|
Author
|
Topic: Ka/Ks and positive selection
|
Kirk Durston
Member
Member # 174
|
posted 08. March 2002 11:34
A prediction of ID: For widely divergent orthologues, the ratio of non-synonymous (Ka) to synonymous (Ks) substitutious may be significantly larger than 1. Orthologues that are not too divergent can be fine tuned to an organism by natural selection, but if the tuning requires a large amount of non-synonymous substituions, ID will adjust those while leaving the rest alone. The way to test for this is to see if there is any correlation between Ka/Ks ratios that are significantly larger than 1 and widely divergent orthologues. This occurred to me after reading, 'Positive selection of a gene family during the emergence of humans
and African apes' by Johnson et al. in Nature
vol. 413, p. 514 (2001) where such seems to be the case. It is referred to as 'positive selection'. I think we ought to regard 'positive selection' as a very strong indicator of ID. In other words, ID predicts positive selection, Darwinism does not.
IP: Logged
|
|
Alex Wild
Member
Member # 165
|
posted 08. March 2002 12:08
quote:
Originally posted by Kirk Durston:
In other words, ID predicts positive selection, Darwinism does not.
This comment doesn't make sense to me. Darwinian evolution very explicity is a theory of positive selection, one that not only predicts it but supplies a physical mechanism for it. Am I missing something here?
IP: Logged
|
|
Kirk Durston
Member
Member # 174
|
posted 08. March 2002 16:55
Darwinism is a theory of natural selection, not positive selection. Non-synonymous substitutions may affect the gene product, synonymous substitutions won't. Natural selection, therefore, should restrict non-synonymous substitutions but not synonymous substitutions, which will accumulate at the normal mutation rate for that exon. For this reason, Ka/Ks ratios governed by natural selection should be less than 1 for functional, evolving genes and this is usually the case. Ka/Ks ratios significantly higher than 1 are evidence for positive selection. In one case, in the paper I cited, the Ka/Ks ratio was a phenonemal 13! Positive selection produces the opposite effect of natural selection so far as Ka/Ks ratios are concerned and, thus, run counter to predictions made under Darwinism, but are exactly would ID would predict. Thus, whenever one observes an instance of positive selection, one can conclude that it didn't get there by natural selection. Something is 'not bothering' with changes that don't make any difference. Another way of putting it is that non-synonymous changes are occuring, or occurred, faster than the mutation rate for that exon. Under Darwinism, Ka/Ks ratios are restrained by natural selection and therefore should be less than 1.
[ 08 March 2002, 17:03: Message edited by: Kirk Durston ]
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 08. March 2002 21:26
I agree with Alex, here. Sort of.
I don't think high ratios are outside of Darwinian predictions. Darwinists certainly don't think so! Adaptive, beneficial mutation are selected! This is non-random and presumably driven by some fitness function.
Of course high ratios could be consistent with some ID hypothesis, too. Perhaps Kirk's idea that large Ka/Ks ratios can be ID indicator can be combined with Behe's IC concept. When a Ka/Ks ratio greater than one produces an IC feature . . .
Just a thought.
IP: Logged
|
|
Kirk Durston
Member
Member # 174
|
posted 12. March 2002 20:44
Let me put it this way: under natural selection, all synonymous mutations are permitted. Not all non-synonymous mutations are permitted, some are eliminated from the population. Given this, the highest that the Ka/Ks ratio should get under natural selection is 1, and this would be for the unrealistic situation where no non-synonymous mutations were eliminated. If Darwinism wants to predict Ka/Ks ratios significantly larger than 1, then they must put a mechanism on the table that explains why non-synonymous substituions are permitted even more than mutations that don't make a diffference. Keep in mind that Ka and Ks represent substitutions *per site*, not total number of substitutions.
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 12. March 2002 22:49
Kirk:
rates of synonymous substitutions tend to be quite constant, and subject to genetic drift (new mutations are generated, old mutations "lost"). Rates are usually in the 10^-9/site/year range.
On the other hand, rates of non-synonymous substitutions are very variable, because of course they are subject to different levels of selection.
