I've been amused by the complete lack of discussion, in all the usual places, of this experiment. To summarize, in 2004, geneticists found some fairly long stretches of DNA that are identical among humans, rats and mice. This means that these genes have not been subject to any viable copying errors or mutations over millions of years. Since as far as we know all genes are equally subject to change, that would seem to indicate that these genes do something rather important such that any change is selected out. Because genes that important must to something interesting, scientists excised them out of some lab mice to see what they would get. What they got was, so far as they could tell, perfectly normal mice. The ultraconserved genes don't seem to do anything at all.
To put it in a nutshell, if this is right than the theory of natural selection has to be wrong. One experiment is obviously not enough to bring us to that conclusion and yet we might want to get ready for the possibility that one of the pillars of modern thought is about to come crashing down.
56 comments:
David, you just made my day!
David, something very strange happened after I posted that comment - i had an email saying there was a PERMANENT FAILURE to deliver because dcohen does not exist. Where are you? Have you angered the Darwinists to the point where they needed to anihilate you?
Talk about your existential crises!
We excised him as an experiment to see if he does something important.
Leave him be, SH, until his keyboard has glean'd his teeming brain! (apologies to Keats)
I wonder if this would work for memes as well. For example, if you excised out of Harry, Duck and Skipper all the memes that make them so passionate about modernism, would they all still keep spouting the same nonsense about religion?
The ultraconserved genes don't seem to do anything at all.
Error detection & correction, anyone?
To put it in a nutshell, if this is right than the theory of natural selection has to be wrong.
To take it out of the nutshell, if this is right, it would have a great deal to do with genetics, and the understanding of random variation.
It might even prove the existence of a non-specific Designer.
But it would have absolutely nothing to do with natural selection.
I meant to note this before, but forgot.
Note that the ultra-conserved sequences are at most 731 base pairs long. The mouse and human genomes contain roughly 3,000,000,000 base pairs. It seems that there's a decent probability that you'd find unchanged sequences that long even in the presence of random mutations.
The best estimate I could find for mammalian / human mutation rates was 175 base pairs per generation. Figuring 20 years / generation, that's roughly 10 base pairs per year. Over 20 Myear we'd have 200 million base pair changes, ignoring duplicates and reversions. That's less than 10% of the genome. Simple clustering would lead you to expect unmodified sequences of less than 1K base pairs.
SH: The last common ancestor of mice, rats and humans lived 85 million years ago. These sequences are identical among all living mice, rats and humans. So, if it happened by chance, it would have to happen in all three species and all of their ancestors over generations that, for many of the species at issue, are measured in months.
The people who discovered these sequences -- before the redacting experiment was done -- put the chances of this being random at 1 in 10^22.
Skipper: If natural selection is compatible with keeping four different sections of the genome identical across 85 million years of divergence, then it is meaningless. No trait can then be said to have survived because it was more fit.
Remember, these genes were selected because they were ultraconserved. If natural selection most conserves trivial or junk genes, then it doesn't exist in any way useful to understanding evolution.
David:
Wrong.
What they got was, so far as they could tell, perfectly normal mice. The ultraconserved genes don't seem to do anything at all.
If they actually don't do anything -- that is, if their presence or absence are entirely irrelevant to the organism -- then there is nothing for natural selection to act upon.
What you really meant to say was that the random mutation part of naturalistic evolution -- if these gene sequences actually don't do anything -- is in for a real beating.
After the common ancestor, these sequences were not selected; they persisted when the odds say they should not have. If they are, in fact, completely irrelevant.
That is a big if. Given the state of play, to go from "don't seem" to "don't" is quite a stretch.
Picking a simple example: One could remove an entire disk drive from RAID storage. To all appearances, there would be no effect whatsoever.
That doesn't mean the disk drive did nothing at all.
Skipper: that's how RAID is designed to work.
Harry to me, 22 months ago:
And you seem to misunderstand the fate of mutations. They do not survive if not selected against. Over time, neutral mutations are elided out of the genome, because of the energy cost.
I also note that the people who do biology for a living find this to be a bothersome result.
Mr. Cohen;
From what little I could gather, creatures with shorter generations have fewer mutations per generation, so the base pair mutation rates are roughly equal (the precise measures have large error bars, so it's hard to say anything more definite).
