|
Author
|
Topic: Convergence or Divergence?
|
Jay
Member
Member # 268
|
posted 14. July 2002 23:44
Hi Nic,
Thankyou too, for the dialogue. I always find your criticism to be mind-opening and rather productive. You've helped me think through the homology inference in some new ways, and even bothered to talk about this obscure little topic of sliding rings!! Thankyou again.
Nic: "It appears that all agree that making the homology/analogy distinction, with Owen's definitions (basically; IMO we've been even a bit more explicit), makes things somewhat clearer."
Yep. I'd agree. And really, I'd love to see this distinction actually applied more in the homology inference in the literature. Sequence fishing (and general pattern fishing) can be so superficial, and is likely to lead us astray at some point(s), perhaps more than we even suspect. Yours/Owen's distinction will be a key tool in defining between the common design inference and the homology inference. I hope that those who sequence fish (and those who infer common design, of course) will pay heed to it.
It will also be good to have ID critics of your sort participate in the design inference itself, as we need those who can get past this barrier of needing to see the IDer before they'll start thinking about the inference. This ID inference apart from knowing the designer is cool stuff, and it's good to see the critics help refine it... and frankyl, as Mike pointed out, you'd make a heckuva teleologist, should you ever decide to jump the fence ![[Smile]](smile.gif)
Nic: "E.g. it has focused us on the question of whether or not the differences in the sliding clamps are necessitated by whatever differences in function may exist. I agree that we lack the data to make a definitive decision at this point."
Indeed. As with all of these origins inferences, the data is never as clear as we want. So really, I can understand why you might interpret these rings as homlogous, just as I might interpret them as common design. Not that the data is arbitrary, but we each have different standards of proof required for each inference. But it's good to have two people who have different opinions for scientific reasons hash it out!
Nic: "Too me it seems like the function is basically "sliding clamp" (surprise surprise), with the purpose of serving as an attachment to hold other components near to the DNA strand, and it seems that a variety of ring structures could serve this purpose."
Well... as with all of biology, it turns out to not really be that simple. The rings are now showing that they have other small mini-domains within them that that are involved in other areas such as DNA repair, and a variety of other cellular functions. This additional data is likely to be key to understanding the functional reasons why the rings were built different. I don't think it's arbitrary at all, as nothing in the cell ever seems to be! But for now, we're just left with a very general (although quite interesting, IMO) pattern of a functional correlation between DNA replication system and ring type.
Nic: "(euks/arch, eubacteria, viruses -- although are these really one group?)"
Hah! Great question! I'd love to know! The sequence data, and now whole genome comparisons, are flooding in and really challenging some core beliefs about how the larger classifications are related. It will be interesting to see! So far, I suspect that the three domains are separate designs. As for the rest (say, the euks), I have no idea. Sometimes it looks like it, sometimes not. My bet is that there's quite a bit of mixture between common descent and common design, and that they both play a large part. But the pattern is likely to be rather complex, as the patterns we are already getting are hard to fathom.
Nic: "The trick will be to see if any of these correlated differences make the structural differences necessary, or if the clamp differences and group differences are simply both products of billions of years of separated evolution."
Indeed. And I have a feeling I'll be up more than a few nights trying to help figure that out! It's cool stuff, but it can give you a real craving for solid answers sometimes, no? But that's also the allure to it, I suppose.
Anyway, thanks for the discussion. I won't treat any further silence as not wanting to aswer, so don't feel as though any further discussion is aimed at you (although reply if you wish!). I know you're busy. I'll be there soon enough in the next few months!
Thanks,
jay
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 15. July 2002 14:19
Concerning tubulin/ftsZ, Nic asks some good questions that serve as good springboards for ID research.
First, if ftsZ and tubulin are not homologs, why do they both use GTP? As Nic points out, there are only a few options for energy-providing molecules, so its hard to attribute too much meaning to this similarity. What's more, what if ftsZ used ATP while tubulin used GTP? I'm not sure those who make the homology inference would attribute too much meaning to this finding.
Nevertheless, the ID theorist could investigate the possibility that the shared GTP usage has a functional basis. One of the hundred -or-so projects I would like to undertake someday is to functionally categorize nucleotide-binding proteins to determine if some pattern emerges, keeping in mind that billions of years of evolutionary noise may have partially obscured a pattern that may have existed in the first cells. Most G-proteins act as molecular switches, where the GDP-bound state represents "OFF" and the GTP-bound state represents "ON", and the hydrolysis acts as a timing mechanism between states. This makes them very useful in pathways of information flow, explaining why many are involved in signal conduction. From a design perspective, it is not hard to see how such on-off switching would fit into the dynamic instability mechanism employed by tubulin and probably ftsZ. But none of this is enough to tease apart the use of GTP rather than ATP.
