubulin are related. In fact, Frances talks about the ability of the mainstream view to be predictive, rather than eliminative. Well, concerning these two proteins, what do possibility 1 and 2 each predict in a positive sense? The fact that scientists can't rule out 2 indicates there are no robust predictions that distinguish the two. Instead, we rely on an intuition about whether the similarities are too uncanny for coincidence.
And this takes us to the third point. One quote concludes, "The structural and functional properties combined make it unlikely that this evolved twice." Yet where is the probability estimate? Where are the calculation behind the "unlikely" claim?
When it comes to science distinguishing between 1 and 2, we find no calculations, no successful attempts to rule out one or the other (intolerance of false positives), and no robust predictions that help us tell 1 from 2. But when it is time for 3 to join this list, suddenly the rules are changes and we now expect 3 to provide what 1 and 2 have failed to provide. Since 3, like 1 and 2, can't provide them, we toss out 3, and return to 1 vs. 2. Go figure.
Finally, let me submit an interesting take. What is behind the conclusion of homology? I suggest it is simple an analogical argument: "which takes note of the fact that two or more things are similar in some respects and concludes that they are probably also similar in some further respect." Clearly, scientists look for similarities between ftsZ and tubulin. On the basis of these similarities, they propose another similarity - ftsZ and tubulin have a common ancestor.
IP: Logged
|
|
|
|
Frances
Member
Member # 169
|
posted 05. July 2002 16:33
Mike Behe states that
quote:
... we need to remember that when we read the scientific opinions/conclusions, they were not arrived at by considering all three possibilities. In every case, its a binary decision between 1 and 2. Thus, any opinion/conclusion that comes down in favor of 1 (for example) is simply playing against 2, not 2 and 3. In other words, the scientific arguments provide no basis for choosing 1 (or 2) over 3, as that dispute is not even on the table.
I disagree since if 1 can be shown to not be supported 2 and 3 automatically become potential candidates. When and if evidence supporting 3 becomes available it will become a likely candidate but to state that it is not even considered is begging the question. One of the papers for instance suggested that we very well end up with Woese's idea of multiple separate origins for instance.
My question remains, how would we be able to formulate and falsify an argument that claims 'common design' and does common design include or not include natural mechanisms? For instance if we were to accept Woese's ideas, would this not include 'common design' in its widest meaning of the term?
quote:
Consider the claim, "Thus, prokaryotes are now known to possess homologues both of tubulin, namely FtsZ, and of actin." This is not known.
It's infered from the data.
quote:
This is an interpretation of the structural similarity. Yet for our discussions, why should we conclude that structural similarity establishes a homologous relationship?
I would say because at the moment, combined with all the data it provides for the best hypothesis.
And it's not just structural similarity, which of course is an important factor but it also includes the finding that the tubulin signature GTP-binding sequence motif, GGGTG(T/S)G is found in FTSZ.
Thus the hypothesis that
quote:
Erickson (3) suggested a possible explanation which would account for the different nucleotide binding motifs found in FtsZ/tubulin and most GTPase proteins and the evolutionary relatedness of FtsZ and tubulin. It was speculated that an ancestral protein containing a Rossman fold-like GTP-binding domain diverged in two directions. In one direction it evolved into the typical GTPase family of proteins and in the other, into the atypical GTPase proteins FtsZ and tubulins (3). Alternatively, the similarity in structure between FtsZ and tubulin could be coincidental and there may not exist true homology between these proteins (21). Nevertheless, it is intriguing to think about the evolutionary history that most likely exists between FtsZ and tubulin and especially to wonder, Could FtsZ be the evolutionary precursor to tubulin and microtub
ules?
(3) Erickson, H.P., (1998). Trends in Cell Biology. Atomic structures of tubulin and FtsZ. 8: 133-137.
From (3)
quote:
The GxxNxD sequence defining this synergy loop
is almost completely conserved in tubulins and FtsZ (Fig. 1), and a surprisingly large number of functional mutations of tubulin also map to this region.
...
It is important to note that the synergy model is
still just a hypothesis. The next step will be to test it by molecular modelling of the interfaces to determine how close the synergy amino acids can come to the GTP. However, this specific hypothesis is also an illustration of a more general potential. Molecular mechanisms like this, which were completely inaccessible before the atomic structure, can now be proposed and tested in detail.
quote:
The evidence for the homology of the two proteins is the structural similarity. Yet possibility 3 can also explain this (on my web page article, I suggest the mechanism of rational protein design).