Thus, darwinian theory would predict that, especially for recently diverged genes subject to intense selection, the ratio of Ka/Ks can indeed largely exceed 1. This seems to be the case of the hominoid gene family you mention in your first post.
[ 12 March 2002, 22:52: Message edited by: charlie d. ]
IP: Logged
|
|
Moderator
Administrator
Member # 1
|
posted 12. March 2002 23:19
Charlie,
quote:
On the other hand, rates of non-synonymous substitutions are very variable, because of course they are subject to different levels of selection.
Ok, this sentence just doesn't make sense to me. How do levels of selection affect mutation rates? I can understand how selection might "weed out" some nonsynonymous mutations, but it cannot increase them relative to synonymous mutations.
In fact, I think almost nobody is understanding what Kirk is trying to say, so I'll try to re-state it as I understand it (feel free to correct me if I'm wrong, Kirk!)
Assuming that there is a 50/50 chance of getting a synonymous versus nonsynonymous mutation, the ratio of nonsynonymous to synonymous mutations should not ever exceed 1. Why? Because selection acts in a purely eliminative way and can at most remove some nonsynonymous mutations (by definition it cannot act on synonymous mutations). Therefore, the only thing selection can do is decrease the ratio of nonsynonymous/synonymous mutations.
Think about it: the synonymous mutation rate is in some sense a measurement of background mutation rate, and if the rate of nonsynonymous mutations goes up, we would expect the ratio of synonymous mutations to increase at least as much. This means the ratio will (assuming purely random mutations) never exceed 1.
Think about what this means if the ratio of nonsynonymous/synonymous mutations indeed exceeds 1. This means that somehow more nonsynonymous mutations than synonymous mutations have occurred. How can this be? Only if mutations are being targeted to nonsynonymous sites! Thus, this implies that the mutations are somehow constrained or guided. As Kirk put it, "Something is 'not bothering' with changes that don't make any difference." Exactly. Somehow, the synonymous mutations that we should expect are not happening! This is remarkable and seems like excellent evidence that mutations are actually directed somehow.
Of course, when you begin to think about things in detail they always turn out to be more subtle and complicated than you originally thought. That is the case here, as I've thought about whether there really is a 50/50 chance of nonsynonymous vs. synonymous. To determine this we have to go to a table of the genetic code and figure out what the average probability of a synonymous vs nonsynonymous mutation are. I found that nine amino acids are completely determined by their first two bases. Therefore, I gave them a 1/3 probability of having a synonymous mutation (the mutation must affect the last base in order to not change the amino acid). Six of the remaining amino acids are 50% determined by their first two codons (in other words, two bases in the 3rd position code for one amino acid and the other two bases code for a different amino acid). I gave those codons a 1/6 probability of synonymous mutation. Continuing in this way for all the codons and averaging over all codons, I came up with a probability of synonymous mutation of 0.2656, or about 1/4. This means that we should have about 1 synonymous mutation for every 4 nonsynonymous mutations. Thus, the "baseline" ratio of nonsynonymous to synonymous mutations should never exceed 4 to 1.
But as Kirk has pointed out, researchers have often found much greater ratios, up to 13 to 1, or in a study I recently looked up (dealing with mutations in salmonids), the ratio was 9 to 1. I am currently beginning to do some research on this and hope to have more data soon.
I would like some feedback on my methods for estimating the probability of synonymous versus nonsynonymous mutations, and whether these seem reasonable to people. Also, I would like to hear from others who have used the term "positive selection" and try to understand precisely what they mean by it--it seems to usually imply that selection somehow increases the ratio of nonsynonymous to synonymous mutations, but from what I tell that would be highly teleological and non-random, and certainly never predicted by Darwinian theory.
John Bracht
IP: Logged
|
|
Jack Foster
Member
Member # 79
|
posted 13. March 2002 02:48
Think about what this means if the ratio of nonsynonymous/synonymous mutations indeed exceeds 1. This means that somehow more nonsynonymous mutations than synonymous mutations have occurred.
Does the ratio compare mutations? Or does it compare nonsynonymous mutations that have been selected with synonymous mutations that have drifted?