Perhaps looking to higher levels may provide a clue. Both ftsZ and tubulin are crucially involved in cell division. And several other GTP-binding proteins are likewise involved. In E. coli, SulA is an SOS-inducible protein that inhibits cell division. It interacts with ftsZ in a GTP-dependent fashion. Obg-like proteins are small GTP-binding proteins that are found in everything from humans to bacteria. In some way, they couple DNA replication to cell growth and cell division. Loss of this protein function results in impaired chromosome partitioning in bacteria. CodY in a protein found in gm+ bacteria that senses GTP levels. It translates a drop in the intracellular pool of GTP as a signal for dividing cells to enter the stationary phase or even sporulate. In eukaryotic cells, excess GTP levels can trigger cell death (apoptosis). Here's an interesting abstract:
quote:
Biochem Biophys Res Commun 1988 Feb 15;150(3):1144-8
Cyclic AMP control of GTP pools in Saccharomyces cerevisiae.
Pall ML.
Program in Genetics and Cell Biology, Washington State University, Pullman 99164-4350.
Previous studies have shown that GTP and cyclic AMP have similar effects on the regulation of sporulation in the yeast Saccharomyces cerevisiae. Declines in either nucleotide can trigger sporulation. These results raise the question whether either nucleotide influences the pool of the other. The current study shows that a cyclic AMP deficiency produces a decline in GTP pools and cyclic AMP readdition quickly increases GTP pools. UTP but not CTP shows a similar pattern of control to that shown by GTP. These results suggest that cyclic AMP effects on sporulation and possibly other cell properties may be mediated in part or in whole by GTP. They provide support for the hypothesis that GTP has a general role in stimulating cellular growth and proliferation.
Another more recent paper from Anticancer Research notes:
quote:
Guanine nucleotides are important substrates for macromolecular synthesis, cell signaling, and integration of metabolic status, and have an evolutionarily conserved role in differentiation, proliferation, and apoptosis. Bacteria, yeast, and mammalian cells are all dependent on an adequate supply of guanylates to maintain proliferation. Depletion of intracellular guanylates, especially by inhibition of de novo synthesis via the IMP dehydrogenase pathway, is a potent signal for inhibition of proliferation, as well as apoptosis. Growth inhibition by depletion of GTP is a conserved pathway from humans to Bacillus.
In other words, since GTP levels are associated with the decision to divide, perhaps the GTP-binding properties of ftsZ and tubulin are involved in one more node of control when it comes to orchestrating such fantastically complex events. I mention this only as speculation for further inquiry and even research.
We then come to the second question raised by Nic - since there are multiple ways to bind GTP, why do tubulin and ftsZ share the same motif. Part of that answer may be tied in up their shared structures coupled to shared basic functions. It is interesting to note that the binding site in both cases is formed at the interface of the dimer. This involves the formation of a cleft between the two subunits which keeps the nucleotide binding site open to the medium, allowing for exchange. When hydrolysis occurs, a conformation change is triggered that cascades outward from the active site. Some speculate that this structural change may bend the GDP filament, where one monomer pushes against the next, thus, generating force on the membrane during cell division.
I mention all this simple because we should not think of the GTP-motif as merely a GTP-binding site. If the function of ftsZ and tubulin was merely to bind up excess GTP (like a sponge) and release it when needed, the shared motif would be hard to explain. But this motif actually serves as the trigger point for coordinated molecular motion that is tied into the overall dynamic of these filaments. Can other motifs truly substitute? An ID research project might attempt to create mosaic FtsZ or tubulin, replacing their GTP-binding sites with other versions. Can this be done without compromising the structure? The function? Here we can see yet another methodological constraint that comes with ID - if we could replace the Gtp-binding from ftsZ/tubulin with another unrelated version, and actually generate a protein with superior abilities, the ID hypothesis would take a big hit.
Nic ended his reply to me as follows:
IMO if this is correct then a few ticks are added to the "similarity in excess of that being necessitated by similar function" side of the scale.
I agree. The question then remains how to interpret these similarities. Are they in excess of functional demands? Only further research and a better understanding of life's complex dynamics will answer these questions. There is so much potential for ID research. It's merely a question of whether enough IDists with the needed skills will ever be able to actualize this potential.
