Home » Creationism, Darwinism, Evolution, Intelligent Design, Video » Nachman’s Paradox Defeats Darwinism and Dawkins’ Weasel

Nachman’s Paradox Defeats Darwinism and Dawkins’ Weasel

The following is a crude 1-minute silent animation that I and members of the IDCS Network put together. God willing, there will be major improvements to the animation (including audio), but this is a start. Be sure to watch it in full screen mode to see the details.

http://www.youtube.com/watch?v=SrIDjvpx7w4

The animation asserts that if harmful mutation rates are high enough, then there exists no form or mechanism of selection which can arrest genetic deterioration. Even if the harmful mutations do not reach population fixation, they can still damage the collective genome.

The animation starts off with healthy gingerbread men as parents. Each spawns ginger kids, and the red dots on the kids represent them having a mutation. The missing ginger limbs are suggestive of severe mutations, the more mild mutations are represented by ginger kids merely having a red dot and not severe phenotypic effects of their mutation. The exploding ginger kids represent Selection doing its thing and removing the less functionally fit from the population. The persistence of red dots on the ginger kids represents persistence of bad mutations despite any possible mechanism of selection.

Nobel Prize winner HJ Muller (of Muller’s ratchet fame) suggested that the human race can’t even cope with a harmful rate of 0.1 per new born. The actual rate has been speculated to be on the order of 100-300.

The animation uses a conservative harmful rate of 1 and argues (with some attempts at humor) that deterioration would thus be inevitable even with a harmful rate of 1 per new born.

I save discussion in the comment section the relevant but technical topics of truncation selection, sexual reproduction, recombination, synergistic epistasis, compensatory mutations, relief from Muller’s ratchet etc. These highly technical topics should be addressed and were not included in the animation. We can discuss them in the comment section.

However, the essential problem of mutation rates and deterioration is depicted by the animation. How this cartoon is illustrative of reality (when we consider the technicalities such as recombination, sexual reproduction, synergistic epistasis), can be discussed in the comment section.

In light of such problems Kondrashov posed the rhetorical question, Why have we not died 100 times over?. Kondrashov attempted to answer the question, but I don’t think the problem has been solved. The animation expresses my skepticism of the long term benefit of “synergistic epistasis”.

And if the conclusion symbolized by the animation is true, then on what grounds can we believe Darwinism is true?

The animation was inspired by a paper by Nachman and Crowell. Two years ago I wrote: Nachman’s U-Paradox. This animation helps illustrate the problem of Nachman’s paradox.

I recommend we should build a non-partitioned WEASEL to feature how Nachman’s paradox will trump Dawkins conclusions that somehow Darwin found the answer to appeance of design. One can get an idea of what it would look like given the animation.

[ Admins, I can't seem to embed the video, can you embed it for me? I think embedded video requires higher privilege than my account has.]

Notes:

1. We could have done the drawings differently to emphasize the mutations are unique and novel and different for each ginger kid, but I save that work for later (including audio).

2. There is a refinement to the animation that is in order based on Nachman’s calculation of average removal rates of harmful mutations assumng trucation selection, “U”=3, and a conservative reproduction rate for humans, but I didn’t get around to it. That is yet another modification for future animations. We’ll need also some technical research on the matter.

  • Delicious
  • Facebook
  • Reddit
  • StumbleUpon
  • Twitter
  • RSS Feed

108 Responses to Nachman’s Paradox Defeats Darwinism and Dawkins’ Weasel

  1. Notes:

    1. To be fair, I believe Nachman is a convinced evoulutionist and in way do I mean to suggest he is a creationist or ID proponent. But the paradox he put forward could (unwittingly) defeat the Darwinian evolution (often symbolized by Dawkins WEASEL).

    2. There are other schools of evolutionary thought that are non-Darwinian such as the mutationists and the neturalists. ID proponents and creationists like Walter ReMine and John Sanford have relied heavily on the work of the neturalists (most notably Kimura). They also drew on the works of Kondrashov, Nachman, and Crowell. The animation was named after Nachman who was the co-author of the paper that inspired this
    animation. I wrote about Nachman and Crowell’s work in Nachman’s U-Paradox.

    Any of the following 3 areas of investigation could separately overturn the Dawkinsian view that life was created by a Blindwatchmaker:

    1. Origin of Life (see the works of Thaxton, Bradely, Olsen, Kenyon, S. Meyer, Yockey,

    Voie, Trevors, Abel, Don Johnson many others)

    2. Population Biology ( Walter ReMine, John Sanford)

    3. Irreducible, Specified, Integrated, Functional Complexity, No Free Lunch
    (Shutzenberger, Eden, Behe, Dembski, Marks, Denton, others)

    I bolded the Population Biology because to date it is an underempahsized and misunderstood field, partly owning to difficult technical issues. This animation was intended to help make clear the essential issues of population biology and how it may challenge the hypothesis of Darwinian evolution.

    A little known area of population biology research funded by the Discovery Institute was pursued by Walter ReMine. ReMine’s work was inspirational to a later book by John Sanford, Genetic Entropy.

    I speculate that problems for Darwinism in population has only begun to be explored.

    It has sympathizers even from secular quarters (the neturalist and mutationist schools of evolution).

    See: Part of the Discovery Institute’s secret
    research program uncovered

    4. I wish to thank Allen MacNeill of Cornell and Joe Felsenstein of WSU for their comments on my writings the last two years. They helped correct and improve my understanding immensely, and though they are on the opposing side of the debate, they have helped increase my appreciation for the hard work and research being done in the field. Thanks also to Dave Wisker for alerting me to Lynch and Estes papers and “Zachriel” of telic thoughts.

    Special thanks to Walter ReMine and John Sanford. My hope is Walter’s contributions to the field of population biology will finally be recognized and appreciated by the rest of the world. He’s paid the price for challenging the mainstream.

  2. I speculate there will be keen interest in this topic not becaue of the creation/evolution debate but because genetic deterioration has significance to medical understanding.

    The medical community is ramping up its ability to track sincle nucleotide polymorphisms.

    I mentioned at UD in 2006:
    Solexa: a development which may lead to measuring claims of ID

    Some of the claims by ID proponents have not been adequately explored because of the cost issues involved in doing large-scale whole-genome sequencing of numerous individuals. Not even Warren Buffet has the trillions of dollars needed to accomplish such a massive amount of gene sequencing. At least not today, but maybe in the future!

    The human genome project took 3 billion dollars and 13 years to complete. By comparison, Solexa might be able to do a comparable job for a few thousand dollars per person (ideally even less) and in a much shorter time frame. (See the UD sidebar on Solexa Genomics.) Solexa might be viewed as an unwitting research partner of the ID movement.

    The fine work of two important ID proponents, Cornell geneticist John Sanford and independent researcher Walter ReMine, might finally get slam dunk empirical confirmation if Solexa succeeds in its grand quest. For example, a fundamental consequence of Sanford’s Genetic Entropy thesis is that there will be an unabated rise in Single Nucleotide Polymorphisms (SNPs) per generation per individual. If confirmed, this data will be more nails in Darwin’s coffin, and then Darwin Day might have to be renamed Darwin Bashing Day (or something else, how about Abe Lincoln Day?).

    Solexa, Inc. is developing and commercializing the Solexa Genome Analysis System, which is being used to perform a range of analyses including whole genome resequencing, gene expression analysis and small RNA analysis. Solexa expects its first-generation instrument, the 1G Genome Analyzer, to generate over a billion bases of DNA sequence per run and to enable human genome resequencing below $100,000 per sample, making it the first platform to reach this important milestone. Solexa’s longer-term goal is to reduce the cost of human re-sequencing to a few thousand dollars for use in a wide range of applications from basic research through clinical diagnostics. For further information, please visit http://www.solexa.com.

    Update: I should add, Solexa’s technology is poised to provide data which will overturn the prevailing ideas about molecular clocks (See: Molecular Clocks: Michael Denton continues to be vindicated). I will elaborate in the comment section if anyone is interested. In brief, we do not have accurate measurements of molecular evolution to the degree needed to overturn certain hypotheses. The increased accuracy provided by Solexa technology could permanently shatter the prevailing molecular clock hypothesis and vindicate various claims of ID proponents.

    Update: Solexa’s technology will also aid in the ID quest of steganography in biology. If the key to rapidly understanding the steganography in junkDNA is through comparative sequencing of various creatures, Solexa’s technology is a welcome friend. I would like to also acknowledge again the fine work of Dr. Pellionisz on JunkDNA. I expect he will be delighted by the work of Solexa. Here is the latest from Dr. Pellionisz comment 87430

    Solexa and Illumina were used in a recent relevatn article on human mutation:

    Human Y Chromosome Base-Substitution Mutation Rate Measured by Direct Sequencing in a Deep-Rooting Pedigree.

    If Rick Sternberg’s thesis is correct that 90-100% of the human genome is functional, then we have to wonder what this mutation rate (as determined through solexa technology) means for genetic deterioration. It would (by my calculation) mean that we could be experiencing 100 harmful mutations per new born.

    That would be difficult to estimate because what constitutes harmful.

    I would suggest however, with respect to the question of design, we could compare the progress of Single Nucleotide Polymorphisms in comparing deeply “conserved” sequences.

    I postulate the results will eradicate any presumption that the “conservation” is due to selection whatsover. There will simply not be enough population resources to sustain the needed purification (as suggested by the animation).

    The benefit of this is that we bypass any need to define “fitness” or functionality. We can measure purifying selection (or lack thereof) without needing to identify these other issues.

    Furthermore, we can extend this line of inquiry to the architectural designs uncovered by the ENCODE project, but I save that discussion for later.

  3. Hi Scordova interesting topic, I have a bit of a off topic question though, Have you heard any word on the progress of Dr. Sanford’s work down in the “salt mines”?

  4. Well I’m waiting to hear from the geneticists whether my heart defect (ARVC+CHF) is genetic i.e. caused by a mutant gene, either passed to me or a new one of my own as it were. So could I be one of the proof’s that could overturn darwin? If so then praise the lord & pass the ammunition!

  5. It’s great to see some modeling being done here, and congrats with the nice animation. However, it would be nice to have some more details, so us readers can check your results.

    Therefore I have a couple of questions:

    (1) You say there is one harmful mutation per offspring. Is that one on average from a poisson distribution, or do you give every single offspring exactly one harmful mutation?

    (2) What is the viability of mutant offspring? (s=0.99, s=0.9, what?)

    I have many more questions, but those will do for the moment.

  6. jistak,

    This following study confirmed the “detrimental” mutation rate for humans, of 100 to 300, estimated by John Sanford in his book “Genetic Entropy” in 2005:

    Human mutation rate revealed: August 2009
    Every time human DNA is passed from one generation to the next it accumulates 100–200 new mutations, according to a DNA-sequencing analysis of the Y chromosome. (Of note: this number is derived after “compensatory mutations”)
    http://www.nature.com/news/200.....9.864.html

    Professional evolutionary biologists are hard-pressed to cite even one clear-cut example of evolution through a beneficial mutation to the DNA of humans which would violate the principle of genetic entropy. Although a materialist may try to claim the lactase persistence mutation as a lonely example of a “truly” beneficial mutation in humans, lactase persistence is actually a loss of a instruction in the genome to turn the lactase enzyme off, so the mutation clearly does not violate Genetic Entropy. Yet at the same time, the evidence for the detrimental nature of mutations in humans is overwhelming for doctors have already cited over 3500 mutational disorders (Dr. Gary Parker).

    “Mutations” by Dr. Gary Parker
    Excerpt: human beings are now subject to over 3500 mutational disorders.
    http://www.answersingenesis.or.....ations.asp

    As well the slow accumulation of “slightly detrimental mutations” in humans, which are far below the power of natural selection to remove from our genomes, is revealed by this following fact:

    “When first cousins marry, their children have a reduction of life expectancy of nearly 10 years. Why is this? It is because inbreeding exposes the genetic mistakes within the genome (slightly detrimental recessive mutations) that have not yet had time to “come to the surface”. Inbreeding is like a sneak preview, or foreshadowing, of where we are going to be genetically as a whole as a species in the future. The reduced life expectancy of inbred children reflects the overall aging of the genome that has accumulated thus far, and reveals the hidden reservoir of genetic damage that have been accumulating in our genomes.” – Sanford; Genetic Entropy; page 147

  7. It is overwhelming likely that most mutations on earth that have occurred have been reproductively neutral. For instance, humans average about 130 new mutations per sexual generation. And I’m doing fine.

  8. Thanks, born. I read the recent Current Biology paper about the mutation rate at a stretch of the human Y chromosome. But what I am interested in here are the details of Salvador’s model.

  9. It’s great to see some modeling being done here, and congrats with the nice animation. However, it would be nice to have some more details, so us readers can check your results.

    Therefore I have a couple of questions:

    (1) You say there is one harmful mutation per offspring. Is that one on average from a poisson distribution, or do you give every single offspring exactly one harmful mutation?

    1 is an assumed minimum.

    If the average number of new harmfuls per human is 100, several standard deviations from the mean would easily be a minimum of 1 per human if the standard deviation is say 5, and even if one were lucky enough to avoid getting a novel harmful mutation, on balance I think it would be of too little help.

    I do not know off hand what the standard deviation is or whether anyone on the planet has a figure. I presume the figures from the Y-chromosme study would result in a 2-sigma mutation rate of at least 100 per individual, at least that is how I read their paper. Anyone with a better guess or a correction to that figure is more than welcome to post their estimate here are UD.