I would argue that unless we have an understanding of the design pathways, 3 could almost explain anything. Homology: Intelligent design reuses parts, analogy: Intelligent design reuses a design 'idea'. How can we constrain 3 to differentiate it from homology/analogy? My argument is that unless we can do this, 3 will always be part of 1/2 and is thus NOT ignored.
quote:
Secondly, even by considering the scientific argument divorced from our discussion, it should be clear how fuzzy this all is. For example, Frances quotes, " Furthermore, it was found that the GTP-binding domain of FtsZ/tubulin are clearly homologous to the GTP-binding domains of other typical GTPase proteins (like p21Ras), which contain Rossman fold structures for GTP-binding." But this clearly homologous relationship seems to be person-dependent, as Kull and Fletterick argue: “the conclusion that the ftsZ fold is related to that of the G proteins is misleading.....the two protein families should not be thought of as being related, at least in a direct evolutionary manner.” (Trends Cell Bio 8:306-307). These homology arguments must be fuzzy if what is clear to one person is misleading to another.
First of all the existance of controversy in science is quite common so to suggest that this is a surprise hardly is convincing evidence. I have not found the paper Mike is refering to but the disclaimer 'at least in a direct evolutionary matter' seems to be an important one.
quote:
The existence of an extra nucleotide-binding region that comes into place upon polymerization and is most likely directly involved in activation of GTP hydrolysis, further reinforces the concept that tubulin and FtsZ form a distinct
family of GTP-hydrolysis proteins.
From
Tubulin and FtsZ form a distinct family of
GTPases , Eva Nogales1, Kenneth H. Downing, Linda A. Amos, and Jan Löwe, nature structural biology . volume 5 number 6 . june 1998
pp. 451-458
Infering homology and ancestry is not always straightforward but
The case for a common ancestor: kinesin and myosin motor proteins and G proteins provides us with some guidelines.
quote:
We conclude that kinesin and myosin, and possibly G proteins, are probably directly related via divergent evolution
from a common core nucleotide-binding motif, and describe the likely topology of this ancestor.
quote:
In the same quote that Frances provides, we find "A relatively new and interesting finding is that FtsZ and tubulin appear to be homologues" and "Alternatively, the similarity in structure between FtsZ and tubulin could be coincidental and there may not exist true homology between these proteins." In other words, no one has ruled out coincidental relationship.
Science hardly has the answers to all questions. Additional research is needed to support these 'recent findings'.
quote:
Yet many now proceed with the conclusion that ftsZ and tubulin are related.
Hypothesis Mike, hypothesis. So far I have found several predictions made for FTSZ and tubulin relating to common ancestry. Perhaps its time to formulate some for common design as well?
good overview article
And of course this intruiging one:
The design plan of kinesin motors
[ 05 July 2002, 16:37: Message edited by: Frances ]
IP: Logged
|
|
|
|
Jay
Member
Member # 268
|
posted 05. July 2002 23:45
Hi all,
Mike, you pretty much summed up much what I was trying to get at in this thread!
Frances: "My question remains, how would we be able to formulate and falsify an argument that claims 'common design' and does common design include or not include natural mechanisms?"
Perhaps a better question is to ask is *when and where* we are able to detect common design, rather than falsifying the possibility of it. After all, homology could be design re-use of parts, as could analogy. The reason we're looking into these systems here is that it may be easier for us to detect any latent common design, and pull this data out and distinguish it from common descent or convergence. It may still be there in other systems, but may be too hard to distinguish.
In response to Mike's reply that we don't know of FtsZ and tubulin are homologous, you say: "It's infered from the data."
*How* is it inferred from the data? Why is this explanation better than convergence or common design?
You reply to this question: "I would say because at the moment, combined with all the data it provides for the best hypothesis.
And it's not just structural similarity, which of course is an important factor but it also includes the finding that the tubulin signature GTP-binding sequence motif, GGGTG(T/S)G is found in FTSZ."
But now you're basically proving the stereotype argument I presented at the start of this thread - that homology is usually infered simply because there is some structural/functional similarity sprinkled with some amino acid similarity. But here's the *real* question - why should that lead us to infer homology? Why is this a better explanation than convergence or common design? Those you quote say: "Alternatively, the similarity in structure between FtsZ and tubulin could be coincidental and there may not exist true homology between these proteins (21). Nevertheless, it is intriguing to think about the evolutionary history that most likely exists between FtsZ and tubulin and especially to wonder, Could FtsZ be the evolutionary precursor to tubulin and microtubles"
But again, where does this 'most likely exists' in reference to homology come from? Have these scientists:
1.) ruled out convergence by testing and eliminating all possible convergence scenarios?
2.) shown that divergence is even remotely possible? Do they really know the steps that it takes to diverge from FtsZ to tubulin? Do they even have the slightest clue?? If not, how can they rule it most likely?
3.) Even considered - ever - the possibility of common design for *any* system? In fact, has this ever been considered in any journal anywhere no matter how speculative? Or have we ruled it out already without ever bringing it up? Are we not in the least bit concerned about getting false positives for naturalistic scenarios?
Frances: "I would argue that unless we have an understanding of the design pathways, 3 could almost explain anything."