I also don't believe that 10^-22 odds, either. I would bet money that it's the odds of those very specific sequences being conserved. But the sequences were selected post facto, so the correct odds are the odds of any such sequence being preserved. It'd be like arguing you don't exist because the odds of your specific DNA sequence arising are infinitesimal, and is a common dodge of anti-Darwinists.
Could you provide a cite for the 85 MYear value? I thought mammals didn't arise until around 65 MYear ago.
so the correct odds are the odds of any such sequence being preserved.
SH, I'm about to jump into water way over my head. Wouldn't that apply only if we assume that the sequence that was excised was the one and only identical sequence among the three billion base pairs?
(Fast duck)
No. Consider the original genome sequence at the time of speciation. Afterwards, various mutations occur and differences arise, with ever smaller preserved sequences. Only after all of that history do we look and pick out whatever sequences have been preserved. The only situation in which we can't do that is if no sequence of sufficient length was conserved anywhere in the genome. Therefore, the odds of finally a randomly preserved sequence is the odds of not ( no sequence preserved ).
Think of a lottery with a billion tickets. The odds of any person winning is 10^-9, but you're guaranteed of finding a billion to one shot winner afterwards. The essence is that you look at the winner after he's won. This kind of post facto choice is a common trope of the anti-Darwinists, focusing on the single, extremely unlikely winner without regard to the untold multitude of other possibilities that would have been equivalent but didn't occur.
Presumably the biologists involved have done some of these calculations (which are a bit involved) and therefore have reason to think that preservation by chance is unlikely. Or perhaps it isn't and the scientists are simply checking whether it happened or not. But 10^-22? I laugh at that estimate.
SH: You can get more information here and in the linked articles. But you seem to be assuming that this is some sort of anti-Darwin pseudo-science. In fact, the people doing these experiments are main stream geneticists. It's possible that they, and all the reviewers, and all the scientists puzzled by this have simply calculated the odds wrong and there's no real issue here, but I'm skeptical that that's the explanation.
Unfortunately, I can't link directly to the 2004 article that computes the probability, but if you have access to an academic database, here's the cite: Ultraconserved Elements in the Human Genome, Gill Bejerano, Michael Pheasant, Igor Makunin, Stuart Stephen, W. James Kent, John S. Mattick, and David Haussler (28 May 2004), Science 304 (5675), 1321. [DOI: 10.1126/science.1098119]
Presumably, we can agree that, whatever the faults of Science, anti-Darwinism is not one of them.
And here's the explanation:
Calculation of p-value for finding any instance of 200 bases absolutely conserved between human, mouse and rat in the human genome: This calculation is done using a Poisson approximation. Each column in the orthologous multiple alignment is considered to be an independent observation of a Bernoulli random variable that is 1 (“heads”) if the bases are completely conserved between the three species (a “3-way identity”) and 0 (“tails”) otherwise. Based on analysis of neutrally evolving (ancestral repeat) sites in each 1 Mb window in the human genome (1, 35), we estimated the mean of this Bernoulli variable (the probability of heads) to be at most 0.7. (The largest percent identity among ancestral repeat sites we obtained for any 1 Mb window with enough ancestral repeat sites to get a good estimate, i.e. at least 1000 sites, was actually 0.68.) The distribution of the number of runs of at least 200 heads in a series of 2.9 billion tosses of a biased coin with probability p = 0.7 of heads can be approximated quite well using a Poisson distribution with mean (1-p)·p200 (43), and the probability of one or more such runs is very close to the mean of the Poisson distribution in this case, which is at most 10-22. This probability is small even if the neutral probability of 3-way identity is as high as 0.9.
Wow, I knew I ducked for a reason.
No, I meant that those promoting this as "Darwin Disproved" might have a bias or willful misinterpretation / misreading of the report.
I'll have to think about that explanation, but it seems like a reasonable approach and the numbers work out. I'll concede the point.
Joe:
Skipper: that's how RAID is designed to work.
Precisely my point.
If someone whose knowledge of RAID went only so far as chemical warfare on spiders when the man of the house is not around (my wife, who will happily admit she would have to be in a coma to care any less than she already does about computer arcana), the removal of a disk drive from drive array without any apparent ill effect would lead to the conclusion that drive did not seem to do anything at all.