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 16. July 2002 20:28
Mike: You are right there is lots of opportunity for research, I fail to see why you consider this ID research though. Are you saying that the questions asked by ID are dealing with different issues then 'regular' science?
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 16. July 2002 23:50
Frances,
You fail to see why I consider this ID research because you believe the only form of research ID can conduct is to study the designers or their lab-notes. I've already explained how the designer-centric view is a useless research avenue and why it makes no sense to adhere to it. I've just illustrated, with example, how ID can generate research (albeit, one needs to think in subtle terms).
Has anyone in the mainstream scientific community suggested the experiments I have? What would be their purpose, from a purely non-teleological origins perspective? Imagine me writing a grant, where I lay out the basic arguments hashed out by Jay, Nic, and me concerning tubulin/ftsZ in the introduction. Does anyone really believe reviewers would be interested to know whether my design inference would be damaged by replacing the ftsZ/tubulin GTP motifs?
We clearly have a communication impasse. So be it. Rehashing these arguments is not sufficient merit for continuing my violation of the 50 Rule.
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 17. July 2002 01:24
Mike, No on the contrary, I believe that ID research will be limited to trying to disprove alternatives while lacking any positive evidence of ID. Your claim that ID can generate research sounds hollow especially when it fails to show that regular science would not follow such paths.
I agree there may be some communication problems here and I do understand your commitment to ID but I have yet to see why ID would forward scientific thinking in areas in which regular science would not venture. After all, ID being mainly eliminative, it will have to rely on disproving regular scientific hypotheses. Have you tried to submit your ideas for grants or have you given up before making the effort?
I also fail to see why the research would need to happen under an ID umbrella.
I applaud you for thinking up ways to do research and I encourage you to pursue your ideas but I fail to see their relevance to ID perse. But I applaud any effort to increase our understanding of the world around us.
IP: Logged
|
|
Jay
Member
Member # 268
|
posted 17. July 2002 02:23
Hi all,
Mike, I like the research ideas. The more I delve into the specifics of possible ID examples, the more research ideas I start to see, with your ideas being a good part of that!
In looking at these sequence similarities between the ring clamps and the ftsZ/tubulins, I noticed a peculiar pattern in what is *really* related by sequence. I haven't really found any literature addressing these things, and they may simply be artifacts, but I'd like to hear other opinion on it.
I've seen that many making the homology inference in cases like FtsZ and the clamps look for very remote sequence similarity as evidence that the two systems are related. I realize that this is often buttressed with other data, but, taking the logic that sequence similarity - even low stuff - is better suggestive of common ancestry, it seems that we can add quite a few odd members to our list of 'homologous' proteins, thereby underscoring why the homology inference really is rather fuzzy in practice in many cases, and why very low sequence similarity may be misleading in comparing systems like these being discussed. I think that looking into this may also provide another subtle way of looking for common design.
So below, I'll try to point out some weird patterns in sequence similarity between the systems that we've been discussing, and then try to draw some larger speculations from that.
Starting with the Clamps, it has been suggested that the PCNAs and the Beta Clamps are likely related because they look so similar, and also because they (supposedly) share some very remote sequence similarity (which we might expect if they really were related). So remote sequence similarity is used to help infer homology between these in the technical literature (some of it, anyway).
When I BLASTed a PCNA ring sequence against all non-PCNA sequences, however, I found that there were quite a few - hundreds, actually - protein types that shared more sequence similarity to PCNA than the beta clamp and PCNA did to each other! Some of the top-scorers that were annotated are (scores of 33 down):
Aspartate aminotransferase, geranylgeranyl reductase, asparaginyl-tRNA synthetase, glycosyltransferase, prolyl-tRNA synthetase, iron-sulfur binding reductase, ATP-dependent DNA helicase, ribosomal protein S8, etc...
Notice the variety of proteins represented here! Note again that there are hundreds like this that fall closer in sequence to PCNA than the Beta Clamp does.
For visual effect, look at just a portion (so it would fit on the line) of the comparison of a part of PCNA with the non-PCNA sequence that best aligned with it - Acylaminoacyl-peptidase (A.A.P.):
PCNA DLSHIGDAVVISCAKDGVKFSASGELGNGNIKL
______DL+__G+_VVIS__KDG__+___G_L___NI+
A.A.P DLATNGNRVVISATKDGDDYGL-GNLYEVNIET
(underscores added in place of spaces so it would look right on this board)
Now, suppose that it just happened that these two proteins happened to share some similar functions and/or structure. This stretch of sequence similarity would doubtless suddenly be a 'clear' call to homology (after all, it's much clearer than the PCNA/Beta comparison that is currently considered!).