    (2) What is the viability of mutant offspring? (s=0.99, s=0.9, what?)

    The viability is not specified (not to mention even in field practice would be almost impossible to directly determine short of a high death rate) and the animation attempts to show that the viability is a moot point!

    If the viability is 1.00 (meaning the mutation is harmful but does not reduce immediate reproductive fitness), then that’s bad, the mutation persists. If the viability is 0, well the individual dies. But the last image of the animation shows, that with the presumption of 1 new harmful per new born, it doesn’t matter who is killed off.

    A note about viability being 1.0 and the mutation being harmful. I have taken issue with characterizing something as “fit” based on reproductive benefit.

    See:
    Survival of the sickest, why we need disease.

    Also, it is a bit suspect to characterize S values that are contingent. You can see the problem in characterizing S based on immediate fitness.

    See the quotation of Lawrence Hurst here: Airplane Magnetos, Contingency Designs, and Reasons ID will Prevail.

    If the S-value is contingent on other mutations, this makes traditional analysis difficult if not impossible. We can sometimes put some sort of estimate if we have recessives and a degree of inbreeding, but for other contingent situations, it is hard.

    However from an analytical standpoint, if we are not tracking the behavior of specific mutations but rather the net increase in harmfuls, one sees that it doesn’t matter if S=1.00 or S=0. If S=1.00 for all every bad mutation, then the bad persists (a comparable example would be blind cave fish that have functionally bad mutation, but reproductively fit mutation that caused blindness). If all S values were zero, there is extinction.

    Thus, this shows there does not exist any conceivable value of viability that can arrest deterioration.

    Importantly, the general hypothesis or related hypotheses can be tested empirically with the advent of Solexa and Illumina technology and large bio-informatic systems. That is if we’re willing to shell out the money and resources for such an enterprise.

  10. Incidentally, contingent S values are esentially the behavior of synergistic epistasis.

    That is, mutation A in and of itself has zero effect on reproductive fitness, mutation B in and of itself has zero effect on reproductive fitness, but the presence both simultaneously results in death.

    I tried to depict what would happen even if that were true with the last slide and selection based on synergistic epistasis.

    The issue is not one of seleciton coefficients, but rather the cost of purification.

    Walter has put forward the cost model of selection with a paper funded by the Discovery institute.

    I think the cost model would be very good for topics like purifying selection as it can be easily visualized (as evidenced by the animation).

  11. I hinted about a future modification to the animation.

    It is possible for Dad to have a bad mutation and Mom to have a bad mutation and because of basic laws of inheritance, some of the kids may not have that mutation.

    Nachman discusses this issue in his paper if Mom and Dad have 3 mutations:

    Estimate of Mutation rate per nucleotide

    For U = 3, the average fitness is reduced to 0.05, or put differently, each female would need to produce 40 offspring for 2 to survive and maintain the population at constant size. This assumes that all mortality is due to selection and so the actual number of offspring required to maintain a constant population size is probably higher.

    So there is a removal rate on the generous assumption of truncation selection and every human females giving birth to 40 kids (OUCH!).

    The removal rate of harmfuls cannot be greater than 1 per human on average (recall the assumption 1 always being added), so any inefficiency (due to less than perfect truncation and/or less than need number of offspring) will result in net increase of harmfuls.

    I should mention, I suggested a mutation rate of 1 novel mutation per new born will lead to deterioration. Muller estimates even 0.1 would be intolerable. I don’t think I’m stating anything that should be controversial.

  12. Salvador, you haven’t answered my questions, which were very specific. Let me rephrase them, in case I was unclear. I am talking about the specifics of the model on which your animation is based.

    In the model, what is the distribution of harmful mutations among the offspring?

    In the model, exactly how harmful are individual mutations?

    Please answer those questions. After all, you make some strong claims in the OP, so it seems only fair that you explain exactly how you arrived at those conclusions.

  13. In the model, what is the distribution of harmful mutations among the offspring?

    Apparently you missed that I said 1 was minimum and was presumed several standard deviation from the mean of 100. Invoking the phrase “standard deviation” conventially means a NORMAL distribution.

    Before accusing me of not answering the question, perhaps you should show a little more astuteness. I don’t take kindly to such flimsy nit picking.

    It is you who didn’t read and comprehend that I answered your question.

    Please answer those questions. After all, you make some strong claims in the OP, so it seems only fair that you explain exactly how you arrived at those conclusions.

    Please be more astute in recognizing that your question was answered. I have little patience for trolling.

  14. 1 is an assumed minimum.

    If the average number of new harmfuls per human is 100, several standard deviations from the mean would easily be a minimum of 1 per human if the standard deviation is say 5, and even if one were lucky enough to avoid getting a novel harmful mutation, on balance I think it would be of too little help.

    Ok jistak, let me spoon feed you.

    If the distribution is normal and the mean is 100 and standard deviation is 5, 1 would be several standard deviations from the mean in a normal distribution. Comprende?

  15. Salvador,

    Apparently you missed that I said 1 was minimum and was presumed several standard deviation from the mean of 100. Invoking the phrase “standard deviation” conventially means a NORMAL distribution.

    No reason to get upset. Just asking some questions here. The technical problem is, the domain of a normal distribution ranges from minus infinity to plus infinity, so it cannot have a minimum of 1. Moreover, number of mutations is a discrete variable, while a normal distribution applies to continuous variables. Maybe that’s nit picking, but if I wanted to reconstruct your model, those details matter.

    Ok jistak, let me spoon feed you.

    If the distribution is normal and the mean is 100 and standard deviation is 5, 1 would be several standard deviations from the mean in a normal distribution. Comprende?

    Yo comprendo muy bien amigo, gracias. OK, so the mean number of mutations per offspring was 100 from a (truncated) normal distribution with SD=5. Thanks, that answers one of my questions.

    Next, in the model, how did the mutations affect viability and/or fecundity of the offspring?

  16. No reason to get upset. Just asking some questions here. The technical problem is, the domain of a normal distribution ranges from minus infinity to plus infinity, so it cannot have a minimum of 1

    No you mis-interpret again.

    Assuming a minimum of 1 is several standard deviations from the mean which covers 99.9999….% of the cases.

    You can thus have a minimum of 1 and that would be a minimum within 99.999..% of the cases in a Normal distribution. So the assumption by any standard is reasonable given the premises. I also pointed out, the outlier would be so improbable to be of no consequence.

    Next, in the model, how did the mutations affect viability and/or fecundity of the offspring?

    I pointed out they were not modeled because they were moot points. If you wish to provide a model that models moot points and numbers that deal with moot points you can provide them and post them here.

    The animation depicts the simplifying assumption that the removal rate is zero, but a more accurate depiction would have a removal rate between zero and one, and I pointed out a future animation will account for non-zero removal rates. I alluded to the calculation by Nachman.

    For this animation, the assumption was whatever fecundity or viability will result in zero removal rates. This was a simplifying depiction for the first go around.

    For future animations, we can have a removal rate between zero and one assuming any conceivable viability and fecundity that can support that number.

    No value of viability or fecundity will affect the basic point of deterioration if the removal rate of harmfuls on average is less than 1.

    Fecundity and viability would affect the removal rate number, but by assumption, it cannot be greater than one given the premise put forward.

    A removal rate greater than one is not possible (given the premises), and a removable rate of 1 is not realistic. Thus removal rates would have to be less than 1. But a removal rate of less than 1 guarnatee deterioration.

    The way to resolve the question of course is not by theoretical modeling but to track things like the actual increase in single nucleotide polymorphisms.

    If Sternberg’s hypothesis that 90-99% of the genome is functional, then an unabated rise in the number of polymorphisms would imply an unabated rise in the number of harmfuls, which imply the removal rate is closer to zero than to one, which would underscore the claim issues with fecundity and viability are moot points.

    You are welcome to post what you think the removal rate should be based on any reasonable value of viability or fecundity that you desire, so long as the constraint of 1 new harmful per new born is modeled.

    With respect to fecundity, Nachman mentioned human females having 40 kids each. I think that is a bit much!

    The bottom line is based on the premis of 1 new mutation per new born, deterioration will happen. The issues of fecundity and viability will only affect the speed and severity.

    I leave the question of speed and severity for others to investigate.

  17. In comment #13, jitsak asked a crucial question. Specifically, it is of absolutely fundamental importance to know two things:

    1) Are the organisms that we are considering diploid or haploid? Humans, like almost all animals, are diploid, but bacteria and all other prokaryotes (and many, but not all protists/protoctists) are haploid.

    2) Are the mutations for which we are calculating coefficients of fitness dominant, recessive, or neutral (and if dominant, what is the degree of penetrance)?

    Without an answer to these questions, one literally cannot calculate anything having to do with the fitness of specific alleles. Here’s why:

    Both Sanford’s “genetic deterioration” model and the video presented in this thread assume that both beneficial and deleterious mutations will be “visible” to selection, and can therefore be subject to selection. However, in diploid organisms both beneficial and deleterious alleles can be “invisible” to selection if they are either recessive or if they are dominant, but lack complete penetrance.

    The opposite would the case for any genes located in the genomes of haploid organisms (bacteria, archaea, most protists, many fungi most of the time, the gametophyte generations of plants and multicellular algae, and even some animals, such as whiptail lizards). This would also be the case for alleles located in the y chromosomes of mammals and other non-paired non-homologous chromosomes.

    In all of these latter cases, dominance does not exist and all genes are immediately “visible” to selection. As far as I can tell, it is only genes like these that are subject to the effects modeled by Sanford and others, and which could (at least theoretically) lead to some kind of “genetic deterioration”.

    Furthermore, as Motoo Kimura (and Jukes and Crow independently) pointed out, neutral mutations (i.e. mutations that are neither beneficial nor deleterious) are not “visible” to selection and can therefore persist in populations indefinitely.

    Even severely deleterious mutations will not be visible at all unless they are dominant (or at least exhibit some dominant “penetrance”). If they are recessive (and/or show no penetrance), then they will not be eliminated from populations except when they appear in homozygous recessives.

    For example, consider the empirical fact that the allele for cystic fibrosis (call it “f”) is extraordinarily deleterious and recessive, whereas the F allele is both normal and dominant with virtually 100% penetrance. The f allele almost always results in death when homozygous (usually in childhood), but has no immediately obvious phenotypic effect when heterozygous. I know this for a fact, as my wife is heterozygous for the cystic fibrosis allele (i.e. she is Ff, whereas I am FF, yet she shows absolutely no phenotypic signs of having any of the symptoms of cystic fibrosis).

    We know (as the result of molecular genetic analysis) that the f allele has the effect of disabling the chloride protein ion channels in animal cell membranes. People who are homozygous FF and Ff can make all the normal (i.e. functional) chloride channels they need. Only ff individuals cannot do this, and hence cannot regulate the transport of chloride ions across cell membranes. Despite the fact that homozygous ff individuals virtually always die, usually in childhood, the frequency of the f allele among people whose ancestors are from Europe (i.e. Caucasians) is approximately one in twenty (or about 5%).

    Given that the ff homozygous condition is literally as deleterious as it can be, how quickly should the cystic fibrosis allele be removed from the human population, and is there some equilibrium value at which the frequency of the cystic fibrosis allele is so low that it only ever appears in heterozygotes (i.e. its frequency is so low that the probability of two heterozygotes interbreeding and having 1/4 of their offspring expressing cystic fibrosis is effectively zero)?

    The answer always surprises my students: let’s take just Europeans. Even though the allele is present in 5% of the population, the probability that two people who carry the allele will mate and produce children is only 1/20 X 1/20 = 1/400. Furthermore, the probability that two heterozygotes will have a child that has cycstic fibrosis is only 1/4 (i.e. 1/4 of their children will be homozygous normal and 1/2 = 2/4 will be heterozygous carriers). Ergo, the probability that two Europeans (chosen at random) will have a child that has cystic fibrosis is 1/20 X 1/20 X 1/4 = 1/1,600. This is why the majority of cases of cystic fibrosis in the United States are the very first appearance of this condition among either of the families of the parents of the affected child.

    But that’s the theoretical calculation. What is the actual frequency of cystic fibrosis? In 1997, about 1 in 3,300 Caucasian children in the United States was born with cystic fibrosis. In contrast, only 1 in 15,000 African American children suffered from cystic fibrosis, and in Asian Americans the rate was even lower at 1 in 32,000.

    Why is the actual frequency of cystic fibrosis so much lower than the predicted frequency? One reason is that the actual frequency of the cystic fibrosis allele varies from subpopulation to subpopulation. Indeed, if one does a calculation based on the assumption that the f allele is a complete recessive, but completely lethal allele, and the F allele is both completely dominant and has no deleterious effects, then the frequency of the f allele should be less than 1/10,000, and therefore the actual rate of appearance of cystic fibrosis should be 1/400,000,000 (i.e. 1/10,000 X 1/10,000 X 1/4). This calculation is based on the assumption that every single individual who is ff dies before reproducing, whereas every single individual who is FF and Ff dies of something else (i.e. unrelated to the presence or absence of the F allele).

    So the interesting question is not “why is the frequency of the f allele not zero”, but rather “why is the frequency of the f allele so anomalously high among Europeans (but not African Americans or Asians, who still exhibit a frequency of the f allele that is slightly higher than that calculated on first principles alone).