Just so here. Just because removing the gene sequences in question produced no apparent ill effects is not, in and of itself, evidence that those genes do not do anything at all.
After all, in a far, far simpler system than a genome, it is possible to remove something without readily apparent ill effect whatsoever. Absent, that is, underlying knowledge of what that drive is doing.
Now, if it can actually be demonstrated that those sequences are even less important to humans, rats, and mice than a kickstand is to a tricycle, then naturalistic evolution is in for some serious revising (although Darwinism, strictly defined, could not possibly be affected, since Darwin did not -- could not -- even know genes existed).
David:
In citing Mr. Eagar, you have clearly failed to read your own writing.
This means that these genes have not been subject to any viable copying errors or mutations over millions of years.
This means one of two things:
1. Just what it says: no mutations whatsoever among a non-trivial number of absolutely useless gene sequences since the last shared ancestor. I agree with SH, it seems the odds are very strongly against this happening by chance.
2. There were mutations, all of them causing a significant decrease in fitness, but for reasons as yet unknown.
If the first, then there simply isn't variation for natural selection to act upon, AND our notions of random mutation -- but not natural selection -- are in for a beating
Otherwise, both random mutation and natural selection stand reconfirmed.
Given other strongly conserved gene sequences (hox being probably the most prominent) we do know some things about), I will buy you dinner at a location of your choosing if it in fact turns out these sequences are completely useless.
You are right, this is a very interesting result.
Just not for the reason you said.
I usually don't jump in on the Darwin threads, but setting the details (and the ratios) aside, for just a moment, why should it be a surprise that a theory first postulated 150 years ago suddenly hits the wall? If not this discovery, surely there will something else that will alter the "model" of natural selection.
Physics has had its moments of intellectual 'catastrophe' in the past. Why shouldn't biology?
jim:
It stopped being about biology long ago.
Peter:
If this post and thread isn't about biology, what is it about?
Let me add a few amateur observations to the fray. Did the scientists confirm that the sequence is the same for all mammalian species that were descendants of the common ancestor of men, mice and rats? Wouldn't that be the expected result if this sequence were overconserved? If it varies between species but is only the same in men, mice and rats, and perhaps some other species, then I'd say its a coincidence.
I'll second Skipper's caveat that apparent and actual functioning are not the same. You could take the fuel filter off of a car without noticing anything adverse for a long while. But eventually the engine will gum up and die.
The study acknowledges that they have not proven that no functional attribute of the mice was deleted or altered. But the theory seems to be that utraconserved genomes should be those that code for the most immediately critical biological functions, as opposed to functions that are necessary from a gradual, inter-generational timeframe. That's an assumption. It doesn't make it true.
Another assumption is that this sequence experienced no mutations over the time period that separated mice and men from their common ancestor. All we know is that this sequence is the same now. The sequence could have mutated independently between mice and men and arrived at the same sequence. If it were the same in all mammalian species, then this possibility is remote in the extreme. But if all we've proven so far that mice and men share this sequence together, without knowing about other species, then it's not as outlandish.
One other possibility is that this sequence codes for a function that is not very complex, but is necessary, and for which there are really no ways to improve by mutation. It could be a maximally optimized gene.Even the most sphisticated modern machines will have simple components that may not have changed in 100 years, like a simple machine screw. If you compared an A380 to a DC3, would it be surprising if one of the parts that they had in common was a 15 cent Cotter pin? Thinking teleologically for a moment, designers achieve improved results by varying the major design components, like engines, drivetrains, etc. They don't usually acheive results by re-inventing the smallest design elements, like screws or wires, although these too are part of the mix. So it shouldn't seem surprising to find the most overconserved elements to be minor design elements.
Skipper,
It's about theology, of course.
If you compared an A380 to a DC3, would it be surprising if one of the parts that they had in common was a 15 cent Cotter pin?
As I was saying, it stopped being about biology long ago.
Yeah, I have to agree with Mr. Burnet on that. The problem is that evolutionary theory takes random mutation as an axiom, making it as likely to modify a 15¢ screw as the engine air intake. In, over time, more likely, because it is precisely the less significant sequences that have the least pressure for conservation.
I still think that there's something fundamentally wrong with their probability analysis, but I need to think about it some more before I would be willing to be more specific. I'd probably want to write a few simulations as well.