IP: Logged
|
|
Drosera
Member
Member # 139
|
posted 13. March 2002 04:18
Uh...
...unless I'm sorely mistaken, there are two separate issues here that *should not* be confused, to wit:
#1 - The raw rate of mutation of any given gene in a given genome in a given population. This rate should be about the same for synonymous and non-synonymous substitutions (although there are of course complexities, certain kinds of mutations are more likely due to known chemical/physical causes, etc.) (and due to the way the code works etc., certain things are more common than others as Bracht points out above, just to clarify).
But #1 is *different* from:
#2, which is the rate at which such mutations ***are fixed in the population of genomes*** i.e. a population of individuals in a population-genetics sense.
Natural selection cannot increase #1 (leaving aside minor exceptions like hypermutator genes perhaps), but can certainly increase #2 for nonsynonymous mutations.
The study being discussed, and all such similar studies measuring the divergence of orthologues, are talking about #2. Here is the key piece that is missing from above analyses: natural selection can be either stabilizing or directional:
If natural selection is just maintaining the old function of a gene, then indeed changes selection can "see" (nonsynonymous) will be selected against and eliminated, whereas those that selection cannot see (synonymous) will get through, and very rarely a particular mutation will spread to fixation in a population by random chance (I say "very rarely", meaning *for a particular individual mutation* -- as the raw mutation rate is much much higher than the fixation rate, a few lucky neutral mutations will always get fixed, and these are the synonymous substitutions that are being measured).
If natural selection is instead favoring the improvement of a new function, for say a duplicated gene, then of course mutations improving that new function (which must be nonsynonymous by definition) will on average be selected, and on average spread through the population to fixation. Synonymous substitutions, though, will continue to accumulate at the same slow rate of fixation of neutral mutations by chance (although there are complexities here with 'hitchhiking' of genetic material closely linked to an adaptive mutation being spread by selection, I'm not qualified to comment further).
This latter process is what Johnson et al are talking about in (part) of their Nature article ( click here for link ) when they talk about 'positive selection'. E.g.:
(less than sign replaced with 'LT', greater than sign with 'GT' to avoid weird UBB error message)
quote:
Alignments of the human paralogous segments revealed that regions corresponding to coding exons were conspicuously hypervariable (10% nucleotide divergence when compared with intronic sequences that exhibited 2% divergence). The increased frequency of substitution suggested rapid genic evolution had occurred along with the genomic dispersal of the LCR16a duplication. Increased substitutions among exons are a hallmark of genes undergoing adaptive evolution8-11. A common test of positive selection is to compare the number of non-synonymous substitutions per site (Ka) to the number of synonymous substitutions per site (Ks)7. Ka/Ks quotients significantly greater than 1.0 are taken as evidence for positive selection. We assessed the Ka/Ks quotient for two of the most rapidly diverging exons (exons 2 and 4; Fig. 3b and Supplementary Information Fig. 3). Genomic subclones were obtained for the various copies of LCR16a from chimpanzees, gorillas, orangutans, gibbons and several Old World monkeys, and the exonic regions were comparatively sequenced (see Methods). Average Ka and Ks for all between-group and within-group comparisons was calculated independently for each exon (MEGA2, Modified Nei-Gojobori method; Tables 2 and 3). A statistical test of the difference of average Ka and Ks values both between species and for multiple copies within species was used as a measure of significance.