    The answer seems to be that, like the allele for sickle-cell anemia, the allele for cystic fibrosis increases the fitness of heterozygotes, relative to homozygous normal FF. Using standard population genetics calculations, it is possible to calculate the degree of this increased fitness and then factor it into the forgoing calculations.

    Physiological and developmental research is now being conducted to determine precisely how the anomalously high frequency of the f allele is being maintained in some (but not all) human populations. Four hypotheses have been advanced for this, all based on the assumption that being Ff confers some resistance to at least one of the following diseases:

    • Cholera: With the discovery that cholera toxin requires normal host chloride transport proteins to function properly, it was hypothesized that carriers of the mutant f allele benefited from resistance to cholera and other causes of diarrhea. Further studies have not confirmed this hypothesis.

    • Typhoid: Normal chloride transport proteins are also essential for the entry of Salmonella typhi into cells, suggesting that carriers of mutant chloride transport genes might be resistant to typhoid fever. No in vivo study has yet confirmed this. In both cases, the low level of cystic fibrosis outside of Europe, in places where both cholera and typhoid fever are endemic, is not immediately explicable.

    • Diarrhea: It has also been hypothesized that the prevalence of cystic fibrosis in Europe might be connected with the development of cattle domestication. In this hypothesis, carriers of a single mutant chloride transport chromosome had some protection from diarrhea caused by lactose intolerance, prior to the appearance of the mutations that caused lactose tolerance.

    • Tuberculosis: Poolman and Galvani from Yale University have added another possible explanation – that carriers of the f allele have some resistance to TB.

    My money is on the diarrhea hypothesis, as there is good evidence that the f allele first appeared among Europeans about 52,000 years ago (based on the neutral mutation rate of “silent” base-pairs in the f allele).

    In the context of this thread, cystic fibrosis is as lethal as a mutation gets, yet its frequency has neither caused humans (including Europeans) to go extinct, nor has it been removed from the collective human genome. How might this eventually occur? The probability of the f allele persisting in the human genome is an equilibrium between the rate of its removal (via both expression among homozygous ff individuals and pure, random accidental disappearance as the result of heterozygotes failing to pass on the allele for reasons unrelated to its phenotypic effect) and the rate of its preservation as the result of the increase fitness of heterozygotes. It is virtually impossible for the f allele to cause humans to go extinct nor for the allele to completely disappear as long as the effective breeding population of humans remains relatively high (this also calculable using standard population genetics, but I won’t go into it here). However, if the effective breeding population declines enough, then inbreeding effects begin to cause the f allele to appear much more often as ff homozygotes. However, this has the effect of removing the f allele from the population, and if the population is small enough, it can completely disappear (this is the so-called “Sewall Wright effect”, which is now usually referred to as “genetic drift”).

    Ergo, if effective breeding populations fluctuate in size enough that they get so small that deleterious alleles can disappear from them by accident as the result of genetic drift, then the Muller/Kondrashov/Sanford problem of “genetic load” (i.e. the accumulation of deleterious alleles, which appear only in heterozygotes) goes away.

  18. Salvador,

    Assuming a minimum of 1 is several standard deviations from the mean which covers 99.9999….% of the cases.

    You can thus have a minimum of 1 and that would be a minimum within 99.999..% of the cases in a Normal distribution. So the assumption by any standard is reasonable given the premises. I also pointed out, the outlier would be so improbable to be of no consequence.

    Fair enough, if the mean is sufficiently high (in this case 100 mutations per offspring), a normal distribution is a good approximation to a poisson distribution. What strikes me as odd, though, is that you would choose 1 as the minimum number of mutations, rather than the more obvious choice of zero.

    I asked:

    Next, in the model, how did the mutations affect viability and/or fecundity of the offspring?

    and you answered

    I pointed out they were not modeled because they were moot points. If you wish to provide a model that models moot points and numbers that deal with moot points you can provide them and post them here.

    They are not moot points at all. The harmfulness of mutations determines how efficiently selection can remove them from the population.

    The bottom line is based on the premis of 1 new mutation per new born, deterioration will happen. The issues of fecundity and viability will only affect the speed and severity.

    I am not so sure about that. On any given locus, recombination creates offspring without mutations on that locus if both parents carry a single mutation on that locus. Then the spread of the mutation depends on the selective advantage of that offspring compared to offspring that do inherit the mutation.

    Look, Dawkins may have lost the source code of his original Weasel model, but I am sure you were not that careless and you can show us the source code of your model, so that we can check for ourselves if your conclusions are justified.

  19. It’s great to see some modeling being done here, and congrats with the nice animation.

    Unfortunately, it’s not a model, it’s an animation. There is no underlying model for the animation.

  20. Hi Scordova, sorry to interupt, but I think you may have missed my question at #4
    Have you heard any word on the progress of Dr. Sanford’s work down in the “salt mines” i.e. ancient bacteria?

  21. Hi Scordova, sorry to interupt, but I think you may have missed my question at #4
    Have you heard any word on the progress of Dr. Sanford’s work down in the “salt mines” i.e. ancient bacteria?

    Profuse apologies. I have not asked him about this work. Nothing to report yet.

  22. Jistak objects:

    They are not moot points at all. The harmfulness of mutations determines how efficiently selection can remove them from the population.

    And I pointed out short of 100% removal efficiency (totally unrealistic given there are numerous genetic diseases that have persisted for centuries), the deterioration will happen. The point is moot with the final outcome.

    Or do you claim selection will can purify harmfuls at a rate greater than 100% on average. :-)

    PS
    see next few comments on what I mean by 100% efficiency. there is a subtlety there.

  23. Fair enough, if the mean is sufficiently high (in this case 100 mutations per offspring), a normal distribution is a good approximation to a poisson distribution. What strikes me as odd, though, is that you would choose 1 as the minimum number of mutations, rather than the more obvious choice of zero.

    No. If 100 is the mean. 85 would be giving the benefit of the doubt Darwinist case, 70 would be generous, 50 would be insanely generous, etc.

    The way to frame the question is why I didn’t pick a higher number than one. The more obvious choice is not zero, but some number higher than one.

  24. jistak:

    I am not so sure about that. On any given locus, recombination creates offspring without mutations on that locus if both parents carry a single mutation on that locus. Then the spread of the mutation depends on the selective advantage of that offspring compared to offspring that do inherit the mutation.

    I alluded to that with this comment:

    It is possible for Dad to have a bad mutation and Mom to have a bad mutation and because of basic laws of inheritance, some of the kids may not have that mutation.

    But I pointed out, with more mutations this becomes problematic and with 3, human females would be having to give birth to 40 offspring AND truncation selection would have to be applied.

    In any case, the purifying selection will not work with 100% efficiency (100% relative to the rate number of new mutations being added, not 100% with respect to preventing fixation or even elimination from the population).

    As I suggested, we can even assume, for the sake of argument that any specific harmful mutation will eventually be removed given enough time.

    The issues is whether it can remove it at the same rate as they are being introduced (that is what 100% efficiency means in this context).

    This is analogous to a waiting line that gets longer and longer over time. The backlog just keeps increasing even though in principal we could speculate any given individual, given enough time gets serviced.

  25. Sternberg’s hypothesis is presented here:

    How The Junk DNA Hypothesis Has Changed Since 1980

    , it seems that 90% or more of chromosomal DNA has some kind of specific developmental function, given the available data.

    If the human mutation rate is 100 to 200 (as bornagain77 quoted), then Sternberg’s hypothesis would suggest that about 100 new harmfuls is a reasonable figure if changes to a nucleotide would result in functional compromise most of the time.

  26. Sal, do you think the harmful mutation rate has increased since Adam and Eve, or remained constant?

    For example, did each of their children have, on average, 100 harmful mutations?

    And they passed those on to their children, who also had 100 new harmful mutations, etc., down to the time of Noah and his family, where many of these likely became fixed in the population.

    And then the process again continued.

  27. You do realize that the human gene complement is 2X, and that we reproduce sexually?

    Maybe you could try the video where not all the gingerbread people have the deleterious mutation to begin with. It would then be a really nice example of natural selection.

    The video essentially shows that a deleterious, penetrent, dominant disorder will be strongly selected against. Which is predicted by evolution, and seen in human disease. Most dominant disorders aren’t fully penetrent, or affect the individual after the age of reproduction.

    In contrast, non-penetrence, recessive genes, etc. explain how a deleterious trait could persist.

  28. I’m pretty sure the mutation rate must have been less before the flood and then picked up, judging by the rather constant life expectancy of about 900 years before it, and the several generations in which it steadily dwindled afterward.

  29. bornagain77 at #5 mentions the detrimental results of first cousins marrying. In many Muslim societies girls are often married to their uncles via arranged marriages. I wonder if, likewise, there are any inherited harmful consequences because of this?

  30. Since I was not familiar with the abbrevation “IDCS” I searched “Thelic Thoughts” for it. There it
    has been used with different meanings. I wonder which one Salvador refers to and if he thinks ID theory will benefit from any of these definitions especially in the light of the DOver case:
    1. Intelligent Design Creation Science
    2. Intelligent Design Creationists
    3. Intelligent Design Critics

  31. Sal, do you think the harmful mutation rate has increased since Adam and Eve, or remained constant?

    We can’t establish scientifically that there was indeed an Adam and Eve. I personally think their was, but it is a personal opinion, thus anything I say on the matter is properly a personal opinion, and not a direct scientific inference.

    That said, I think the mutation rate is higher by some unknown factor based on the supposition we live in an environment that will likely cause more mutations (more chemicals, UV light, harsher environment).

    The rate is determined somewhat by quantum effects which cause copy errors at the molecular level.

    For example, did each of their children have, on average, 100 harmful mutations?

    And they passed those on to their children, who also had 100 new harmful mutations, etc., down to the time of Noah and his family, where many of these likely became fixed in the population.

    And then the process again continued.

    Yes the harmfuls would accumulate with some removal. Severe diseases would obviously be removed, but lesser ones can persist, and there have been genetic maladies that have indeed persisted. Many of the maladies are mild (like myopia).

    Many of the harmful are hidden as recessive traits of discovered in the event of inbreeding.

    With respect to the question of Noah (if indeed he was real, and that is not an established fact), we could in principal investigate the Table of Nations in Genesis chapter 10 and 11. This actually describes the structure of a detectable phylogeny, imho. This would be especially feasible to verify because Y-chromosomes are passed only from father to son.

    Along similar lines, there is published in secular journals research on the Cohen Haplotype, which scientists suspect describes the house of Levi.

    We could even, more easily take say Arabs and Jews. Their Y-chromosomes ought to align with the phylogeny suggested of Abraham, Ishmael, and Isaac.

    The Y-chromosome phylogeny should also be broken into 3 discrete lineages for Noah’s 3 sons. In turn, the table of nations should describe the number of branches on the next level of hierarchy, so on and so forth.

    This research may become feasible as Solexa and Illumina become more in use and bio-informatic systems become cheaper.

    The genetic catalogs should continue to rise mainly because of the importance to modern medicine.

    Furthermore, Solexa technology gives us reasonable mutation rates. If these mutation rates were about the same to the time of Noah, we could, in principle, not only construct a phylogeny, but even estimate maybe when the patriarchs were alive.

    A phylogeny that aligns with the Table of Nations and the genealogies (such as suggested by th one in Luke chapter 3) would lend credence to the idea that the tedious listing of names was indeed a real family tree and not some fabrication.

    But I stress again, these are bordering on religious topics and creationism, not really Intelligent Design. For sure, these topics are of interest to me, but I stress there should be a demarcation between speculation and more tangible inferences.

    The subject of genetic deterioration is testable empirically.

    The existence of Noah, Adam, Eve is a matter of speculation and circumstantial evidence.

    PS
    See:

    Y-chromosomal Aaron and Y-chromosomal Levi regarding the Cohen Haplotype.

  32. From Wiki:

    In this case illustrated above with table and graphic, Joseph Felsenstein`s methods for making coalescense statistically independent comparisons using 21 familiar markers haplotypes shows a well shaped, defined, and geographically distributed Cohanim Tree.

    Despite two thousand years since the destruction of the Second Temple, and the spread of the Jewish population into the Diaspora, the deadly Crusades, Cohanim families managed to survive the persecutions and kept their lineages intact, imprinted in their Y Chromosomes as a unique and common signature. This signature, distinctly reflecting the Cohanim ancestral haplotype, visibly identifies today and recognizes these 21 Jewish priest families, directly related to one common Cohanim ancestor who lived 2400 +-300 years ago, around the times of Zadok, the High Priest that anointed King David. This is because both Cohanim Jews, Askenazi and Sephardi, for thousands of years preserved their genealogical lineages since the Temple period. Of course they did not think in terms of the DNA, however, they have faithfully followed Cohanim Jewish tradition until present day. As a result this is exactly what the haplotype tree shows. Families which haplotypes sit on the tree next to each other on flat branches live (or lived) in close territories and more likely share a recent common ancestor, as the tree shows. The Cohanim Tree places the correspondent families in the branches based on respective mutations. DNA results confirmed, by positioning the families in their respective places of origin, that the geographical location is correctly connected in genetics according to the Jewish tradition and records found in each one of these 21 different Cohanim families.

    Which suggest to me, there has been a strong cultural tradition to maintain accurate genealogical records for a long time.

    The research on the Cohen haplotype was able to resolve a common ancestor to within 2400 years ago.