Duck:
No, not theology. More metaphysics. I can understand the impatience and frustration at those who deny evolution in favour of a childlike creationsim (although that is an almost uniquely American battle born in American politics). However, I continue to be amazed by the dogged defence of natural selection in the face of all kinds of improbabilities and weaknesses. Why is it so hard for you folks to say something like "I believe in naturalism and materialism, but we still have a lot of work to do to figure out what drives evolution."?
Peter, photocopy this comment, date it and store it in your archives. Ready? I believe in naturalism and materialism, but we still have a lot of work to do to figure out what drives evolution.
To me the most proven point about evolution is common descent. I think that natural selection is most likely, because it seems intuitive and it is borne out on the macro level, but I admit that there are still gaps to be filled. I'm really not dogmatic about what is used to fill those gaps, other than to prefer not to know over believing in something just to close all the gaps. I'm fine with provisional assumptions because I'm willing to alter those assumptions when they are proven wrong.
That's something that the theologians can't do.
Gee, Duck, I hope you aren't the sign of a trend. I would really miss my beloved just-so stories. Oh well, I guess I'd still have Skipper to amuse me with his theories on the evolution of the Corvette Stingray and iambic pentameter.
I would prefer to say the most proven point about evolution is common ascent (at least until the "decoupling", as Harry calls it), seeing as it goes from the simple to the complex, but otherwise you can copy this comment for your archives too. What do we argue about now? Oh, I know, of course theologians can alter assumptions when they are proven wrong. They do it all the time. I know, you guys say they can't, but thankfully they don't pay any mind.
The problem with theologians being proved wrong is that they are purportedly inspired by divine revelation to get it right. The whole enterprise is built upon revelation. If you can't get revelation right, it sorta destroys all credibility for the enterprise.
And the stakes being what they are, salvation or damnation, their little mistakes cost souls. I don't know of any souls that are at stake in the debate over natural selection.
And don't worry! I imagine we will continue to find subjects to disagree over. ;-)
I don't know of any souls that are at stake in the debate over natural selection.
Yes, you do.
Elaborate, Joe.
Is it too much to suppose we shall never hear again that evolution is not a science because it can't be disproved?
Much too much. Evolution isn't a science. It is natural history.
But natural selection is a scientific theory. A flimsy one.
SH - They estimate a 30% change that any one base pair is mutated. Then, crudely, in a single 200-bp segment the chance of it being ulta-conserved is (.7)^200 = 10^(-31). There are 3 billion such segments in the genome, to take your number, so, ignoring overlap, no more than 3 billion (3 x 10^9) 200-bp segments; so the chance of an ultra-conserved 200-bp sequence in the genome is on the order of 3 x 10^(-22).
So their number of 10^(-22) is plausible, presumably their calculation was more realistic than the above.
Clearly, the assumption that base-pair mutations are random events independent of the mutation status of their neighbors is wrong. So far, the only mechanism accepted for non-independence is selection, and there seems to be no phenotypic event for selection to act on. So there must be some unknown non-selection effect that makes the probability of a mutation on one bp of this sequence correlated with the probability of a mutation in its neighbor.
That's the state of affairs. What's missing is a theory of the mechanism. It could be a host of things.
I don't take it as a refutation of the existence of either "random" mutations or selection, but as a proof that some other effects also exist.
paulj:
What's missing is a theory of the mechanism. It could be a host of things.
Can you suggest some possibilities? In previous discussions about how Darwinism is ultimately tautological, we've tried to think of theoretical alternatives without much success. If the whole game is about survival by definition, how can the mechanism be anything other than a drive to survive without bringing the whole ediface down?
PaulJ;
As far as I can tell from Mr. Cohen's quote and the numbers cited, their analysis was just what you wrote. What my experience of randomness tells me, though, is that you don't get such smooth distributions, you get a lot of clumping, which could easily yield sequences that this analysis would determine are extremely improbable. I would have to run some simulations to verify that, however.
The other thing that bothers me is that their universe of outcomes varies along the number of mutations axis, i.e. every possible number of mutations is an outcome. I would think that a better model would restrict the universe to outcomes with a relatively constant number of mutations. I.e., presume 30% mutations and then modify that many pairs.
First, SH: You get clumping in any process. The question is, what is the distribution of clumps, and is it consistent with a random process?
