445 thoughts on “Evolution Visualized

  1. JoeCoder: OMagain, welcome to our discussion! Selection is weakest in organisms with large genomes, long generation times, and low reproductive rates. So in general the more generations we’re able to observe, the less likely the species has an issue with too much deleterious load.

    I explained this a couple pages back but with so many comments it’s easy to miss it.

    There’s a funny quality to this line of inference. Reminds me of the dragon in Carl Sagan’s garage. There’s mutational meltdown, and if you don’t see it it just means the meltdown is weaker. So how long until it’s going to meltdown regardless?

  2. Petrushka, this is only an argument against evolution. I am not claiming that this is a positive argument for God or even any kind of design.

  3. Well just fuck 150 years of consilient evidence.

    If MA says it can’t happen, then it can’t. End of discussion.

  4. I’m going to ask again. can someone set the MA parameters to model the Lenski experiment and see what happens in 60,000 generations.

    If the program can’t do that, then what’s the point?

  5. JoeCoder: I am not claiming that this is a positive argument for God

    And there he goes with the appeal to motive fallacy. The reason we don’t accept it is because somehow we secretly hate God and if Mendel’s accountant argues against evolution, that must mean God must exist!. Yeah, yeah that must be it.

  6. Put another way, wouldn’t it make sense, if necessary, to modify the MA program to model an experiment that has actually been conducted to see if it tracks the observations?

  7. Petrushka did you see my comment on the previous page? I modeled it for 2000 generations and fitness stayed almost completely constant. 60k generations might take hours, but I don’t know why the program wouldn’t be able to do it, given the plot for 2000 generations. Maybe you could give it a try yourself?

  8. JoeCoder:

    So it could be none. You have no evidence to support the claim that MA models anything in reality.

    One of the main utilities of a model is to figure out what will happen in cases we don’t have the ability to observe.Theoretically, what would this evidence look like?

    It’s your model. Make a prediction and test it.

    So here’s our complete list of reasons why Mendel is an awful program that can’t be trusted:

    1.It was written by creationists–Ew!

    We have evidence of creationists lying for Jesus. While that doesn’t mean all creationists are untrustworthy, it would be foolish to trust any on matters of science that conflict with their religious views.

    2.It uses programming practices from 1994.

    It uses programming practices that were poor even in 1994, to the point that the code cannot be used to understand the model.

    3.None of us have read the documentation but we know it’s not sufficient.

    I read the intro paper you provided. It’s not sufficient to re-implement MA.

    4.Mendel predicts species go extinct from deleterious load, and we haven’t measured whether this actually happens.

    There is no evidence that MA models anything we observe in real world biology.

    5.It’s not accurate enough.
    6.It’s too accurate!

    It’s neither, because it has never been tested against real biology.

    Mendel’s Accountant looks a great deal like it was constructed to wow the rubes, not to test a scientific hypothesis.

  9. JoeCoder:
    Petrushka, my purpose in mentioning neanderthals was to highlight a possible case where deleterious load contributed to their extinction.And also to say that even if we knew exactly how they died, it would still be difficult to say whether load contributed to their extinction?

    So your example can’t be shown to support your claim.

    My main argument is that evolution can’t work in higher animals because del. mutations arrive faster than selection can remove them and they go extinct.

    And yet we observe just what you say can’t happen. Time to throw out reality again and cling to that model!

  10. If I assume that MA doesn’t show a change of fitness when the parameters are set to model the Lenski experiment, then i have to ask why it shows meltdown for mammals and such.

    Could it be that by ignoring such phenomena as gene duplication, it is ignoring a mechanism that protects against mutations? If a gene has duplicates and one of the dupes mutates detrimentally, what is the effect on overall fitness?

    This is not a rhetorical question. I’d like to know if population genetics addresses this. And whether it is a relevant question.

  11. JoeCoder: Petrushka did you see my comment on the previous page? I modeled it for 2000 generations and fitness stayed almost completely constant. 60k generations might take hours, but I don’t know why the program wouldn’t be able to do it, given the plot for 2000 generations. Maybe you could give it a try yourself?