    What will we find for the common ancestor of all men as our mutation rate info gets better resolution. :-)

    All this data should also tell us if the neutralist crticism of selectionism is correct, namely, natural selection doesn’t do much to purify out new mutations and the majority of molecular evolution is not subject to selection.

    If the majority of molecular evolution is not subject to selection, then it is presumptuous to argue Darwinism was the principal mechanism that created the features of man.

  33. Sal, do you think the harmful mutation rate has increased since Adam and Eve, or remained constant?

    Now that is priceless.

  34. Salvador,

    And I pointed out short of 100% removal efficiency (totally unrealistic given there are numerous genetic diseases that have persisted for centuries), the deterioration will happen. The point is moot with the final outcome.

    Yes, there are numerous recessive diseases that are shielded while rare. Classic population genetics predicts a low equilibrium frequency.

    As long as you refuse to provide the model details that I asked for, I tend to agree with Mung that there is no underlying model to the animations. The animations are what you want to be the truth.

    You can prove us wrong by showing your source code.

  35. Hi Scordova,
    I think that a more realistic model would have included e.coli bacteries or other simple organisms of this kind. According to Plotkin, the fitness landscape of bacterias and even human tend to stabilize after a period. Please see: “Speed Limit To The Pace Of Evolution “.

  36. Kyrilluk, states “the fitness landscape of bacterias and even human tend to stabilize after a period.”

    Not according to this evidence:

    Upon closer inspection, it seems Lenski’s “cuddled” E. coli are actually headed for “genetic meltdown” instead of evolving into something better.

    New Work by Richard Lenski:
    Excerpt: Interestingly, in this paper they report that the E. coli strain became a “mutator.” That means it lost at least some of its ability to repair its DNA, so mutations are accumulating now at a rate about seventy times faster than normal.
    http://www.evolutionnews.org/2.....enski.html

    Since “selection pressure” is removed from humans as with the e-coli, the actual evidence suggests we are headind for meltdown sooner.

    Something of interest for graham1:

    Another interesting line of genetic evidence, which has recently come to light and which is extremely antagonistic to the materialist, is the Genetic Adam and Genetic Eve evidence, which is exactly what they call it in the scientific literature. This genetic evidence offers another line of strong support for the Biblical view of the sudden creation of man.

    Human Evolution – Genetic Adam And Eve – Hugh Ross – video
    http://www.youtube.com/watch?v=1cfHsFtw02g

    CHROMOSOME STUDY STUNS EVOLUTIONISTS
    Excerpt: To their great surprise, Dorit and his associates found no nucleotide differences at all in the non-recombinant part of the Y chromosomes of the 38 men. This non-variation suggests no evolution has occurred in male ancestry.
    http://www.reasons.org/interpr.....lutionists

  37. Hugh Ross is an old-earth creationist, and has been scolded by Creation Ministries International. They have some serious theological problems with Hughs position, but it was all a bit deep for me.

  38. Graham1, Well there are some serious problems with the physics of Young Earth Creationists. Their molecular biology is fairly strong as far as that goes, but there is some serious discrepancies, in the contorting they do to physics, in order to have the evidence fit their model. Which as far as I am concerned puts them in the same boat as evolutionists for letting their ideology drive their science!

  39. Graham,

    Here is a debate, on John Ankerberg, between Dr. Ross and Kent Hovind, who was probably the best Young Earth debater, prior to his arrest for tax evasion.
    Judge for yourself whether who is more consistent in their science and their theology.

    Kent Hovind vs Hugh Ross (Part 1, disc 1 of 2)
    http://www.youtube.com/watch?v=WNuHuG517lI

  40. As long as you refuse to provide the model details that I asked for, I tend to agree with Mung that there is no underlying model to the animations. The animations are what you want to be the truth.

    You can prove us wrong by showing your source code.

    The model is not run on a computer or requires source code.

    The animation makes vizual a simple inference.

    If there is one new harmful mutation per new born human, deterioration will not be arrested. There exists no values of viability or fecundity or any concievable method of selection or drift that will arrest the long term deterioration.

    Do you disagree with that?

    As long as you refuse to provide the model details that I asked for, I tend to agree with Mung that there is no underlying model to the animations. The animations are what you want to be the truth.

    I’ve provided the details.

    Model: one new harmful mutation per new born human. All other details will generally be moot with respect to the final conclusion.

    The power of the description is its simplicity. I wanted an argument that won’t be difficult to conceptualize or comprehend.

    You are now accusing me of not providing the details. Did I ever say the model of deterioration came from source code? No! Does it reflect reality? If anything it may understate the severity of reality. If the aniation is wrong, it may be wrong to the extent it is not showing enough deterioration. I put only one dot in the animation per new ginger kid, perhaps 50 or 100 or even 200 might have been more appropriate!

    And then you make a silly complaint that I put only one mutation when the minimum would have been zero. That’s sophistry, jistak.

    You can prove us wrong by showing your source code.

    I told you there is no source code. The model is conceptual, and the animation makes vizual this conception.

    If there is one new harmful mutation per new born human, deterioration will not be arrested.

    Do you disagree with this simple obvious fact?

    I look forward to hearing whatever obfuscation, distortion, and twisted Darwinistic sophistry you will attempt to put forward to refute a very simple and logical inference.

    So far all you’ve put forward is sophistry. That suggestion of using zero as a minimum number of mutations. Now that was some of the best twisted sophistry I’ve seen on the net yet. Do you recommend using zero as the number mutations in all models of mutation that use a normal distribution? LOL…..

  41. The power of the description is its simplicity.

    It seems oversimplified to the point of having little demonstrative value. As near as I can tell from your description, the presence of even one harmful mutation is enough to terminate the offspring. The assumption that any harmful mutation is fatal is unwarranted. Some mutations are trivially harmful and won’t affect survivability. Second, you make no provision for beneficial mutations.

    From this perspective, even Dawkins little program, WEASEL, is a better model as it allows both harmful and beneficial mutations and doesn’t necessarily weed all beneficial mutations out of the gene pool (as demonstrated in the video of the program running, which shows some reversions.)

  42. In comment #37 hummus man wrote:

    “It seems oversimplified to the point of having little demonstrative value. As near as I can tell from your description, the presence of even one harmful mutation is enough to terminate the offspring. The assumption that any harmful mutation is fatal is unwarranted. Some mutations are trivially harmful and won’t affect survivability.”

    This was precisely the point of my comment on the genetics of diploidy and Mendelian dominance and its relevance to this discussion. However, the moderators have seen fit to disallow any discussion of these points, at least when posted by me.

    “Second, you make no provision for beneficial mutations.”

    Not only that, but the model under discussion also does not take into consideration mutations that are deleterious when homozygous, but beneficial when heterozygous. Not including this in the model renders its applicability completely nugatory, along with virtually all of the arguments based upon it.

  43. I need to amend my previous comment slightly for more accuracy. The fourth sentence in the first paragraph should read:

    Some mutations are trivially harmful and won’t affect survivability in a single generation, like you are assuming.

  44. Let me get this straight.

    In the video of Richard Dawkins’ WEASEL program, as we see the letters tick over on the display, we are watching the output of a computer model designed to illustrate that cumulative selection can reach a pre-selected target phrase faster than a purely random search.

    The gingerbread men video, on the other hand, is only an animation, not the output of a computer program, designed to illustrate the effect on a population of one harmful mutation per generation with no other factors being taken into account?

  45. My comments have finally appeared, almost 24 hours since they were posted. Interesting; as they appear in the order that existed when I posted them, rather than when they were approved, one would either have to know that they existed or start reading from the beginning of the comment thread to notice them. This means that most people who have been on this thread since the beginning have either not noticed them, or have conveniently not referred to them.

    IOW, delaying the posting of comments from the opposition serves to quite effectively remove them from the flow of the debate, even though they are eventually allowed out of moderation.

    Or have I made a false “design inference”?

    Let’s find out: how long will this comment linger in moderation. I am posting it on Sunday 15 November 2009 at 13:00 EST.

  46. Sorry: the last paragraph in my comment #42 should not have been a blockquote. It was my response to the second half of hummas man’s comment #37.

    Furthermore, after reading the objections and defenses posted here, it seems to me that the video depicts only a qualitative (i.e. not quantitative) process that may or may not be valid, and which could only be applied to haploid organisms (such as bacteria) and which lacks any empirical verification.

    But I do worry about the long-term prospects for gingerbread men, especially considering the impending advent of the Christmas season. To quote Ferdinand the Duck,

    “Christmas is carnage!”

    …indeed

  47. Sal in comment #40:

    “Model: one new harmful mutation per new born human. All other details will generally be moot with respect to the final conclusion.” [emphasis added]

    Thank you, Sal. With this clarification, it is very clear that your qualitative model bears no relationship to either a rigorously defined mathematical moderl nor biological reality at all, for the reasons that I have outlined in comments 17, 42, and 45 (assuming they eventually appear).

    That is, until you have factored into your model the following:

    1) diploidy (in humans)
    2) Mendelian dominance
    3) penetrance
    4) the degree to which the mutations are deleterious
    5) the degree to which the mutations are either selectively neutral or selectively beneficial among heterozygotes

    it cannot be reasonably applied to any actual biological system except haploid organisms such as bacteria (and certainly not to humans or any other diploid eukaryote), and then only if points 4 and 5 are addressed.

    I hope you will not take this as a criticism of yourself or your motives, only in the adequacy of your model and the reasoning upon which it is based. It is not meant in the former sense, as I hope you realize.

    P.S. This year is turning out to be an unusually good one for Finger Lakes wines. I hope you can drop by and join me in a winery tour the next time you are in Ithaca (or environs).

  48. jistak,

    Mom has mutation J1
    Dad has mutation J2

    Average number of harmfuls in parent generation : 1.0

    I designate their children as Kid_1, Kid_2, etc.

    We assume each of their kids will have 1 new harmful.

    The kids may or may not inherit a harmful from mom or dad (basic Mendelian inheritance).

    To illustrate the idea, here is the list of harmfuls for each kid, where K1 is the novel muation for Kid_1, K2 is the novel harmful for Kid_2 etc.

    Kid_1: K1

    Kid_2 : K2 J1 (from mom)

    Kid_3 : K3

    Kid_4 : K4

    Kid_5 : K5 J2 (from dad)

    Kid_6 : K6

    Kid_7 : K7

    Kid_8 : K8 J1 (from mom) J2 (from dad)

    let us tally the number of harmfuls for each kid

    Kid_1: 1

    Kid_2 : 2

    Kid_3 : 1

    Kid_4 : 1

    Kid_5 : 2
    Kid_6 : 1
    Kid_7 : 1
    Kid_8 : 3

    The average number of mutants in the kids is

    (1 + 2 + 1 + 1 + 2 + 1 + 1 + 3)/ 8 = 1.5

    We can do the same for any other set of parents with the same initial condition that mom and dad start off with 1 mutation each.

    It is easy to see, that there will never be the case the children on average have less mutations than their parents.

    Thus by way of extension, under the premise of 1 new harmful per new born, it is easy to see that the children will on average have more mutations than the parents.

    If this is the case, do you think any value of fecundity or viability will change the fact that the children of parents will have on average more mutations than their parents?

    Please inform the reader the reasons you think there exists values of fecundity and viability which will ensure the average number of mutations in the children will be less than their parents on average.

    If you can’t do this, you aren’t doing much to refute the assertion that with respect to the final outcome, fecundity and viability are moot points given the premise of 1 new harmful per new born.

  49. Another scenario:

    Mom has mutation J1 J2 J3
    Dad has mutation J3 J4 J5

    Average number of parent mutations: 3.0

    it is remotely possible if Mom and Dad have 40 kids, we’ll have:

    Kid_1 : K1 J1 J2

    ….

    Kid_39 : K39

    Kid_40 : K40

    Kid_39 and Kid_40 live, the rest are killed (truncation selection)

    Average number of mutations of living children is 1.0.

    This is he case described by Nachman using truncation selection. But the reduction is only temporary, and it can’t reduce the average number of mutations per individual below one. And realistically speaking, truncation selection doesn’t happen in the wild, not to mention, if there is insufficient reproducitve excess, the average number of mutations in the children will be greater than 1.0. Thus over time, on average, the number of mutations in kids will keep increasing.

  50. Sal in comments 48 and 49:

    Once again, you have neglected to factor in dominance and recessiveness, and so your analysis is irrelevant.

    P.S. Please see what you can do about removing me from permanent moderation, so that my comments will be timely, rather than afterthoughts. Thanks!

  51. Sal:

    In a thread at Telic Thoughts on this same subject (see XX ), you implicitly acknowledged my criticisms by stating

    Müller’s ratchet will apply to large linkage blocks in the genome and the Y-chromosome of humans, the Y-chromosome and linkage blocks exist in sexually reproducing species like humans.

    I agree that Müller’s ratchet should apply to alleles located in the y chromosome of mammals (as well as to alleles in the chromosomes of haploid eukaryotes and the DNA of prokaryotes and viruses), but I do not see how it could apply to alleles located in “large linkage blocks” or anything else located in the autosomes or X chromosome of mammals or other eukaryotes. In my understanding, “large linkage blocks” are just another name for chromosomes. Indeed, that’s how they were first defined in the early 20th century, before the congruence between linkage groups and chromosomes was empirically verified in the early 20th century by Bridges, Stephens, and Sturtevant.

    If you disagree, please explain how Müller’s ratchet might apply to alleles located in the autosomes or X chromosome of mammals (or provide a link to such an explanation). Thanks!