The model they used, a Poisson process, basically says: we going to run 3 billion events in succession, each with a 0.7 chance of non-mutation and 0.3 chance of mutation. Poisson statistics can calculate the frequency of "runs" of different length -- e.g. how many runs of 1 non-mutation, how many runs of 2 non-mutated base pairs, etc. It gives a probability of having even one run of 200 non-mutated base pairs of 10^(-22). So it seems Poisson statistics -- indicative of a random chance of mutation of 30% for each base pair, and independence of neighboring base pairs -- is not the right model. The mutations are too "clumped" for a Poisson process.
So, clearly, either: (1) the probability of mutation of certain base pairs is much less than 30%, or (2) the probability of mutation of one base pair, at least in certain regions of the genome, is not independent of the probability of mutation of neighboring base pairs, or both.
Peter - I didn't mean that there has to be a complete replacement/alternative to natural selection, only that some process must exist besides selection that makes the probability of mutation in a base pair correlated with the probability of mutation in its neighboring base pairs. Selection does this because it acts on genes (or transcriptional control regions of genes): the base pairs in a gene produce a specific protein whose function depends on every nucleotide, and if selection is conserving the protein, then it's conserving every base pair in the gene as a totality, i.e. blocking changes in any nucleotide.
But it's not selection that's enforcing the shared low mutation rate in this case.
As for other mechanisms, I am not a molecular biologist (though my wife is, and I stayed at a Holiday Inn Express), so I hesitate to make suggestions. I would guess that epigenetic transcriptional control mechanisms, which control the phenotype of cells (e.g. whether your cells, which all share the same genome, turn into heart, brain, muscle, or blood cells; or whether another genome forms a caterpillar or butterfly) are important. Epigenetic states are generally preserved through cell replication and can be inherited. If the mutation rate in this region is suppressed when genes are epigenetically silenced, and if these genes are silenced (explaining their lack of influence on the mice), it's possible they could have been silenced throughout the mammalian lineage going back tens of millions of years. In other words they might be "junk" genes that are silenced for reasons of energy-conservation.
No, I think the modeling problem is more subtle than that.
The root is that in physical reality, the number of modified base pairs isn't variable, as it is in the Poisson model. We may not know precisely what it is, but it's effectively a fixed percent. If it wasn't, all of those "genetic clocks" would be hogwash. So I think the more accurate model is "we have N base pair mutations, randomly distributed". If we take that as basis, then we don't have independence of base pair mutations which is a significant difference.
My intuition is that such a model would be much clumpier than the Poisson model, i.e. have a much higher probability of long unmodified chains, but it's possible I am wrong about that as well. As I noted, I would need to run some simulations to get an order of magnitude sense of how much clumpier, which, sadly, I don't have time for right now.
SH - There is a "genetic clock" becuase in any random process, if it goes on long enough, the number of mutations becomes strongly peaked around N mutations. For instance, in a sum over normally distributed events the width of the distribution scales as 1/sqrt(N), N the number of events; or, when rolling dice, the probability distribution is flat for one die, peaked at 7 for two dice, more strongly peaked at 10-11 for three dice, and so forth. The longer the time period, the more precise the evolutionary clock becomes (as long as the random process doesn't change its character -- not necessarily a good assumption).
Of course, the Poisson distribution they assumed also comes up with a distribution that is highly peaked around 30% of the genome being mutated. So, of 3 billion base pairs, 900 million +- a relatively small number will end up mutated.
Your alternative model is "we have N base pair mutations, randomly distributed." But "randomly distributed" usually means the next event is independent from previous events. If I understand you, your process will start with 3 billion base pair spots, run a random variable at each one, and assign either a mutation or non-mutation. At the first base pair the probability of mutation is N/(3 billion); if after y base pairs you've had x mutations, you want to make the probability of a mutation there as (N-x)/(3 billion - y), so that at the last base pair the probability of mutation is either 0 or 1 depending on whether you've previously used N or N-1 mutations.
Well, your model isn't biologically realistic, but in any case it gives a very similar probability of not getting a 200 bp run as the model they used.
I might add that in the model I've ascribed to you, the probability of getting a long run of non-mutations actually is less than in the Poisson process, because with every non-mutation added to a run of non-mutations the pool of available non-mutations is reduced and the probability of a non-mutation on the next try decreases. That doesn't happen in the Poisson process.