In the case of exon 2, highly significant Ka/Ks quotients (P LT 0.0005) were observed among all comparisons involving either humans or chimpanzees. The most extreme positive selection was observed between humans and Old World monkeys (Ka/Ks = 13.0, P LT 10-5) and between chimpanzees and Old World monkeys (Ka/Ks = 11.8, P LT 10-5). This level of amino-acid replacement translates into 43% amino-acid divergence between these species and a rate of amino-acid replacement of 1.0 10-8 changes per site per year for this exon. This is far in excess (20-fold) of most typical estimates7 of protein divergence between Old World and great ape species. Highly significant differences were also observed when comparing paralogues between chimpanzee and human sequences (Ka/Ks = 5.0, P LT 0.0001) with an average amino-acid divergence of 23% among the paralogous exons. To identify more precisely when the major episode of positive selection occurred, we estimated the number of synonymous and non-synonymous nucleotide substitutions per site for each branch of a phylogenetic tree using the method first proposed by Zhang et al.12. A major burst of positive selection seems to have occurred after the separation of the human and chimpanzee lineages from the orangutan (LT 12 Myr ago, Ka/Ks = 35.0), with subsequent protein-diversification events occurring during the emergence of chimpanzee and human species (see Fig. 3). A comparison with gorillas (Table 2) confirms that the major effect occurred in a common ancestor to humans and African apes. In stark contrast, the paralogues within orangutan and gibbon species have not experienced bursts of rapid positive selection (Table 2).
(etc., apologies for missing formatting)
The take home messages that *I* get from this paper are that:
1) Gene duplications are common in the recent evolutionary past of humans and African apes
2) The rapid modification of genes under the influence of natural selection is also commonplace.
This kind of thing is, we must realize, the kind of stuff that biologists read about and publish every day, and I think that this explains a great deal of the skepticism of biologists towards ID, which has yet to come to grips with the commonality of gene duplication and adaptive evolution of new genes in any substantial way as far as I can tell.
Click here if you want to see what comes up under 'related articles' for the Pubmed reference of the above article.
Drosera
PS: Basically we shouldn't confuse the mutation rate with the amino-acid replacement rate in a population of genomes.
PPS: Can't resist posting the summary of Johnson et al:
quote:
Our analysis has revealed an extraordinary degree of evolutionary plasticity, at the level of both the genome and the gene. We provide evidence for the evolution of a hominoid gene family by recent duplication and positive selection. Can additional examples within the human proteome be expected? Preliminary analysis of the human genome suggests that as much as 5–7% of all human sequences may have been duplicated within the last 30 Myr of evolution. The abundance of segmental duplications may be an important reservoir for the emergence of other hominoid genes that do not possess definitive orthologues in the genomes of model organisms.
[ 13 March 2002, 04:26: Message edited by: Drosera ]
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 13. March 2002 09:32
I think Drosera explains it well: mutation rates are one thing, substitution rates are another.
Mutation rate is more or less constant and uniform, regardless of effects on aa sequence (well, not entirely, since certain sequences are more mutable than others, but as a general approximation). Substitution rates however refer to the rates at which new mutations become fixed, and are found in all corresponding genes in the population.
So, mutations occur randomly and at a constant rate. Then, synonymous substitutions "drift", most of them are lost, but some fluctuate randomly to a point in which eventually they may become fixed; since the rate by which this occurs is essentially a stochastic process, it is quite constant in time (Ks).
On the other hand, non-synonymous substitutions are subject to selection: bad ones are eliminated, but good ones actually spread, sometimes very fast (a lot of non-darwinians have a problem understanding this "positive" aspect of selection, I think). Thus, Ka's can vary dramatically.
Another nice example of a recently diverged pair of genes displaying the effects of strong selection and a Ka/Ks ratio >>1 is here , for those with access to Nature Genetics (it was discussed a few days ago on the ARN boards). Here's the abstract:
quote:
Although the complete genome sequences of over 50 representative species have revealed the many duplicated genes in all three domains of life, the roles of gene duplication in organismal adaptation and biodiversity are poorly understood. In addition, the evolutionary forces behind the functional divergence of duplicated genes are often unknown, leading to disagreement on the relative importance of positive Darwinian selection versus relaxation of functional constraints in this process. The methodology of earlier studies relied largely on DNA sequence analysis but lacked functional assays of duplicated genes, frequently generating contentious results. Here we use both computational and experimental approaches to address these questions in a study of the pancreatic ribonuclease gene (RNASE1) and its duplicate gene (RNASE1B) in a leaf-eating colobine monkey, douc langur. We show that RNASE1B has evolved rapidly under positive selection for enhanced ribonucleolytic activity in an altered microenvironment, a response to increased demands for the enzyme for digesting bacterial RNA. At the same time, the ability to degrade double-stranded RNA, a non-digestive activity characteristic of primate RNASE1, has been lost in RNASE1B, indicating functional specialization and relaxation of purifying selection. Our findings demonstrate the contribution of gene duplication to organismal adaptation and show the power of combining sequence analysis and functional assays in delineating the molecular basis of adaptive evolution.