  52. Salvador,

    If this is the case, do you think any value of fecundity or viability will change the fact that the children of parents will have on average more mutations than their parents?

    Please inform the reader the reasons you think there exists values of fecundity and viability which will ensure the average number of mutations in the children will be less than their parents on average.

    I can explain why in mutation-selection equilibrium (MSE) the average number stays the same from one generation to the next, even if children receive on average more than one deleterious mutation.

    The reason is recombination. As mutations accumulate before MSE is reached, there will be variation in the population in the number of mutations per chromosome. Therefore, when two partners mate, it’s likely that they have different numbers of mutations on each chromosome. Because of recombination, there will be variation among the offspring in the number of mutations they carry. Even if all offspring receive an extra deleterious mutation, if the variation in numbers of mutations is large enough, and if selection is strong enough, then the number of mutations need not increase from one generation to the next.

    If you don’t believe me, I have just coded a little c++ program that proves my point. Do you want to see it?

    Finally, I find it pretty sad that you make such grandiose claims in the OP based on a cartoon without an underlying model. If you would try to pull a stunt like this in the competitive world of science, you would be toast.

    cheers

  53. Continuing OT: any YEC should watch Ross vs Hovind (it won’t hurt OECs of course).

  54. 54

    I’m no geneticist, so please correct me If I misunderstand.

    But here’s an observation from an outsider perspective that might help:

    It would seem that the lack of any neutral or beneficial mutations invalidates the conceptual model.

    Wouldn’t an individual’s reproduction rate be dependent on the net effect of all mutations, instead of just on the number of harmful mutations?

    how can we justify ignoring the effect of beneficial mutations or mutations that are neutral or harmful at the time of introduction and then become beneficial at some later time due to changing circumstances?

    It would seem that the author is claiming that if all offspring are guarunteed to have a net harmful effect from mutation then genetic collapse of the population will surely follow.

    I don’t think anyone would disagree with this assertion, however the question is whether or not, in reality, all offspring are actually guaranteed to have a net harmful effect from mutation?

    I doubt this is the case but I am willing to be persuaded by a relevant study or logical explanation.

  55. Sal:

    For your model to be a reasonable representation of biological reality it must also take into account a variant of natural selection that can remove deleterious alleles from gene pools when effective breeding population sizes become very small. This form of natural selection is often referred to as either “negative selection” or “purging selection”. Here is a definition (from Wikipedia) with references, followed by three empirical studies showing that negative/purging selection does in fact take place:

    NEGATIVE SELECTION:

    Negative selection, in natural selection, is the selective removal of alleles that are deleterious. This can result in stabilizing selection through the purging of deleterious variations that arise. It is also known as purifying selection.

    Purging of deleterious alleles can be achieved on the population genetics level, with as little as a single point mutation being the unit of selection. In such a case, individuals bearing the allele selected against might simply have less offspring on average generation after generation.

    In the case of strong negative selection on a locus, the purging of deleterious variants will result in the occasional removal of linked variation, producing a decrease in the level of variation surrounding the locus under selection. The accidental purging of non-deleterious alleles due to such spatial proximity to deleterious alleles is called background selection.[1] This effect increases with higher mutation rate but decreases with higher recombination rate.[2]

    References:

    1. Charlesworth, B., Morgan, M. T. and Charlesworth, D. 1993. The effect of deleterious mutations on neutral molecular variation. Genetics 134, 1289-1303.

    2. Hudson RR, Kaplan NL (December 1995). “Deleterious background selection with recombination”. Genetics 141 (4): 1605–17. PMID 8601498.

    http://en.wikipedia.org/wiki/N.....election%2 9

    ************

    How are deleterious mutations purged? Drift versus nonrandom mating.

    Glémin S.

    Evolution. 2003 Dec;57(12):2678-87.

    Accumulation of deleterious mutations has important consequences for the evolution of mating systems and the persistence of small populations. It is well established that consanguineous mating can purge a part of the mutation load and that lethal mutations can also be purged in small populations. However, the efficiency of purging in natural populations, due to either consanguineous mating or to reduced population size, has been questioned. Consequences of consanguineous mating systems and small population size are often equated under “inbreeding” because both increase homozygosity, and selection is though to be more efficient against homozygous deleterious alleles. I show that two processes of purging that I call “purging by drift” and “purging by nonrandom mating” have to be distinguished. Conditions under which the two ways of purging are effective are derived. Nonrandom mating can purge deleterious mutations regardless of their dominance level, whereas only highly recessive mutations can be purged by drift. Both types of purging are limited by population size, and sharp thresholds separate domains where purging is either effective or not. The limitations derived here on the efficiency of purging are compatible with some experimental studies. Implications of these results for conservation and evolution of mating systems are discussed.

    http://www.ncbi.nlm.nih.gov/pubmed/14761049

    ************

    Testing alternative methods for purging genetic load using the housefly (Musca domestica L.).

    Meffert LM, Regan JL, Hicks SK, Mukana N, Day SB.

    Genetica. 2006 Sep-Nov;128(1-3):419-27.

    When a population faces long-term inbreeding, artificial selection, in principle, can enhance natural selection processes for purging the exposed genetic load. However, strong purge pressures might actually decrease fitness through the inadvertent fixation of deleterious alleles and allelic combinations. We tested lines of the housefly (Musca domestica L.) for the effectiveness of artificial selection to promote the adaptation to small population size. Specifically, replicate populations were held at average census sizes of 54 for nine generations or 30 for 14 generations while being subjected to artificial selection pressure for increased fitness in overall mating propensity (i.e., the proportion of virgin male-female pairs initiating copulation within 30 min), while also undergoing selection to create differences among lines in multivariate components of courtship performance. In the 14-generation experiment, a subset of the lines were derived from a founder-flush population (i.e., derived from three male-female pairs). In both experiments, we also maintained parallel non-selection lines to assess the potential for natural purging through serial inbreeding alone. Sub-populations derived from a stock newly derived from the wild responded to artificial selection for increased mating propensity, but only in the short-term, with eventual rebounds back to the original levels. Serial inbreeding in these lines simply reduced mating propensity. In sub-populations derived from the same base population, but 36 generations later, both artificial selection and serial inbreeding increased mating propensity, but mainly to restore the level found upon establishment in the laboratory. Founder-flush lines responded as well as the non-bottlenecked controls, so we base our major conclusions on the comparisons between fresh-caught and long-term laboratory stocks. We suggest that the effectiveness of the alternative purge protocols depended upon the amount of genetic load already exposed, such that prolonged periods of relaxed or altered selection pressures of the laboratory rendered a population more responsive to purging protocols.

    http://www.ncbi.nlm.nih.gov/pubmed/17028969

    ************

    Purging of inbreeding depression within the Irish Holstein-Friesian population

    Sinéad Mc Parland, Francis Kearney and Donagh P Berry

    Genetics Selection Evolution 2009, 41:16doi:10.1186/1297-9686-41-16

    The objective of this study was to investigate whether inbreeding depression in milk production or fertility performance has been partially purged due to selection within the Irish Holstein-Friesian population. Classical, ancestral (i.e., the inbreeding of an individual’s ancestors according to two different formulae) and new inbreeding coefficients (i.e., part of the classical inbreeding coefficient that is not accounted for by ancestral inbreeding) were computed for all animals. The effect of each coefficient on 305-day milk, fat and protein yield as well as calving interval, age at first calving and survival to second lactation was investigated. Ancestral inbreeding accounting for all common ancestors in the pedigree had a positive effect on 305-day milk and protein yield, increasing yields by 4.85 kg and 0.12 kg, respectively. However, ancestral inbreeding accounting only for those common ancestors, which contribute to the classical inbreeding coefficient had a negative effect on all milk production traits decreasing 305-day milk, fat and protein yields by -8.85 kg, -0.53 kg and -0.33 kg, respectively. Classical, ancestral and new inbreeding generally had a detrimental effect on fertility and survival traits. From this study, it appears that Irish Holstein-Friesians have purged some of their genetic load for milk production through many years of selection based on production alone, while fertility, which has been less intensely selected for in the population, demonstrates no evidence of purging.

    http://www.gsejournal.org/content/41/1/16

  56. You should also check out “founder-flush” speciation, as this is a variant of negative/purging selection.

  57. A commentator at another ID website (Mung at http://telicthoughts.com/mutat.....ent-248496 ) pointed out that Müller’s Ratchet should also apply to mitochondrial DNA. To which I replied that, indeed, Müller’s Ratchet would apply to mitochondrial DNA, and also to the DNA of chloroplasts. In my (lame) defense, I usually think of mitochondria and chloroplasts as modified prokaryotes, as they almost certainly evolved from prokaryotic ancestors via serial endosymbiosis (as proposed by Copeland and expanded by Margulis).

    Which raises an interesting point: since the mitochondria in every multicellular eukaryote are virtually always inherited via mitochondria contained in the maternal egg cell, it would seem likely that there would be a “founder-flush” event that purges deleterious mutations every generation, as the sample of mitochondria in each egg cell would be an almost infinitesimal fraction of the total mitochondrial population of the eukaryotic mother (who made the egg cells). If this were indeed the case, it would explain why there is an unusually low frequency of deleterious mutations in mitochondria (with the exception of some rare forms of muscular dystrophy).

  58. ok, so Sal has offered some adjustments to his model. I think this is an important step forward.

    So what we have now is a rate at which mutations are introduced (1 per individual in the population), and a rate at which mutations are inherited (.5 per individual in the population).

    Now how many generations does it take us before even these figures are known to be absurd? (I am using for an anology the chessboard, with one grain on the first square, two grains on the second, etc.)

    Also, we still need to know the level at which the accumulated mutations become lethal (else we have no reason to terminate our gingerbread man).

    But since everyone in the population is accumulating at the same rate and has the same probability, why wouldn’t the entire population go extinct at once?

    So I think the animation should show accumulation, then “poof,” no more gingerbread men.

    And if we can extrapolate backwards in time, I think it becomes obvious that we could never be here.

    Therefore, I conclude, we are all merely a part of a computer simulation.

  59. Just a little musing – If Scordovas or any ‘model’ of mutation accumulation is run in reverse does it not lead through a serial ‘removal’ of mutations to a specific ‘perfect’ gene with no ‘mistakes’ ?

  60. A little addendum to my post #52.

    The problem with Salvador’s examples in 48 and 49 is that they do not allow for variation between parents in the number of harmful alleles they carry. No variation = no response to selection. If he had allowed for (inevitable) variation, then the average number of harmful alleles per offspring may decrease (depending on the details) by more than one during selection, thus compensating for any new harmful mutation acquired by the offspring.

  61. Allen:

    P.S. Please see what you can do about removing me from permanent moderation, so that my comments will be timely, rather than afterthoughts. Thanks!

    I working on starting another discussion forum. Not only in that place will you not be moderated, but you can even be a moderator if you should ever choose.

    I don’t have the authority to have you removed from any moderation here at UD. I’m sorry. As you can see, even I post infrequently at UD. I cannot even post at most threads at UD without being on the moderation queue (a peculiarity of the software which is a legacy of DaveScot which no one here has been able to uncork).

    Please accept my regrets. The most I can offer is we can reconvene elsewhere where we are not under so much restriction.

    I apologize for the delays in your display of comments. Thank you for your informative criticisms.

    I will attempt to address them, and also acknowledge where I feel you have made a valid critique, and where I must amend or withdraw my claims.

    Thank you again for your willingness to provide pointed and insightful objections.

    Sal

  62. jistak:

    I can explain why in mutation-selection equilibrium (MSE) the average number stays the same from one generation to the next, even if children receive on average more than one deleterious mutation.

    The reason is recombination. As mutations accumulate before MSE is reached, there will be variation in the population in the number of mutations per chromosome. Therefore, when two partners mate, it’s likely that they have different numbers of mutations on each chromosome. Because of recombination, there will be variation among the offspring in the number of mutations they carry. Even if all offspring receive an extra deleterious mutation, if the variation in numbers of mutations is large enough, and if selection is strong enough, then the number of mutations need not increase from one generation to the next

    I alluded to that issue in comment Comment #49

    Average number of parent mutations: 3.0
    ….
    Average number of mutations of living children is 1.0.

    And you said:

    if the variation in numbers of mutations is large enough, and if selection is strong enough, then the number of mutations need not increase from one generation to the next

    Which I alluded to when I said:

    This is the case described by Nachman using truncation selection.

    But this assumes sufficient reproductive excess and truncation selection. Nachman himself said truncation selection was unrealistic.

    However, there is a point where if there is insufficient reproductive excess and several harmfuls being added to each offspring, even under truncation selection, the MSE will not exist, or at least the MSE point will be carrying so many mutations, one has to wonder if the organism is viable.

    There will be an improvement to future animations.

    I mentioned that there have to be correctons to in the animation in the OP:

    2. There is a refinement to the animation that is in order based on Nachman’s calculation of average removal rates of harmful mutations assuming truncation selection, …That is yet another modification for future animations. We’ll need also some technical research on the matter.

    The the dots of the parents are shown to be always inherited by the kids. That will happen only in situations where Muller’s ratchet directly applies (like Y-chromosomes) and female mitochondira (some controversy over possible recombination, but still a reasonable assumption).

    That adjustment needs to be made so that sometimes the dot from the parent is not always inherited. I mentioned this situation in comment #48.