No, the process would pick a base pair at random roughly 900,000,000 times and mutate it.
paulj:
All of this is very dependent upon the suspicion that the gene sequences do not, in fact, have any function at all.
It is worth noting that conclusion is based upon deleting them in their entirety, not (IIRC) making changes in the way random mutations would.
My bet is that there is a reason -- other than highly fortuitous dice rolls -- these sequences are so highly conserved between disparate mammalian species.
Either they, in fact, do something; or, piecemeal mutations are very bad in ways that wholesale deletion is not.
Both leave Darwin undisturbed.
Of course, an intelligent designer could presumably manage wholesale deletions (which is, after all, the IC argument). But, It has not chosen to do so.
SH - Are you making the point that some base pairs will have mutated multiple times? Yes, but that's not going to affect the outcome much. We're still constrained by the observed mutation rates to end up with about 70% of the nucleotides unmutated.
Your more complex model only matters if you have a random process that is not identical for all base pairs and independent between base pairs. If you allow an independent random process to act on each base pair individually, then the outcome is adequately modeled by a Poisson process. And the point of the paper was that a Poisson process doesn't work, thus refuting the assumption of independent random events drawn from a uniform distribution.
hey skipper - You lost me about three times in there. Darwin hasn't been disproved, but we can't really know whether he is disturbed, assuming God has told him about this finding.
I don't think we can conclude much yet except that there are as-yet unknown processes in genetics waiting to be discovered. Whether those processes have a bearing on phenotypic evolution is unknown.
For the science-interested, here's an article indicating the lack of knowledge and uncertainties regarding the epigenetic role in genetics: http://www.sciencedaily.com/releases/2007/11/071130160450.htm. A quote:
“It’s like a puzzle where you only have a few pieces,” Pinto said. “There’s not a clear cut signal and consequence. We will learn new insights into the cellular mechanisms that govern the basic process of chromosome transmission from mother to daughter cell.”
No. My experience has been that selection models and independent coin flip models tend to have different amounts of clustering, which is a critical parameter.
Skipper: That really strikes me as whistling past the graveyard. Any genetic mechanism that results in the most conserved genes being junk or only weakly related to fitness cannot be squared with any strong form of natural selection.
It does fit nicely with my understanding of natural selection, which is that, unless a mutation kills outright the phenotype will survive.
We really don't know enough about how the genome functions to make any conclusions.
Cells tend to program in messiness and carry a lot of redundancy in them. They're not perfectly-engineered machines.
paulj:
I don't think we can conclude much yet except that there are as-yet unknown processes in genetics ...
Which is precisely the point I have tried to make several times. Their are two assertions required to make this extremely improbable result astonishing:
1. The conserved gene sequences have absolutely no function; and,
2. Piecemeal changes in those sequences will not have effects sufficiently deleterious to cause their disappearance from succeeding generations, even though the intact sequences themselves have no function.
There is very little evidence for the former assertion, and none for the latter.
If both could be demonstrated, then, and only then, will the random mutation portion of naturalistic evolution come into question, and in such a way as to quite possibly leave continuous supernatural intervention at the molecular level the best explanation.
David:
It does fit nicely with my understanding of natural selection, which is that, unless a mutation kills outright the phenotype will survive.
You need to work on your understanding. See, in particular, chronospecies.
How do you deal with the fact that Darwin himself stated that IF the single cell is found to be anything but a simple lump of protoplasm, that he would consider his own theory disproven, and microbiology has indeed proven that the single cell is a complex machine run by trillions of proteins interacting in a designed, purposeful manner. And this doesn't even address the advances in the other cutting edge science such as quantum physics and cosmology which also have found strong indications of design. And what about the Cambrian explosion? NO fossils have EVER been found indicating that several proto-species (before specific animals evolved into subspecies better suited for various ecosystems such as prototype fox became desert fox, arctic fox, etc.) did not just appeared from thin air. If there haven't been fossils found by now predating the Cambrian explosion, there never will be. There isn't and never will be real scientific evidence that all life evolved from a single blob of protoplasm. Like the Soviet Union and Communist China, politics in the US has censored science to suit its own ends.
Post a Comment