    Is it even possible to set up the program with realistic values without requiring a supercomputer to run? I mean in the Lenski experiment, a typical flask contains several hundred billion bacteria, each with a >4 megabase genome.

    Putting in such parameters give me integer overflow memory errors.

  12. So many people here who just don’t understand Mendel’s Accountant and who seek to be critical of it even though they have never coded one themselves. That must be it.

  13. Mung: So many people here who just don’t understand Mendel’s Accountant

    Do you understand it?

    Do you know if it is accurate?

  14. JoeCoder: Mendel’s Accountant is wrong because it disproves evolution. But we KNOW evolution is true, so Mendel’s Accountant has to be wrong! I’m so glad that’s finally settled

    It would help if Mendel’s Accountant actually used the accepted mathematics and models of population genetics instead of relying on a wizard behind a curtain that no one can see.

  15. Rumraket, I’ve written a GA, so it must follow now that I understand GAs.

    So all the nay sayers here need do is write a Mendel’s Accountant for themselves and they will understand it.

  16. Long ago in this thread but actually earlier today:

    JoeCoder: In my simple Javascript version of Mendel’s probability selection algorithm, I give Barney fife a fitness of 1.0 and Genghis Khan a fitness of 1.1. Then I apply Mendel’s algorithm 1 million times to see how often each survives truncation. Genghis Khan survives about 545k times and Barney Fife survives 455k times. Dividing those numbers gives 1.2, meaning Khan 20% more likely to survive truncation than Barney Fife.

    I am wondering if this is actually too favorable, since I would think a fitness of 1.1. would mean Genghis Khan should only be 10% more likely to survive. But maybe I’m not accounting for something.

    Exactly, and others here have tried to point this out too. If the fitnesses stand in the ratio of 1.1 : 1.0 then Genghis Khan should be 10% more likely to survive. But owing to (perhaps) the truncation you find him 20% more likeliy to survive. The truncation part is not in standard population-genetic models.

    That’s why MA does not help me see what is going on — it has too many arbitrary processes that the authors probably thought were important to include. Similarly for the multiplication or division by a uniform random number.

  17. Mung:
    Rumraket, I’ve written a GA, so it must follow now that I understand GAs.
    So all the nay sayers here need do is write a Mendel’s Accountant for themselves and they will understand it.

    That was done at AtBC.

  18. Joe,

    That description by JoeCoder is confused on a number of levels. Details later, but for now I’ll just point out that JoeCoder’s Javascript test does not test how often Barney and Genghis survive truncation.

    It tests how often Genghis’s fitness remains greater than Barney’s after each has been divided by a separate random number from the interval [0,1).

  19. Dropping in to leave a few references for the gene duplication frequencies:

    High Spontaneous Rate of Gene Duplication in Caenorhabditis elegans

    The gene duplication rate in C. elegans is quite high, on the order of 10^−7 duplications/gene/generation. This rate is two orders of magnitude greater than the spontaneous rate of point mutation per nucleotide site in this species and also greatly exceeds an earlier estimate derived from the frequency distribution of extant gene duplicates in the sequenced C. elegans genome.”

    Basically just look at the references of this one:
    Rapid Increase in frequency of gene copy-number variants during experimental evolution in Caenorhabditis elegans

    Background

    Gene and genome duplications are the primary source of new genes and have played a pivotal role in the evolution of genomic and organismal complexity [1, 2, 3, 4]. The rates of spontaneous gene duplication and deletion are extraordinarily high and speak to the enormous potential of these structural variants for generating new adaptive variability [5, 6, 7, 8, 9, 10]. However, most gene duplicates are eventually lost from populations due to a variety of reasons: genetic drift or natural selection, inherent instability of tandem duplications, and relaxed selection against detrimental mutations [5, 11, 12, 13, 14]. Although, gene duplications and deletions contribute significantly to the immense standing genetic variation related to gene copy-number observed in natural populations [15, 16, 17, 18], the relative importance of genetic drift versus natural selection in determining their evolutionary fate remains obscure.