    We can then add mothers and fathers and show the children with some dots from mom and some dots from dad, and some situations where none of the dots are inherited.

    Thanks for your criticisms. I find them valuable.

  63. Allen,

    Like you I’m pressed for time. Perhaps we can continue our discussion at TT where you will be freer to comment.

    You are correct that large linkage blocks are not subject exactly to the same kind of ratcheting as Muller’s ratchet as the Y-chromosomes, but there is comparable problem since bad mutations can be randomly carried over along with the good on a large linkage block.

    The notion came from John Sanford’s book. Perhaps that consideration should be given to using different terminology so as to avoid confusion.

    Thank you again for your valuable criticisms.

    I hope we can continue at TT later today.

    regards,
    Sal

  64. Back in 2006, there was an interesting ARN discussion with Sal on this very subject:

    http://www.arn.org/ubbthreads/.....st30321663

    Note: I post on ARN as ‘KC”

  65. This is based on a modification of Gregor’s Bookkeeper

    An original isogenic population with a fitness of one. Mutation is considered damage to an allele. If both alleles are damaged, then the fitness for the entire organism is zero. Otherwise, we can adjust the effect. With a slight deleterious effect of 0.1 per gene, the population is more stable than if the damage is completely hidden.

    Here is a typical scenario, each with a stable population after a hundred generations.

    Population = 100
    Offspring per individual = 2
    Recessive = -0.1
    Average Fitness = 0.62

    Population = 100
    Offspring per individual = 1.5
    Recessive = -0.1
    Average Fitness = 0.41

    Population = 200
    Offspring per individual = 1.5
    Recessive = -0.1
    Average Fitness = 0.53

    No beneficial mutations, no variance, no phylogenetic noise. Doesn’t change the overall result anyway.

    Population = 100
    Offspring per individual = 2
    Recessive = -0.0
    Average Fitness = 0.58

    With silent recessives (i.e. Recessive = -0.0), the Offsprings have to be at least ~1.9 per individual for a stable population.

  66. Note: I appreciate the removal of the ban on my comments. I’ll endeavor to address the current topic and add value to the discussion.

    Clarification: Per scordova’s scenario, mutation means damaged or harmful.

    If both alleles for *any gene* are damaged, the organism has a fitness of zero and is not viable. For *each gene* with one damaged allele, the organism’s fitness is reduced by one-tenth (depending on the setting of Recessive). Consequently, ten damaged genes means the organism is no longer viable.

    The algorithm uses Roulette Wheel Mating.

  67. 67

    scordova,

    Sorry for coming into this late, but I had a weekend offline for once.
    I’m very interested in these kinds of models. I understand that you haven’t implemented yours in a programming language, but do you have a written mathematical description of what is going on behind the scenes in your animation?

  68. 68

    For *each gene* with one damaged allele, the organism’s fitness is reduced by one-tenth (depending on the setting of Recessive). Consequently, ten damaged genes means the organism is no longer viable.

    Isn’t this a rather poor representation of reality? enough so to invalidate the model?

    In reality wouldn’t the reduction in fitness caused by any given mutation be a quasi-normally distributed variable? With the majority of mutations doing relatively little damage, but extreme cases certianly existing where a single mutation could cause either an arbitrarily large or and arbitrarily small reduction in fitness?

    why is it valid to do away with all that complexity and just assume that all mutations cause a given, constant amount of reduction in fitness?

  69. Outdoor_Engineer: Isn’t this a rather poor representation of reality?

    Gregor’s Teeth is not devised to be a model of all aspects of biology. It is designed to test the specific claim that if every newborn has a harmful mutation, then genomic deterioration is inevitable, and that no other factors matter. In fact, it depends on a number of variables, some of which are identified in the algorithm.

    Outdoor_Engineer:In reality wouldn’t the reduction in fitness caused by any given mutation be a quasi-normally distributed variable?

    The original Gregor’s Bookkeeper uses a gamma probability density function. It skews so that most mutations only have a very small effect, with only a few being strongly significant. It shows that fitness can increase even when the vast majority of mutations are slightly deleterious (within the limitations of the model).

    Outdoor_Engineer:With the majority of mutations doing relatively little damage, but extreme cases certianly existing where a single mutation could cause either an arbitrarily large or and arbitrarily small reduction in fitness?

    The accumulation of slightly deleterious mutations is an interesting question, but such a discussion cannot progress without resolving some misconceptions.

    Outdoor_Engineer:why is it valid to do away with all that complexity and just assume that all mutations cause a given, constant amount of reduction in fitness?

    Not all mutations cause the same change in fitness. Most are neutral—or nearly so, and deleterious mutations far outnumber favorable mutations.

  70. Along those lines:

    This video gives a overview of the effects of mutations (Kimura’s distribution):

    Evolution vs Genetic Entropy
    http://www.youtube.com/watch?v=mmbRbyv2PA0

  71. Salvador:

    However, there is a point where if there is insufficient reproductive excess and several harmfuls being added to each offspring, even under truncation selection, the MSE will not exist, or at least the MSE point will be carrying so many mutations, one has to wonder if the organism is viable.

    Of course, if the rate of harmful mutation is high enough, then deterioration and extinction is inevitable. Yet your original claim that one harmful mutation per offspring is sufficient to guarantee this is incorrect for the reasons I explained, and as Zachriel also pointed out.

    You can find some c++ code here to verify this. The parameters of the model are:

    N=population size
    K=litter size
    NC=#chromosomes
    NG=#biallelic loci per chromosome (0=good allele, 1 =bad allele)
    r=per chromosome probability of crossing over event (at most 1 event/chromosome)
    s=selection coefficient
    mu=expected # mutations per offspring (good->bad, bad->good equally likely)

    Individuals are non-selfing diploid hermaphrodites and mating is at random.

    Offspring viability=(1-s)^M, where M is the number of bad alleles. In other words, multiplicative fitness scheme and additive gene action.

    If you play around with it, you’ll see that even with mu>1, an equilibrium can be reached.

  72. Sorry, the link doesn’t work. Try this.

  73. The fact that a deleterious or neutral allele is linked to another allele in a “large linkage block” (i.e. in the same chromosome without an intervening crossover chiasma) doesn’t really apply here. Once again, it matters if the allele is recessive or dominant (and if dominant, what its degree of penetrance is). Furthermore, if a deleterious or neutral allele is linked to an allele that confers increased fitness (i.e. those individuals who have it survive and pass it on more often than other individuals who do not have it), then the increase in frequency of the beneficial allele will “drag” the deleterious and neutral allele frequencies along with it, so long as the benefit of the one allele outweighs the deleterious effect of the other.

    This phenomenon is similar to genetic drift, and has hence been referred to as “genetic draft” by John Gillespie, the population geneticist who discovered it (based on a suggestion from Will Provine). That is, the slightly deleterious or neutral alleles will “draft” along with the beneficial allele(s) so long as they are all linked in the same non-recombining chromosome.

    And again, if the deleterious allele is recessive, but beneficial when heterozygous, then your model is completely out to lunch (yum, yum, gingerbread men for dessert ;-).

  74. Link posted in another thread:

    http://mendelsaccount.sourceforge.net/

    Zachriel has looked at this program and commented on it.

  75. Of course, if the rate of harmful mutation is high enough, then deterioration and extinction is inevitable. Yet your original claim that one harmful mutation per offspring is sufficient to guarantee this is incorrect for the reasons I explained, and as Zachriel also pointed out.

    Before we move forward on this, the animation was the haploid or asexual case. There are obvious analogs to the issue of Y-chromsomal heredity……

    Before moving to the diploid case, is it correct to say 1 harmful per offspring in that case will lead to deterioration, independent of viability or fecundity, or for that matter selection strength?

    Thanks for you input.

    Sal

  76. Yet your original claim that one harmful mutation per offspring is sufficient to guarantee this

    But the animation I provided was haploid, so why is Zach modeling diploid?

    The haploid model will conceptualize the diploid model where the parameters are sufficient for deterioration.

  77. Salvador T. Cordova: But the animation I provided was haploid, so why is Zach modeling diploid?

    Everyone knows that gingerbread people include gingerbread men, gingerbread women and gingerbread children.

    You didn’t provide a model. and there was a lot of discussion about human reproduction. Even bacteria often exchange genes. Your argument was clearly provincial (human-centric).

    Genetic load in non-recombining organisms is much less because of the very large populations and smaller genomes. Most offspring are perfect clones. However, if your point is that bacteria are due for extinction, I think you might be mistaken.

  78. In comment #75, Sal Cordova wrote:

    “There are obvious analogs to the issue of Y-chromsomal heredity…”

    An interesting observation. When one examines the genomes of haploid organisms, what is immediately striking is that they tend to be very small. That is, they contain very little information, compared with the genomes of diploid eukaryotes. This includes the y chromosome of mammals, which contains almost no genetic material beyond the coding sequences for a handfull of proteins (including TDF, which causes undifferentiated embryos to develop into males).

    Why this huge difference? One testable (i.e. falsifiable) hypothesis is that, since all alleles in haploid genomes are “visible” to selection, it may be the case that deleterious mutations (and, perhaps, even neutral mutations) immediately subject to “purifying selection” and removed, leaving only those versions of the alleles that confer functionality.

    This would also explain why prokaryotes tend to exchange genetic material when their genomes are damaged (for example, when they are exposed to mutagens, such as UV light). To state this in terms of selection: those bacteria who have the ability to exchange genetic material would be much more likely to survive deleterious mutations, since they could use the exchanged genetic material to replace the mutated parts of their genomes.

    This idea — that sex (defined as exchange of genetic material) first evolved as a means of repairing damage to the genome (i.e. deleterious mutations) was most vigorously proposed by Lynn Margulis and Dorian Sagan in their (1990) book, Origins of Sex (see http://www.amazon.com/Origins-.....038;sr=8-1 ).

    “Before moving to the diploid case, is it correct to say 1 harmful per offspring in that case will lead to deterioration, independent of viability or fecundity, or for that matter selection strength?”

    Yes, with two qualifications:

    1) that selection does not remove the deleterious mutations, and

    2) that some compensating mechanism (such as diploidy and/or sex) does not mitigate the deleterious effects of the mutations.

    Please note as well that none of this applies to neutral mutations, which cannot be removed by selection (except by accident, such as genetic drift) unless they accumulate to the point that their replication entails a significant energetic cost.

  79. As I have pointed out in another thread (see http://www.uncommondescent.com.....ent-340238 ), sex (defined as the exchange of genetic material between individuals) was proposed as a solution to the problem of the accumulation of deleterious mutations in haploid organisms in the 1930s by Hermann Müller, the originator of “Müller’s Ratchet”, upon which Sanford’s GE model is based. Müller argued that the ability to counteract the accumulation of mutations was the principle selective advantage that diploid sex provides, and therefore can explain its evolution (but not necessarily its origin).

    So, Salvador, if you can indeed extend Sanford’s GE model (and, presumably, your gingerbread man simulation) to apply to diploid, sexually reproducing eukaryotes, I urge you to submit it to a peer-reviewed journal for evolutionary biology as soon as you have verified its validity. If the model can then be tested empirically using either field or lab observations, you would definitely be eligible for a Crafoord Prize (the “Nobel prize for biology”), as this topic (i.e. the evolutionary origins of sex) has been one of the hottest areas of debate among evolutionary biologists for a century and a half.

  80. In comment #77, Zachriel wrote:

    “…if your point is that bacteria are due for extinction, I think you might be mistaken.”

    Indeed, if Sanford’s model is valid and (for the reasons discussed here) applies mostly (perhaps entirely) to haploid organisms such as bacteria, then the continued survival of bacteria is surprising.

    Unless, of course, one assumes (as Dr. Behe does, at least in the case of the malaria parasite) that some (and perhaps all) bacteria (and other haploid organisms) are favored by the Intelligent Designer, who continuously “tweaks” the genomes of every line of prokaryotes to remove the deleterious mutations that would otherwise lead to their eventual extinction.

    Indeed, it would seem that Dr. Sanford’s GE model is contradicted by Dr. Behe’s observation that the malaria parasite has remained virtually unchanged over many millions of generations. Why has the genome of Plasmodium falciparum not degenerated according to Sanford’s model? Does the Intelligent Designer have a soft spot for Plasmodium falciparum, as Dr. Behe speculates in The Edge of Evolution? Or might natural selection actually provide a solution to what would otherwise be incommensurate paradoxes?

  81. Salvador,

    Before we move forward on this, the animation was the haploid or asexual case. There are obvious analogs to the issue of Y-chromsomal heredity……

    I am a bit confused now. If you say that the animation was the asexual case, then that seems to imply there is an underlying model. I thought there isn’t.

    Before moving to the diploid case, is it correct to say 1 harmful per offspring in that case will lead to deterioration, independent of viability or fecundity, or for that matter selection strength?

    If there is zero recombination and every offspring has at least one new deleterious mutation and zero beneficial mutations, then yes that is correct I think. But are these assumptions reasonable? I think not.

  82. I am a bit confused now. If you say that the animation was the asexual case, then that seems to imply there is an underlying model. I thought there isn’t.

    The animation clearly depicts one gingerbreadman giving birth to two offspring each. That doesn’t look like asexual haploid reproduction
    to you?

    If you say that the animation was the asexual case, then that seems to imply there is an underlying model. I thought there isn’t.

    The underlying model 1 new harmful per new offspring. Dawkins model wans’t substantially more detailed than mine, yet is has been celebrated and defended by Darwinists.