Once more for good measure, when we BLAST the Beta Clamp, we find, for example, that a *yeast* proteasome component is actually one of the top scores against the bacterial Beta Clamp. The following is a 53% identity over 32 amino acids:
PCNA ______LTISFNPTYLIDSLKALNSEKVTISFISAVRP
_________+++SF+PT_LI___K_LN+_K___SFIS_VRP
Proteasome MSLSFSPTALIGLDKNLNAPK--FSFISNVRP
Again, this would be seen as a rather strong argument for homology if common engineering elements were shared. In other words, a common engineering idea seems to give license for looking for common descent evidence in otherwise unappealing places (this sequence similarity, by itself, is not appealing enough to make people suggest that these two are actually related).
But... can we now turn this logic back and argue that perhaps these stretches of sequence similarity in apparently unrelated protein types may be suggestive of good evidence for high sequence convergence, or perhaps even for common design elements in unrelated protein types? Does the logic of sequence similarity significance and common design features run backwards?
I suspect that we may be able to use the fact that the sequences of apparently unrelated proteins sometimes fall between two very similarly engineered systems to help us infer that the sequence similarity between the two systems are not indicative of common descent. For example, in the case of the rings, the fact that hundreds of different protein types look more related to each type than they do to each other indicates that any similarity that they do share in sequence may be due to chance or perhaps even common design.
Also, I believe that it would be well worth looking into some of these apparently unrelated proteins and asking why their sequences are still so similar. I have a feeling that we may find lots of short commonly designed and used sequence stretches (as well as structural motifs) that are interspersed in a variety of unrelated proteins - in other words, lots of small markers of common design.
Thanks,
jay
Note: I went ahead and looked at the tubulins quicky, in case Mike or anyone else is interested.
We find a similar pattern of sequences of unrelated proteins falling between similarly engineered proteins with tubulin/FtsZ. I BLASTed alpha tubulin and came up with these top scores from bacteria (scores of 35 down):
actinomycin synthetase I, folylpolyglutamate synthase, RNA pseudouridylate synthase, NADH dehydrogenase, DAPA aminotransferase, ABC transporter, etc...
These were all found before any FtsZ, in fact, out of 250 hits in bacteria, I only found one hypothetical FtsZ! BLASTing from beta tubulin gave the same kind of results, indicating that many things appear to be much closer to tubulin than FtsZ, in terms of sequence.
[ 17 July 2002, 02:34: Message edited by: Jay ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 18. July 2002 21:55
Aaack, the formatting is all screwed up and I can't edit my previous post for some reason (mods, please delete that one).
Here is the post with minor edits and long links/sequences deleted:
Hi all,
Well, I finally got around to figuring out how blast works myself >-) .
Jay has put up results he got from BLAST indicating lack of sequence relationship between PCNA and beta clamps. But the articles I quoted all refered to the use of the more sensitive PSI-BLAST, rather than BLAST. Neither I nor anyone else in the thread has claimed that BLAST detected similarities, and we've all agreed that the raw percent sequence similarity was no better than chance, so I'm not sure why Jay felt that BLASTing was relevant.
Human PCNA record
[link deleted]
PSI-BLAST works through an iterative procedure (see quote at the end of the post).
A summary of results for iteration 1:
- We get the typical pattern of decaying sequence similarity for organisms traditionally thought to be more distantly related . E.g.:
(percent identical similarity; there is also a "positives" index which is usually much higher, e.g. it might be ~50% for something like Methanococcus)
Humans -- 100%
Rats -- 98%
African clawed frog -- 89%
Anopheles (mosquito) -- 70%
Arabidopsis (plant) -- 66%
Archaeabacteria tend to have values like this:
Methanococcus -- 26%
Pyrococcus -- 25%
Score another point for common descent.
(and BTW, this puts to rest the claim that radical, high-speed sequence divergence is necessary to explain the sequence difference between PCNA and beta clamps...sequence is only minimally conserved even within PCNA clamps)
Interestingly, one bacterial beta clamp shows up on the first iteration; however it is in Thermoplasma:
Thermoplasma: 20%
...which IIRC is peculiar in numerous ways (massive LGT) and therefore may even have a beta clamp that is a chimera with archaeal PCNA clamps.
(I think I posted a quote on this earlier).