    Copy-number changes in evolution: rates, fitness effects and adaptive significance

    Abstract
    Gene copy-number differences due to gene duplications and deletions are rampant in natural populations and play a crucial role in the evolution of genome complexity. Per-locus analyses of gene duplication rates in the pre-genomic era revealed that gene duplication rates are much higher than the per nucleotide substitution rate. Analyses of gene duplication and deletion rates in mutation accumulation lines of model organisms have revealed that these high rates of copy-number mutations occur at a genome-wide scale. Furthermore, comparisons of the spontaneous duplication and deletion rates to copy-number polymorphism data and bioinformatic-based estimates of duplication rates from sequenced genomes suggest that the vast majority of gene duplications are detrimental and removed by natural selection. The rate at which new gene copies appear in populations greatly influences their evolutionary dynamics and standing gene copy-number variation in populations. The opportunity for mutations that result in the maintenance of duplicate copies, either through neofunctionalization or subfunctionalization, also depends on the equilibrium frequency of additional gene copies in the population, and hence on the spontaneous gene duplication (and loss) rate. The duplication rate may therefore have profound effects on the role of adaptation in the evolution of duplicated genes as well as important consequences for the evolutionary potential of organisms. We further discuss the broad ramifications of this standing gene copy-number variation on fitness and adaptive potential from a population-genetic and genome-wide perspective.

    Who wrote this?:

    I have no desire to keep defending widely accepted points (like most mutations being substitutions)

  20. “Rumraket, I’ve written a GA, so it must follow now that I understand GAs.”

    If you get a girl pregnant, do you think that means you understand fetal development?

  21. AhmedKiaan: If you get a girl pregnant, do you think that means you understand fetal development?

    I don’t have to get a girl pregnant. I’ve been a fetus!

    In fact, I am just an overgrown fetus right now.

  22. I’m wondering if anyone who takes MA seriously can imagine for us what we would expect a species at the edge of genetic meltdown to look like. What kind of symptoms would you expect to see?

  23. JoeCoder: Petrushka, this is only an argument against evolution. I am not claiming that this is a positive argument for God or even any kind of design.

    If genomes are going to pot as rapidly as Sanford et al. say, then life has not existed for long on Earth. Now say, “But I…” Go ahead. Make my day.

    Mendel’s Accountant was designed to generate “evidence” for a young Earth. We would not have heard anything from the developers if they had failed to satisfy that requirement. Now project their severe bias onto the “Darwinists.” Go ahead. Make my day. Again.

    JoeCoder:
    Tom English wrote:“The notion that one should infer the model from the code is abominable.”

    Tom, The original Mendel paper describes probability selection mode:

    So does the manual:

    And lo and behold, multiplying the fitness by a random value between 0 and 1 makes the probability of reproduction proportional to an individual’s fitness ranking within the population.That’s what I reproduced with my JS code.

    You did not. If you cannot understand the example that keiths gave you, then you’re not remotely qualified to discuss the matter. I have not the least doubt, however, that you can keep an argument going forever.

    Tom, have you read the paper that describes Mendel’s algorithm?Have you read the manual?

    Don’t get cute with me. It was you who, with no apparent compunction, directed people to the source code.

    I read the original paper years ago. (I’ve also seen Sanford fit an exponential curve to longevity figures from Bible, including the age of Jesus at the time of his crucifixion. The man is loony tunes.) Why do you suppose the authors introduced a genetics simulation system at a computing conference? I mean, it’s supposed to be generally useful, not just a means for gussying up arguments that genomes have been going to hell in a hand basket since the Fall of Man — right? Do you suppose they simply did not bother trying to get a description of the system through review by geneticists? I doubt it highly. I have to believe that they tried repeatedly, and failed. If people are discriminating against Mendel’s Accountant because it was developed by creationists, then why has it been in included in scholarly reviews of genetics simulation software? If word of the system has been transmitted by reputable scholars in reputable journals, then why has it been used in only one [ETA: published] study (excluding those by the authors of MA)?