And I would argue that unless we have an understanding of what it takes to diverge between things like FtsZ and tubulin, homology could explain anything with any sort of similarity. And again, see above as to why we're not interested here in whether design could explain everything - we're interested in whether it is a better explanation for certain systems. This is much different.
Frances: "First of all the existance of controversy in science is quite common so to suggest that this is a surprise hardly is convincing evidence."
And why all of the controversy? Often, it suggests a problem. And, here's what I think Mike was getting at - are we sure that, in this controversy, we are fairly considering all options? I don't think we are.
Thanks,
jay
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 06. July 2002 00:45
Jay states that quote:
And why all of the controversy? Often, it suggests a problem. And, here's what I think Mike was getting at - are we sure that, in this controversy, we are fairly considering all options? I don't think we are.
Based on what evidence have you come to this decision? Where was common design eliminated or for that matter convergence? The best hypothesis seems to be at present homology based on the sequence motifs and the structural simularity. Plausible evolutionary pathways have been formulated and will be tested. If homology fails to be supported then we still have convergence or common design. Of course both will be hard to test or eliminate, especially common design due to lack of a mechanism allowing us to determine probability. I fail to see at this point why we should entertain common design unless we include common natural designs which would include for instance multiple common ancestors ala Woese. The evidence linking eucaryotes to (their ancestral) prokaryotes seems to be growing though.
So far the argument seems to be that the sequence alignment showed low similarity yet the two proteins showed strong similarity and shared motifs. Seems to be the best working hypothesis until new data becomes available.
Perhaps in the mean time you can explain to me what the common design hypothesis is in a non ad-hoc manner. That is how do we determine if homology or analogy is evidence of a design? We have to explain why a designer would use two very different sequences for the same end result for instance.
We also need to understand why other examples of analogy are or are not evidence of common design.
For instance
quote:
The bacterial cell division protein FtsZ is a homolog of tubulin, but it has not been determined whether FtsZ polymers are structurally related to the microtubule lattice. In the present study, we have obtained high-resolution electron micrographs of two FtsZ polymers that show remarkable similarity to tubulin polymers. The first is a two-dimensional sheet of protofilaments with a lattice very similar to that of the microtubule wall. The second is a miniring, consisting of a single protofilament in a sharply curved, planar conformation. FtsZ minirings are very similar to tubulin rings that are formed upon disassembly of microtubules but are about half the diameter. This suggests that the curved conformation occurs at every FtsZ subunit, but in tubulin rings the conformation occurs at either beta- or alpha-tubulin subunits but not both. We conclude that the functional polymer of FtsZ in bacterial cell division is a long thin sheet of protofilaments. There is sufficient FtsZ in Escherichia coli to form a protofilament that encircles the cell 20 times. The similarity of polymers formed by FtsZ and tubulin implies that the protofilament sheet is an ancient cytoskeletal system, originally functioning in bacterial cell division and later modified to make microtubules.
In all fairness the issue seems hardly resolved
quote:
In the case of sequence comparisons, for example, misato has a higher overall sequence similarity to tubulins than does the bacterial FtsZ protein, which has been proposed to be a tubulin homolog on the grounds that it is a GTPase, forms microtubule-like structures, and has some sequence similarity to tubulins (20). However, some enzymes can have properties strikingly similar to those of cytoskeletal proteins, even though they are not thought to be related to tubulins in an evolutionary sense. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) has as much sequence similarity to tubulins as does FtsZ, and GAPDH can bind GTP, actin filaments, microtubules, and colchicine. Furthermore, GAPDH can polymerize to form cytoskeleton-like oligomers (21). These data highlight the difficulties of assigning homologies, sensu strictu, to one of the best-characterized cytoskeletal protein families (22). They also put into perspective the emerging multifunctionality of proteins that have so far been considered to have only a single conserved property.
Also see this one
quote:
The chymotrypsin family of serine proteases is notable in being both ancient and large (Table 1), but the extreme proliferation appears to be confined to eukaryotes; only rarely are family members found in bacteria. This raises the possibility that other families that appear to be confined to certain branches of the tree of life are actually more ancient, but that they have
simply become extinct in other lineages, or that a relationship has gone undetected.
The latter is the case for eukaryotic tubulin and bacterial FtsZ, both of which use GTP for polymerization to form similar intracellular fibers and are believed to be ancestrally related (40). This relationship was not detected by pairwise sequence comparisons, but rather by recognition of a tubulin motif in FtsZ. Potentially homologous proteins have also been identified by structure determination, such as the detection of similar folds for kinesin and myosin motor proteins (41).
From: Gene Families: The Taxonomy of Protein
Paralogs and Chimeras Steven Henikoff, Elizabeth A. Greene, Shmuel Pietrokovski, Peer Bork,
Teresa K. Attwood, Leroy Hood
SCIENCE VOL. 278 24 OCTOBER 1997
[ 06 July 2002, 01:13: Message edited by: Frances ]
IP: Logged
|
|
|
Printer-friendly view of this topic