Even down at 20% amino acid sequence similarity, the significance E-value is 7e-07 (the default cutoff for a significant result is 0.005)
There are three proteins in the "non-significant but close" category; my very brief check indicated that they didn't seem like clamps of any sort.
So, in the first iteration, all we get is PCNA clamps and one peculiar beta clamp. This is pretty much what Jay would predict based on his previous statements I think.
So, what happens if we click the 2nd iteration button? (this technology is great).
A summary of results for iteration 2:
This time we have 150 hits (we had 113 on the previous IIRC although I'm not looking at that page at the moment).
In the "significant" category we have all the PCNA clamps from before, plus some additional ones. Still, the Thermoplasma beta clamp is the only other clamp detected. However, in the non-significant-but-close category, there are a number of other eubacterial beta-clamps, mixed in with various random non-clamp proteins.
Let's try another iteration.
Iteration 3:
This time there are 190 significant hits.
A quick text search reveals at least a dozen beta clamps are in the "significant" list now, the most significant with an E-value of 4e-12. There are also a number of beta clamps in the non-significant category, with some non-clamp proteins. Notably, no non-clamp proteins appear to be in the "significant" list.
Iteration 4:
240 hits
Best non-Thermoplasma beta clamp: 4e-32
23 beta clamps score better than 1e-10, after that I stopped counting.
The best non-thermoplasma beta clamp has only 11% sequence identity with human PCNA, but PSI-BLAST still gives a very significant result (e-value = 4e-32). Here is the comparison result:
code:
<snip>
[b]Score = 139 bits (350), Expect = 4e-32
Identities = 29/253 (11%), Positives = 88/253 (34%), Gaps = 28/253 (11%)[/b]
<snip>
(forgive the screwy formatting above)
I conclude that Jay is not using the right technique, and that when the right technique is used, his arguments are not supported. Notably, statistically significant similarities to human PCNA are detectable for a large number of bacterial beta clamps, many orders of magnitude above the minimum significance threshold, and many orders of magnitude above non-clamp proteins.
Furthermore, (this is important) the similarities are spread out across the sequence (which IIRC is an important criterion for distant homology, as it doesn't rely on the kind of short stretches of sequence that Jay talks about, which could converge by chance or for a specific molecular interaction).
Here is the blurb on PSI-BLAST:
quote:
http://www.ncbi.nlm.nih.gov/BLAST/Why.html#PSI
PSI-BLAST is designed for more sensitive protein protein similarity searches.
Position-Specific Iterated (PSI)-BLAST is the most sensitive BLAST program, making it useful for finding very distantly related proteins. Use PSI-BLAST when your standard protein-protein BLAST search either failed to find significant hits, or returned hits with descriptions such as "hypothetical protein" or "similar to..."
The first round of PSI-BLAST is a standard protein-protein BLAST search. The program builds a position-specific scoring matrix (PSSM or profile) from an alignment of the sequences returned with Expect values better (lower) than the inclusion threshold (0.005 by default). In the second iteration the PSSM becomes the query in the search. Any new database hits below the inclusion threshold are included in a new PSSM. The PSI-BLAST search is said to have converged when no more new database sequences are added in subsequent iterations.
[...]
This cannot fairly be described as "fishing", and it would be nice if this kind of mildly inflammatory language were not used for what appear to be standard, widely-accepted, peer-reviewed, statistically valid techniques, standard enough that they are included on the NCBI webpage. Such a claim would at the very least have to take on the method on its own statistical level to have any weight at all.
Thanks,
nic
PS: I request that further BLAST-type analyses provide links as I have done; it's quite easy to do.
PPS: Any bets on what PSI-BLAST will get for FtsZ-tubulin?
Chlorobium ftsZ
[link deleted to avoid screen-stretch]
PPS: Short version of parameters:
I used PSI-BLAST, here.
The only defaults I changed in either case were clicking "low complexity filter" (which probably wasn't used anyhow) and increasing the number of sequences displayed to e.g. 500 or 1000.
[ 19 July 2002, 03:10: Message edited by: yersinia ]
IP: Logged
|
|
Jay
Member
Member # 268
|
posted 19. July 2002 13:32
Hi Nic,
Thanks for the info. You are quite right, there are some Beta Clamps that do look significant. I wonder why me and some of these other authors missed those guys! I suppose its because nothing comes up on PSI-I at all, and when you run the Beta and PCNA on a compare-2-sequences BLAST, it still says that nothing significant was found.