    If there is zero recombination and every offspring has at least one new deleterious mutation and zero beneficial mutations, then yes that is correct I think.

    Thank you!

    But are these assumptions reasonable? I think not

    Unless we find a comparable case that is approximated in diploid organisms.

    For starters, we have the Y-chromosome in humans and femail mitochondrial DNA (widely believed to be subject too little if not zero recombination).

    Secondly, the issue of large linkage blocks (more on that later).

    And thirdly, their may be a point where the mutation selection equilibrium breaks down and can be approximated by the deterioration in the haploid model.

    This was the original intent to begin with a haploid model that would convey the basic point. and then show it’s approximate analogy to relevant diploid cases.

    More on these, in subsequent posts.

    By the way, does the haploid model presented in the animation remind you of Muller’s Ratchet? :-)

  83. scordova: The animation clearly depicts one gingerbreadman giving birth to two offspring each. That doesn’t look like asexual haploid reproduction
    to you?

    Hmm.

    scordova: Nobel Prize winner HJ Muller (of Muller’s ratchet fame) suggested that the human race can’t even cope with a harmful rate of 0.1 per new born. The actual rate has been speculated to be on the order of 100-300.

    The animation uses a conservative harmful rate of 1 and argues (with some attempts at humor) that deterioration would thus be inevitable even with a harmful rate of 1 per new born.

    You are extrapolating the animation to the human race.

    scordova: And if the conclusion symbolized by the animation is true, then on what grounds can we believe Darwinism is true?

    Indeed, extrapolating to all of life!

    As I mentioned, organisms that don’t recombine tend to have small genomes, so most of their offspring are exact clones; and also have large populations, meaning fixation of nearly neutral mutations take a very long time.

  84. Sal in comment #82:

    “The animation clearly depicts one gingerbreadman giving birth to two offspring each. That doesn’t look like asexual haploid reproduction to you?”

    Yes, it does, but it is just as clearly not sexual reproduction, unless one defines the horizontal genetic transfer and binary fission of bacteria as sexual reproduction. Once again, your simulation does not model sexual reproduction in diploid eukaryotes (and certainly not in humans), as there is no indication in the simulation that the mutations can be “hidden” by dominant, non-mutant alleles.

    “Unless we find a comparable case that is approximated in diploid organisms.”

    As I have repeatedly pointed out, there is no comparable case in diploid organisms, except for alleles located in the Y chromosome of mammals (and some, but not all haploid eukaryotes, such as cellular slime molds and some fungi, and then only during part of the life cycles of those organisms).

    Secondly, the issue of large linkage blocks (more on that later).

    “Large linkage blocks” is just another term for chromosomes. When Müller described his “ratchet” in the late 1930s, it wasn’t entirely clear that the “linkage groups” discovered by Bridges, Stephens, and Sturtevant were exactly the same thing as eukaryotic chromosomes. Now we know they are, and know that two processes (and perhaps more) eliminate the problem of Müller’s Ratchet in eukaryotes:

    1) sexual recombination (via independent assortment during metaphase I of meiosis, followed by gametic fertilization), and

    2) chromosomal recombination (i.e. “crossing-over between homologous chromosomes during prophase I of meiosis, followed by first-division segregation of alleles in meiosis I).

    These two processes, combined with diploidy, completely eliminate Müller’s Ratchet, which explains why Müller proposed it as the groundwork for his hypothesis for the evolution of sexual reproduction in eukaryotes.

    “…their may be a point where the mutation selection equilibrium breaks down and can be approximated by the deterioration in the haploid model.

    If you can find one, publish it as soon as possible! No one else has found one in nearly a century of population genetics research.

    “…to begin with a haploid model that would convey the basic point. and then show it’s approximate analogy to relevant diploid cases.

    As far as anyone has ever been able to determine (either theoretically or empirically), there is no analogy to relevant diploid cases.

    “…does the haploid model presented in the animation remind you of Müller’s Ratchet?”

    Of course it does. That’s because Sanford’s GE model is based on Müller’s Ratchet (as I have repeatedly pointed out), and is therefore not applicable to diploid eukaryotes (nor even to haploid organisms that can exchange genetic material horizontally…which appears to be almost all of them).

  85. Allen wrote:

    Large linkage blocks” is just another term for chromosomes.

    Thank you for your comment, and this important enough to clarify.

    I don’t believe that is Sanford’s usage.

    Another common term is Haplotype blocks.

    The most obvious and extreme form of selection interference is when there is tight physical linkage between benefician and deleterious mutations. This results in an irreconcilable problem referred to as “Muller’s Ratchet”. One of the most obvious requirements of natural selection is the ability to separate good and bad mutations. This is not possible when good and bad mutations are physically linked.

    John Sanford, Genetic Entropy
    p. 81

    and

    Essentially all of the genome exists in large linkage blocks (Tishkoff and Verrelli, 2003)

    John Sanford
    page 81

    The reference to Tishkoff and Verrelli is Patterns of human genetic diversity: implications for human evolutionary history and disease

    The say “low levels of linkage disequilibrium (LD)”. I read that to mean high linkage. Anyone agree or disagree?

    Dr. Sanford interprets this to mean the genome is composed of large linkage blocks. I have seen medical papers that rely on the assumption of large linkage blocks to detect diseases. If they use a few Single Nucleotide Polymorphisms to identify the presence of a large linkage block.

    PS
    This study seems mildly relevant:

    Human Haplotype Block Sizes Are Negatively Correlated With Recombination Rates.

  86. If you can find one, publish it as soon as possible!

    Simulations are being done. God willing they will be published. Walter ReMine informs me 10 mutations with truncation selection will result in mutational meltdown in diploid population with comparable reproduction rates as humans.

    The enigma has not escaped Muller, Crow, Kondrashov, Nachman, Crowell. They assume sysnergistic epistasis (a form of truncation selection). To make the matter of synergistic epstasis moot, the simulation is run with truncation selection.

    The whole thing of arguing over fitness and function can be very amorphous and nebulous and speculative. As Lewontin points out, fitness is hard, if not impossible to define since it is context dependent (unlike mass in classical physics).

    So where can we determine if there is genetic deterioration without direct reference to fitness and function?

    Sternberg unwittingly hints of a fruitful area of research:

    Another counterintuitive result of the ENCODE project and other comparative genomic analyses is that known functional sections of the mammalian genome such as protein-coding segments appear to be diverging without constraint, whereas a host of “junk” sequences are under some type of selective pressure—including most human “noncoding” DNA stretches. The same has been repeatedly detected for the fruit fly genome, where most nonprotein-coding sequences appear to be under functional constraint—with the species-specific differences having the statistical hallmarks of being “adaptive”. Even the Y chromosome of the fruit fly, long presented as “exhibit A” in the gallery of garbage DNA, has been shown to have diverse effects on the phenotype of this insect. Such results are exactly the opposite of what Orgel and Crick and Doolittle and Sapienza predicted.

    from
    How the junk DNA hypothesis has changed

    If realtime SNP growth is unabated in these “conserved” regions then we have reason to believe natural selection had little to do with their orgination.

    If these “conserved” regions are functional, then an increase in the SNP in these regions in the current day (real time measurments in newborns today, and subsequent tracking) this would suggest selection had nothing to do with the orgination of these functions.

    It would also suggest whether selection can even select against these defects.

    I hypothesize, the directly measured increase of the SNP’s will be inconsitent with accepted mainstream ideads. This implies genetic deterioration of some sort.

    We’ll see. I have faith empirical results over time will help decide which ideas are closer to the truth.

    We simply need to look at DNA of great grand parents, grand parents, parents, and kids, to get some good estimates of the real-time rise in SNPs. It is heartening to see that Solexa and Illumina technology is helping to inform the debate (as I predicted it would a few years back here at UD).

    The study I referenced above estimates a rate of 100 mutations per individual per generation.

    If 90% of the genome is functional, it is hard to conceive that the human race could tolerate 90 functional defects per new born for very long without suffering. I’m not saying humans will go extinct.

    Microporidia didn’t go extinct, it just lost most of its functionality and was reduced to being a parasite. We all might hope we don’t live to see that day for humans.

  87. Mr Cordova,

    Microporidia didn’t go extinct, it just lost most of its functionality and was reduced to being a parasite.

    That is exchanging cause and effect. Better to say that because of its success in the niche of parasite, Microsporidia lost functionality it no longer needed (or even interfered with being a good parasite).

  88. Salvador,

    For starters, we have the Y-chromosome in humans and femail mitochondrial DNA (widely believed to be subject too little if not zero recombination).

    Y chromosomes tend to degenerate, presumably because of lack of recombination. That fits nicely with Muller’s ratchet. Also as expected, we don’t see the same thing with autosomes. As for mitochondrial DNA, many mitochondria are transmitted by the mother to the egg, so there is a much smaller probability that deleterious mutations will get fixed. Also, much mitochondrial DNA has migrated to the nucleus over millions of years.

    Still, there are some interesting differences between the evolution of nuclear and mitochondrial DNA. I recently attended a lecture by Jerry Coyne, where he described the evolutionary genetics of speciation in a pair of sister species of Drosophila. Although there was little introgression between species at the level of nuclear DNA, there was enormous introgression at the level of mitochondrial DNA. It seems that the mitochondria of one species can easily spread in another species, even if hybridization occurs extremely rarely. An evolutionary puzzle!

    And thirdly, their may be a point where the mutation selection equilibrium breaks down and can be approximated by the deterioration in the haploid model.

    Come on. That’s just hand waving. It’s obvious that with sufficiently high mutation rates, even recombination cannot rescue from mutational meltdown. But you need to use a diploid model with recombination to figure out what the likely conditions are for this to happen, and then you can compare those conditions with empirical data. I suggest you get working on that and show us the results, and we’ll give you some feedback. Perhaps, in the end, the results can be published in Evolution, one of my favorite journals!

    Keep up the good work.

  89. I think the darwinists jumping in with their nitpicking and asking for unreasonable levels of detail are utterly missing the point. scordova’s work here is certainly a thought-provoking presentation, and can also be used as a platform for visualization of mutations spreading through populations. As well as rigorous mathematical and biological lab work the ID movement of course has to do public outreach and this visualization work can help with both.

  90. Sal

    We simply need to look at DNA of great grand parents, grand parents, parents, and kids, to get some good estimates of the real-time rise in SNPs. It is heartening to see that Solexa and Illumina technology is helping to inform the debate (as I predicted it would a few years back here at UD).

    This sounds like a project that could be undertaken and provide some significant proof for a reasonable price. Will you commit to undertake the required work, now that the cost of sequencing is much much lower and certainly within the reach of organizations and people connected to the, for example, Discovery Institute. The most recent cost I saw was $350,000 but that was some time ago. Alot of money, sure, but if it were to prove ID then I’m sure somebody would pay. And I’m sure it’s cheaper today then it was yesterday.

    to get some good estimates of the real-time rise in SNPs

    What would your prediction be, specifically, of the number you’d expect to find to confirm ID and the number you’d find and then conclude that ID was disconfirmed?

    Unless you are specific now, in advance, you’ll be accused of making a postdiction.

  91. I think it’s fair to call it an imaginative piece of work.

  92. Victor Tussle,

    What would your prediction be, specifically, of the number you’d expect to find to confirm ID and the number you’d find and then conclude that ID was disconfirmed?

    Unless you are specific now, in advance, you’ll be accused of making a postdiction.

    What’s wrong with a postdiction? Is there anything different in that and a prediction, logically?

  93. Victor Tussle:

    This sounds like a project that could be undertaken and provide some significant proof for a reasonable price. Will you commit to undertake the required work, now that the cost of sequencing is much much lower and certainly within the reach of organizations and people connected to the, for example, Discovery Institute. The most recent cost I saw was $350,000 but that was some time ago. Alot of money, sure, but if it were to prove ID then I’m sure somebody would pay. And I’m sure it’s cheaper today then it was yesterday.

    The project of tracking the rise of SNP’s and other mutations is being done independent of ID. It is done because of the intense medical interest in tracking hereditary diseases.

    And unlike Darwinism which is quick to mislable something as “beneficial” so long as it makes more babies, the medical community is more inclined to view mutations as a bad thing. See my objection to characterizing things as “fit” when by any other standard, they are diseased: Survival of the Sickest, Why We need Disease.

    Consider:

    Number of Rare Genetic Conditions on the Rise

    The number of people living with identified rare genetic diseases is on the rise in Australia, because more conditions are being recognised.

    It is thought that up to 1.5 million Australians are now living with unique, rare and often recently identified genetic conditions which go largely under the radar.

    “It isn’t so much that the number of people affected are going up it’s more that the number of diseases that we recognise is increasing,” says University of Melbourne Professor of Medical Genetics Bob Williamson.

    “There are now seven or eight thousand different genetic diseases known to occur, although many of them only affect a handful of children.”

    Is this merely more recognition? We’ll see.

    The enigma is, why have these diseases persisted (selection ain’t working so well). Thousands upon thousands of hereditary diseases persisting for centuries.

    We’ll see.

    In the mean time there is another area of research that could cast doubt on Darwinism, purifying selection, the enigma of Ultra Conserved Sequences.

    More in a subsequent comment on this thread.

  94. Jistak wrote:

    Come on. That’s just hand waving. It’s obvious that with sufficiently high mutation rates, even recombination cannot rescue from mutational meltdown. But you need to use a diploid model with recombination to figure out what the likely conditions are for this to happen, and then you can compare those conditions with empirical data.