Zhang et al, Nature Genetics, Published online: 4 March 2002, DOI:10.1038/ng852
So, I guess Kirk's original point could be re-stated this way: if, for a recently duplicated gene under selective pressure, mutation rates for non-synonimous sites were found to be significantly higher than those of synomymous sites (or even better, mutation rates for sites whose changes have adaptive value, vs. non-adaptive sites), this would be strong evidence for some sort of teleological mechanism, perhaps design. Of course, one would have to control for all the caveats that became obvious during the studies on bacterial "adaptive" mutations.
I think this would be better tested in a metazoan system, but I am not sure how one would go along designing such an experiment, short of maybe some transgenic-type set-up.
[ 13 March 2002, 09:40: Message edited by: charlie d. ]
IP: Logged
|
|
John Bracht
Member
Member # 5
|
posted 13. March 2002 09:33
I just realized that I accidentally posted as "moderator" last night. I apologize for the mistake; several of us login as moderator from time to time to do administrative work and that's what I had done. I want to make it clear that I am NOT the moderator, and my posting last night under that name was a mistake.
John Bracht
IP: Logged
|
|
Kirk Durston
Member
Member # 174
|
posted 13. March 2002 13:56
It is entirely possible that I am missing something here, but thus far I have not seen an explanation, under natural selection, fixation, drift, etc. that predicts what Johnson et al. observed. Indeed, they state that the ratios were up to 20 times greater than had been estimated (I read 'predicted').
Several factors have been raised in this topic (mutation rate, fixation rate, gene duplication, speed at which a favorable mutation will spread throughout a population, drift, non-synonymous substitutions/site (Ka) and synonymous substitutions/site (Ks)). I think we can all grant that gene duplication occurs and that if a mutation increases the fitness of an organism, then it is likely to increase its frequency within a population and the rate of frequency change can be rapid if, among other things, the mutation offers a significant increase in fitness. Here are some things to consider in defense of my contention that Ka/Ks ratios significantly larger than one is predicted under ID but not under Darwinism.
1) Johnson et al. Indicate that if a paralogue has become non-functional, then adaptive and purifying selection constraints are neutralized and the Ka/Ks ratio should approximate unity.
2) Both non-synonymous and synonymous mutations for a given site can become fixed in a population, but what does the fixing is different between the two and the number of options for fixing is different between the two. First, if there is any natural selection at all, on the bases of fitness, not just any non-synonymous mutation will do. Natural selection will eliminate the lethal, or badly deleterious non-synonymous mutations right off the bat, and the remaining subset will form the options for fixation within the population. Let us call this subset Sa. On the other hand, any synonymous substitution for a given site will do as an option for fixation. Let us call this subset Ss. Assuming approximate equal fitness among the remaining members of Sa, we could expect some of the members of Sa to be lost due to drift. Of course, some of the members of Ss will be lost due to drift as well. Now here is the question:
Does Sa have fewer options than Ss?
If there is no selective pressure at all, then as Johnson et al. note, Ka should approximate Ks and, by inference, Sa should approximate Ss. But if there is any natural selection in operation at all, then as I have argued above, Sa should contain fewer members than Ss and, we would predict, Ka/Ks should be less than one. Indeed, this is what we most often observe so far as the empirical record is concerned.
3) I find large differences between paralogues within the same genome troubling. This seems to imply that each of the paralogues has a function within the genome; it isn't simply a case of mindless duplication. Of course, there are a lot of complicating factors here, as multiple paralogues could all be read, produce enzymes which all have identical structure and function, and used within a genome. Here is a thought: a given protein may be used in more than one way within a cell. It may be an essential component within a protein complex which, itself, has a different function that the protein has as a stand-alone. Or the same protein may be essential in more than one protein complex, with slightly different versions being more advantageous in their respective protein complexes (slight differences in the same overall 3-D structure).