Now, I still wonder why so many other proteins are intermixed with these results. For instance, I found things like bacterial fumarate dehydratase, and several Rad proteins, auxin response factor, transposase, beta adaptin, tartate dehydrogenase, etc... all mixed in with the Beta Clamp results. And these were all fairly significant hits, some of them better than any of the Betas. The list grows very fast when you start getting to the less significant results (which still includes some more Betas). All this to say that Beta Clamp and PCNA do not seem to share an exclusive relationship at all down at this level of similarity. There are lots of other 'unrelated' looking things as well. It would be interesting to see if they all share a common folding pattern.
One other point of interest is that the PCNA seems to align in approximately the same place on Beta for each hit. Since Beta is much bigger than PCNA, this means that there is a block of sequence ( >100 aa's ) in the Beta Clamp that does not align to PCNA at all, as far as I can see. If the two rings were once composed of a set of 6 monomers, it raises the question of why only part of the Beta Clamp is still similar to PCNA.
But I'm glad that you pointed out some of these more similar matches. I'll try to jam them into an alignment and see what kind of overall pattern we get. But I'll give it to you that things get fuzzy on the boundaries! It's more fuzzy than I previously thought, with some Betas now showing some significant similarity! Ah well, nothing is easy in this field.
One experiment that we could do to now is to cut out the part of the Beta Clamp that seems to align with PCNA and see if it can fold by itself.
We might propose as a null hypothesis that amino acids that serve as necessary contact points during folding are found in all regions of the Beta Clamp's whole subunit, making it more likely that it was always a whole subunit. In other words, I suspect that the amino acids all over the Beta's subunits function as a whole in an interdependent way during folding and during final structure. This would lead us to possibly suspect that there is more to the difference between Beta and PCNA subunits than we previously thought, or than simple sequence comparison would tell us.
One other similar experiment that we might do (I'll try to do it eventually) is to see if we do see modularity in the subunits of PCNA and Beta. It would be interesting to find out if we see 3 modules in Beta, and 2 in PCNA, as predicted. We could then see if those modules were what aligned best to each other. This would help verify whether the 6-1 hypothesis is likely or not.
Even with the sequence similarity between the two types, the overall architecture still suggests that they may still be independent. But we will need to look into folding patterns (which may be hard to find literature on) and patterns of sequence conservation in the alignments that would suggest whether the subunits could have been built from common modules or not. It will be interesting to see!
Thanks,
jay
[ 19 July 2002, 14:48: Message edited by: Jay ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 19. July 2002 20:20
Jay writes,
quote:
Thanks for the info. You are quite right, there are some Beta Clamps that do look significant. I wonder why me and some of these other authors missed those guys! I suppose its because nothing comes up on PSI-I at all, and when you run the Beta and PCNA on a compare-2-sequences BLAST, it still says that nothing significant was found.
Now, I still wonder why so many other proteins are intermixed with these results. For instance, I found things like bacterial fumarate dehydratase, and several Rad proteins, auxin response factor, transposase, beta adaptin, tartate dehydrogenase, etc... all mixed in with the Beta Clamp results. And these were all fairly significant hits, some of them better than any of the Betas. The list grows very fast when you start getting to the less significant results (which still includes some more Betas). All this to say that Beta Clamp and PCNA do not seem to share an exclusive relationship at all down at this level of similarity. There are lots of other 'unrelated' looking things as well. It would be interesting to see if they all share a common folding pattern.
Are you talking about your results or mine? It was pretty clear in my case that PSI-BLAST first gets you all the PCNA clamps, and the very next thing is lots of beta clamps, with none of these other proteins in between.
Perhaps you're not clear on the difference between BLAST and PSI-BLAST? (sounds like a term from Bablyon 5, I know) The reason that the first iteration of PSI-BLAST doesn't get the beta clamps is that the first iteration is apparently just a regular protein BLAST. The results returned from that search are then used (through what I'm sure is some fairly complex math) to build a position matrix that helps to line up new candidate sequences. New sequences are added in the next iteration and another position matrix calculated, etc. A sequence matrix based on all the PCNAs appears to reliably pull dozens of beta clamps out of the haystack of millions (?) of sequences, far ahead of other proteins.
All this is despite the no-better-than-chance overall sequence similarity (~10%) you've repeatedly mentioned (and cited from several papers).
Regarding your other suggestions, I would expect that PSI-BLASTing just one domain from a PCNA, rather than the whole PCNA, would get you similar results (although there may be a bit more noise with a shorter sequence).
nic
IP: Logged
|
|
|
Printer-friendly view of this topic