    And you don’t think statements like the following by Nachman aren’t handwaving by the same standard?

    we estimate that the genomic deleterious mutation rate (U) is at least 3. This high rate is difficult to reconcile with multiplicative fitness effects of individual mutations and suggests that synergistic epistasis among harmful mutations may be common.

    However, many mutations are unconditionally deleterious and it is improbable that the reproductive potential on average for human females can approach 40 zygotes. This problem can be overcome if most deleterious mutations exhibit synergistic epistasis; that is, if each additional mutation leads to a larger decrease in relative fitness

    While extreme truncation selection seems unrealistic, the results presented here indicate that some form of positive epistasis among deleterious mutations is likely.

    Just invoke “synergistic epistasis” and mutations magically become damaging enough to weed it out the population. Just invoke “positive epistasis” and magically the bad mutations become good. This line of deduction is known as circular reasoning.

    No model for viability, selection coefficients, fecundity (all the complaints you and Mung lodge against me).

    Fine.

    But I will make a subsequent comment about this regarding ultra conserved sequences and the potential for an unabated rise in SNP’s and the impotence of purifying selection.

    By the way, thank you for agreeing that there is likely a point where there are enough mutations when no amount of selection can be of help. Muller said for humans, 0.1, Nachman (if there is no Synergistic Epistasis), 3.

    I will argue reasons empirically why synergistic epistasis is unlikely based on results of ultra “conserved” sequences between mice and humans.

  95. What would your prediction be, specifically, of the number you’d expect to find to confirm ID and the number you’d find and then conclude that ID was disconfirmed?

    Unless you are specific now, in advance, you’ll be accused of making a postdiction.

    Prediction: The ultra “conserved” regions will experience an unabated rise of SNP’s (single nucleotide polymorphisms).

    This will be inconsistent with the claim purifying selection (the purging of “bad” mutations”) has been operating on the genome of mice and men for the last 100 million years.

    BERKELEY, CA — Three years ago, “ultraconserved elements” were discovered in the genomes of mice, rats, and humans. These are DNA sequences 200 base pairs in length or longer — some are over 700 base pairs long — showing 100-percent identity among the three species. They have been perfectly conserved since the last common ancestor of mice, rats, and humans, which lived some 85 million years ago.

    These and other highly conserved sequences are thought to have persisted with little or no change because they are indispensable, performing functions vital for viability or reproduction. Scientists in the Genomics Division of the Department of Energy’s Lawrence Berkeley National Laboratory and DOE’s Joint Genome Institute set out to test this hypothesis by engineering four different “knockout” mice, each lacking one selected ultraconserved element.

    If truly indispensable, mice lacking an ultraconserved element should either die or be unable to produce viable offspring. Remarkably, as the researchers report in the September, 2007 issue of PLoS Biology, the knockout mice in this study showed almost no ill effects at all.

    So much for synergistic epistasis. :-) Synergistic epistasis would predict instant death or ill effects. Hardly a dent eh? Maybe Nachman’s “fix” to his own paradox is no fix at all, but rather circular reasoning refuted by hard nosed empiricism.

    If these ultra-conserved regions are not being purified and purged via purifying selection, then why does it seem they have been arranged like a bunch of coins oriented to all heads?

    These knockout experiments are prima facie evidence that selection can’t see these regions, yet somehow they appear to be magically ordered like a bunch of coins oriented heads, or plagiarism, or (gasp) common design.

    Unabated rise of SNP’s in this region would confirm selection has a hard time seeing these regions.

    Finally, Kimura showed by cost arguments (comparable to the ones presented in the animation), that 90% of molecular evolution is non-Darwinian. This is symboliized by the much celebrated and widely accepted “Neutral Theory of Molecular Evolution”. If 90% of the molecules in the genome are functional, and 90% of the molecules are not subject ot selection, the implication is that most of the functioning in the genome was totatlly originated independent of Darwninian blindwatchmaker processes. This is not “argument from ignorance” but rather a brutal proof by contradiction.

    Thus we have found design and function not attributable to natural selection. Where then is the source of this design and function? Whatever the source, we can have good reason to remove Darwinism as a plausible explanation.

    Continued degeneration of function will be evidence selection can’t arrest deterioration. If selection can’t arrest deterioration, circumstantially speaking, it would cast doubt that it had anything to do with the creation of such function.

    And in light of Kimura’s work and Sternberg’s hypothesis, it should seem painfully evident that selection doesn’t account for the creation of most function in the genome, and perhaps most of the rest of the human body.

  96. For the readers’ benefit, not all designs need to be “functional” in the sense of reproductive success.

    In the world of human affairs, if one came across a configuration of coins that were all heads, or like the configurations (where H=heads, T=tails):

    HTTTT HTTTT HTTTT HTTTT HTTTT

    or

    HHH HTHTTT HHH THTTHT HHH …..

    etc.

    These are recongizable linguistic type designs. Whether they have function or not, they are recognizable designs.

    The ENCODE project discovered comparable linguistic architectures in the genome. When I examined some of the sequences myself using data mining tools, the design was astonishing, definitely non-random (the above coin examples were non-random). These designs are especially found in the regions often thought to be expendable junk, hence not subject to selection (like those ultra conserved sequences).

    These designs are subject to deterioration. If they are non-functional, they will be especially immune to the mechanisms of selection to purge them even if we assume sexual recombination mechanisms.

    The haploid animation that I put forward can then be modified to deal with the diploid case, and it can be seen that selection will not arrest the deterioration of these designs.

    So these linguistic designs could be powerful designs that can’t be attributed to selection, if there is a rise of SNP’s in these regions in the present day, real time.

    Also, even on the assumption these are regions not subject to selection (from the prejudicial view that they were Junk DNA), how then do they have linguistic structure?

    Linguistic structures are recognizable designs. I gave some illustration with coins what linguistic designs could look like.

    Now, what would be especailly bad for the Blindwatchmaker argument is if these linguistic regions are both functional and invisible to selection. I’ve provided arguments above why I think that is the case based on Kimura and Sternberg.

    Further, testing of the real-time rise of SNP’s in these regions would be empirical confirmation that selection is mostly impotent save a few exceptional cases.

    The animation would thus be materially correct provided it were modified to illustrate the diploid case and say Nachman’s number of U=3 (U=number of mutations per individual). The work published in 2009 on human mutation rates using solexa technology suggests U = 100.

  97. scordova: And you don’t think statements like the following by Nachman aren’t handwaving by the same standard?

    we estimate that the genomic deleterious mutation rate (U) is at least 3. This high rate is difficult to reconcile with multiplicative fitness effects of individual mutations and suggests that synergistic epistasis among harmful mutations may be common.

    However, many mutations are unconditionally deleterious and it is improbable that the reproductive potential on average for human females can approach 40 zygotes. This problem can be overcome if most deleterious mutations exhibit synergistic epistasis; that is, if each additional mutation leads to a larger decrease in relative fitness

    There are several differences. Nachman isn’t trying to “prove evolution.” It’s already well-established, hence he is interpreting his finding in the light of the known facts, and in the light of plausible mechanisms. In addition, he’s not claiming to have demonstrated synergistic epistasis, but suggesting it as an avenue of additional research.

    scordova: Nachman’s number of U=3 (U=number of mutations per individual).

    That’s Nachman’s number of deleterious mutations. The total number of mutations is much higher. What recent study indicates that deleterious mutations are that high in humans?

  98. Everyone knows that gingerbread people include gingerbread men, gingerbread women and gingerbread children.

    You didn’t provide a model.

    You keep repeating falsehoods zach.

    Model: Haploids, 1 new harmful per newborn.

  99. Analogous Diploid case:

    Mentioned in the OP at UD.

    U=3, 3 harmfuls per newborn

    Result:

    Kid gets about 1 or more harmfuls from each parent. Kid gets 3 additional.

    Why is this. With 40 kids, 2 might have no mutations from parents carrying 3 mutations. If parents have fewer than 40 kids, then, it’s likely all the kids have 1 mutation or more!

    See the resemblance now to the haploid model.

    Did I get the chance to make these clarifications? Maybe, but I was so distracted by the strawmen and derailments being put forward I didn’t get the chance to add these. I make amends now.

    But the data was right there in Nachman’s paper.

    Comprende?

  100. What recent study indicates that deleterious mutations are that high in humans?

    I made the inference several times in this thread.

    The nature paper said about 100 new mutations. Sternberg said about 90% functional. This would strongly suggest U = some number around 100.

    Whether this leads to meltdown is another story. The issue is whether selection can police these functional regions. Which it obviously does not (in light Kimura’s work). Ergo: Selection doesn’t create most of the function in the genome.

  101. scordova: Model: Haploids, 1 new harmful per newborn.

    Is that the average mutations per individual? What is the distribution of effects of the harmful mutations? How often do reversions occur? How frequent are beneficial mutations and what is their distribution? What is the population? The size of the genome? What about recombination which is common in haploid organisms?

    scordova: Sternberg said about 90% functional.

    You still didn’t point to any such study. What specific research has Sternberg done to support the claim?

  102. Is that the average mutations per individual?

    Minimum. See the discussion above.

    How about trying to model it with the other parameters to your choosing.

    Feel free to report the results. :-)

    Do you think that Jistak’s provisional agreement with me is generally well founded regarding 1 harmful per newborn in haploids?

    You still didn’t point to any such study. What specific research has Sternberg done to support the claim?

    See the Evolution and News Report for a bibliography. Disagree with Rick? Do you needed peer-reviewed by Darwinists before you can even consider the conclusions.

    Feel free to express on what grounds specifically Sternberg is wrong.

    We’ll know in the coming decades if junk DNA is junk.

    PS
    I do express my thanks that you’ve tried to stick to the topics at hand and you’ve never attacked me personally despite the fact that I’ve not always been so nice to you.

    For that I salute you.

  103. I do express my thanks that you’ve tried to stick to the topics at hand and you’ve never attacked me personally despite the fact that I’ve not always been so nice to you.

    Unlike myself, who has apparently expressed a revulsion for Salvador that can only be expressed as contempt and repugnance.

    Now I’ve been charged of making the claim here at UD that I find Sal repugnant, even though I cannot recall the occasion.

    See my response to Sal here:

    http://telicthoughts.com/mutat.....ent-248736

    I repeat what I said there: Sal, If I called you repugnant, I apologise.

    I can’t find anyplace in this thread where I have done so.

  104. Model: Haploids, 1 new harmful per newborn.

    Sal, this is not a model.

    Perhaps you mean to model the case of haploid organisms who experience 1 new harmful mutation per newborn.

    If that is the case, you need to develop such a model.

    Let us say that each newborn dies, as a result of the harmful mutation.

    GREAT! Now that is something we can actually model.

    For each newborn that enters the population, add one harmful mutation to that individual.

    For each harmful mutation in an individual in the population, remove that individual from the population.

    Are we making progress yet?

    Well, to answer that, we need to ask how well our model tracks with observations of actual organisms.

    Care to comment?

  105. Mung,

    It wasn’t in this thread, it was 2 years ago here at UD.

    Sal, this is not a model.

    Baloney. Look at the animation. The seleciton model doesn’t matter. Who gets killed doesn’t matter. With haploids, the number of offspring doesn’t matter.

    The model is minimally simple to drive home the theme: if there are enough harmfuls mutations, no model of selection can prevail against deterioration. Therefore with respect to haploids,the other details are moot. Do you understand what the word “moot” means.

    Care to comment?

    I just did. Do you understand what the word moot means with respect to the matter at hand.

    Hint: moot means with respect to any other possible modeling parameters or details, the parameters and details don’t affect the final outcome. Thus it doesn’t add anything to the clarity of the model.

    Perhaps you mean to model the case of haploid organisms who experience 1 new harmful mutation per newborn.

    Look at the friggin animation!!!! It illustrates visually what will happen!!!! Sheesh.

    Zach has yet to run it with those specifications. He’s been running the diploid model. I’ve objected, and Zach, who’s been quick to provide sim work, has been noticably quiet to run the simulation that is consistent the animation.

    Why the silence?

  106. 106

    scordova at 102,

    “Is that the average mutations per individual?”

    Minimum. See the discussion above.

    How about trying to model it with the other parameters to your choosing.

    Feel free to report the results.

    Now you’re just teasing.

    Why so coy? I’d really like to understand your underlying model, possibly to the extent of being able to implement it in software. Could you please describe it at the level of detail suggested by Zachriel?

  107. I’d really like to understand your underlying model, possibly to the extent of being able to implement it in software.

    Haploids: 1 new harmful per new born

    Try implementing that. Are you having problems counting up to 1?

  108. 108

    scordova at 107,


    I’d really like to understand your underlying model, possibly to the extent of being able to implement it in software.

    Haploids: 1 new harmful per new born

    Try implementing that. Are you having problems counting up to 1?

    Ah, it could be worse, I could be having problems demonstrating common courtesy.

    Zachriel at 101 politely detailed the additional information required to understand your model:
    Is that the average mutations per individual? What is the distribution of effects of the harmful mutations? How often do reversions occur? How frequent are beneficial mutations and what is their distribution? What is the population? The size of the genome? What about recombination which is common in haploid organisms?

    Without knowing those kinds of parameters, it is not possible to replicate your model. I am more than willing to go to the effort of doing so; I hope you are willing to provide the additional information.

Leave a Reply