4) I have spent a fair amount of time going over Johnson's paper and what I see them doing is publishing their findings and observations rather than providing anything approaching a rigorous explanation of their results. The high Ka/Ks ratios are labeled 'positive selection' but nowhere can I find an explanation of how it can occur other than references to an 'extraordinary degree of evolutionary plasticity', 'adaptive evolution', 'bursts of rapid positive selection' and a correlation between high ratios and gene duplication.
5) Johnson et al. note that coding exons under study had greater nucleotide divergence than the intronic regions and explain it as 'a hallmark of genes undergoing adaptive evolution.' This too, is troubling under natural selection, references notwithstanding.
As I said at the outset, I may be missing something here but I still do not see any explanation as to why high Ka/Ks ratios should be predicted under natural selection, duplication, drift, etc. If it was, it should be the norm given the proposed evolutionary history of life in all its diversity and disparity.
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 13. March 2002 15:18
Comparative rates of "positively" selected non-synonymous substitutions over the neutral background can simply be adjusted to <=1 by extrapolating rate changes over evolutionary time. Such extrapolations can sometimes seem "extreme" compared with rates measured over experimental time, but... you know. I'll admit I haven't read the cited article, so I'm wondering why the authors didn't do this? Maybe Kirk Durston is paying "too close" attention?
By pushing events into the remote past, extrapolating over sufficiently large intervals, adjusting rates appropriately, and accepting the likelihood of a few remarkable coincidences, the phenomenon can be saved. We can thank the Neo-Darwinists for the "larcenous" bookkeeping. What's really important in the end is not how the books are balanced, but that they do balance. Fits in well with the corporate culture, does it not? Sort of the "Enronization" of evolutionary biology.
Kirk Durston and Mr. Moderator, aka John Bracht, wonder about the concept of "positive selection." In the literature it is an a posteriori conditioning of the relevant random variables. We must never forget that natural selection is omnipresent, omniscient, and omnipotent. The "theologically" minded refer these attributes exclusively to some divivity, but in evolutionary biology we refer conventionally to "natural selection" and not "God," even if the two do closely resemble each other. Of course, natural selection lacks that certain "graceful providence" of a personal touch. But its all the same in the end... sort of.
Sorry, I'm not being very positive. Positively, we wouldn't have to engage in the "creative accounting" and could accept the observations as determinative if we consider as a live possibility that the programmatic constraints perform the relevant conditioning, and that the genome in some real sense actively "selects" optimizing transition probabilities, searches out a least path in its design space, minimizes randomization, adjusts rates/frequencies to convergence, and possibly even induces specific modifications, not as trial-and-error, but as trial-and-error-and-statistical induction.
Whoa! Is that crazy or what?! Especially that part of accepting as given the observations!
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 13. March 2002 16:27
Kirk:
I think you are almost there. Let me see if I can answer/clarify some of your points.
quote:
If there is no selective pressure at all, then as Johnson et al. note, Ka should approximate Ks and, by inference, Sa should approximate Ss. But if there is any natural selection in operation at all, then as I have argued above, Sa should contain fewer members than Ss and, we would predict, Ka/Ks should be less than one. Indeed, this is what we most often observe so far as the empirical record is concerned.
Right. This is generally true for all examples in which aminoacid substitutions are predominantly harmful, such as in the case of single gene-encoded proteins, especially those for which structural constraints are essential (e.g. histones). However, as soon as duplication occurs, one of the genes becomes significantly more able to diverge freely, because the “back-up” will exert the original function, while any advantageous mutation in the second will be picked up and selected. In fact, as you correctly state next…
quote:
This seems to imply that each of the paralogues has a function within the genome; it isn't simply a case of mindless duplication. …. Here is a thought: a given protein may be used in more than one way within a cell. It may be an essential component within a protein complex which, itself, has a different function that the protein has as a stand-alone. Or the same protein may be essential in more than one protein complex, with slightly different versions being more advantageous in their respective protein complexes (slight differences in the same overall 3-D structure).
…duplicated genes are free to diverge and “be drafted” into new functions, or “specialize” for existing alternative functions, which single genes usually can’t do. In the Nature Genetics RNase I case I referenced before, that’s exactly what happened. Leaf-eating monkeys were stuck with a sub-optimal RNase I for their diet, because the same enzyme could not mutate freely (it probably also plays a role in anti-viral immunity). However, after duplication, one of the paralogues accumulated aa mutations that progressively changed its optimal catalytic pH, making it a better digestive enzyme. The uniqueness of this study is precisely that the authors can follow the adaptive role of the mutations, showing how they favorably altered the enzyme’s function. Which brings me to your next point:
quote:
I have spent a fair amount of time going over Johnson's paper and what I see them doing is publishing their findings and observations rather than providing anything approaching a rigorous explanation of their results. The high Ka/Ks ratios are labeled 'positive selection' but nowhere can I find an explanation of how it can occur other than references to an 'extraordinary degree of evolutionary plasticity', 'adaptive evolution', 'bursts of rapid positive selection' and a correlation between high ratios and gene duplication.
Again, you are right. Unlike the RNase I example, unfortunately, in most cases of inferred divergence/adaptation we have at best a vague idea of what selective pressures shaped the evolution of the genes involved, and often we have no clue at all (such as in the morpheus genes of your example). It is therefore hard to make specific points about adaptive value. What the Johnson article wants to say is that this is a remarkably divergent gene family, with evidence of strong selective pressure, whose real function is entirely unknown and may have been important during hominoid evolution. I don’t think is meant as anything more than a descriptive paper, though.
quote:
Johnson et al. note that coding exons under study had greater nucleotide divergence than the intronic regions and explain it as 'a hallmark of genes undergoing adaptive evolution.' This too, is troubling under natural selection, references notwithstanding.
That is odd, although it seems to be occurring specifically for the exon 4 and flanking intron sequences. I do not see in the paper where they call this a hallmark of adaptation; in fact, they go to great lengths trying to explain the observation by non-adaptive mechanisms. Since it is due to a selective increase in the Ks in the exon, the authors mention some potential mechanism by which this may occur, including the possibility of alternative splicing and frameshifts for exon 4 (as already observed in humans), which obviously would affect the attribution of synonymous and non-synonymous substitutions compared to the reference cDNA. Whatever the mechanism, an increased rate of synonymous substitutions is indeed unusual, tricky to reconcile with any straightforward evolutionary hypothesis, not just according to a darwinian model, but also ID-based (as far as I can think of, although maybe you have some better idea).
quote:
As I said at the outset, I may be missing something here but I still do not see any explanation as to why high Ka/Ks ratios should be predicted under natural selection, duplication, drift, etc. If it was, it should be the norm given the proposed evolutionary history of life in all its diversity and disparity.
Actually, I think you got all the ideas already, as I pointed out before. You just have not put them all together. As you say, this is a phenomenon very characteristic of duplicated genes, which because of their nature are more free to undergo fast divergence than single genes, and it is due to the genes “specializing”, or “acquiring new functions” under high selective pressures. Not just “mindless duplication”, but active selection of “slightly different versions … more advantageous in their respective protein complexes”, or functions. The adaptive pathways are usually hard to track, and that’s why we are stuck with inferring adaptation “backwards” from the observation of high Ka/Ks ratios. However, high Ka/Ks ratios are entirely plausible, and in fact expected, in a neo-Darwinian context.
What do you think of the alternative proposal to your prediction I made in my previous post? Would the finding of high mutation rates targeted to potentially adaptive sites be a reasonable prediction based on ID?
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 13. March 2002 16:39
Janitor:
What exactly do you mean by "adjusting comparative rates of "positively" selected non-synonymous substitutions over the neutral background to <=1 by extrapolating rate changes over evolutionary time"?
We are talking about ratios of substitutions at side-by-side sites in the same gene in the same organism. You can't assume they have different evolutionary time-frames.
I don't see much space for 'creative accounting" here. The observations ARE determinative.
IP: Logged
|
|
|
Printer-friendly view of this topic