    Have you bothered to search for recent reviews of genetics simulation software? I have.

  24. Hello Rumraket! Your first paper is in worms and not humans, and the others just say the rate is “high”. So it’s not quite what I need, but it’s still interesting that worms have as many duplications as point mutations.

    However this paper gives specific numbers for humans: “The human de novo singlebase mutation frequency is ~2 × 10^-8, corresponding to ~120 new point mutations per diploid genome… compared with point mutations, the number of de novo segmental deletions and duplications is one and two orders of magnitude lower”

    So there’s very roughtly 100 times more point mutations than duplications.

    Michael Lynch likewise estimates: “an average human gamete acquires approximately 38 de novo base-substitution mutations, approximately three small insertion/deletions in complex sequence, and approximately one splicing mutation. Transposable-element insertions, microsatellite instabilities, and segmental duplications and deletions of total or partial gene sequences will almost certainly sum to several additional events per gamete.”

  25. Patrick wrote:

    I read the intro paper you provided. It’s not sufficient to re-implement MA.

    I was able to figure out a great deal by reading the source code comments, although having looked at some new areas today I now agree it could be still improved, and that a detailed pseudocode listing would be better. Maybe if I ever find time I’ll join their project and see if I can improve it. Still, I don’t know of a better tool for modelling this than Mendel. I also have no reason to be suspicious when Mendel shows declining fitness for high mutation rates. On that, see my reply to Tom below.

  26. Petrushka wrote:

    If I assume that MA doesn’t show a change of fitness when the parameters are set to model the Lenski experiment, then i have to ask why it shows meltdown for mammals and such.

    Because mammals have a much higher per-generation rate of mutation. Deleterious mutations arrive faster than selection can remove them.

    Could it be that by ignoring such phenomena as gene duplication, it is ignoring a mechanism that protects against mutations?

    An interesting thought. Probably somewhat, yes. But two considerations:

    1. There’s only about 1 gene duplication for 100 point mutations. See my reply to Rumraket.
    2. Gene duplications can also be deleterious since you no longer have the correct number of gene products produced.

    But having Mendel model all the various mutational phenomenon would make it more complex and harder to verify, so it’s a tradeoff.

  27. Dr. Felsenstein wrote:

    The truncation part is not in standard population-genetic models.

    A clarification: When Mendel multiplies the fitness by a random number and then removes those with lower fitness, this is done in probability selection mode. Truncation selection mode does the same but does not first multiply by a random number, giving a stronger selection effect. A third mode, partial truncation selection, interpolates between these based on a constant given in the UI. In the papers published with Mendel, probability selection and partial truncation selection (k=0.5) are typically used, with the authors saying that probability selection is likely the most realistic.

    From my demo with Barney Fife and Genghis Khan it looks like Mendel is making probability selection twice the strength as it should really have. So this is being more generous to evolution than what it should. Although I’ll wait and see what kieths has to say about this.

    That’s why MA does not help me see what is going on — it has too many arbitrary processes that the authors probably thought were important to include.

    I agree that makes it more complicated. Yet above Petrushka is asking why Mendel doesn’t add greater complexity to account for gene duplication. Although having now looked further, there are some places in the Mendel source that I agree could use better clarification.

  28. Hey Tom,

    If genomes are going to pot as rapidly as Sanford et al. say, then life has not existed for long on Earth.

    I responded to this same argument earlier today. See here.

    I have to believe that they tried repeatedly, and failed.

    Do you have a source for this? In the Biologic Information volume, the half dozen or so papers that used Mendel passed peer review at Springer. But the journal was threatened with boycott by people who had never read the papers so they weren’t published. They then passed peer review again by World Scientific.

    I also don’t understand why Mendel is suspect for showing declining fitness when mutation rates are high? Take a look at the formulagiven earlier by Dr. F:

    Using the usual mutational load calculation, if the total number of mutations per haploid genome was 1 per generation, the mutational load should be 1 – e^-1 or about 0,63. The reproductive excess needed to survive this would be about 1.718

    A reproductive excess of 1.78 means 2.78 offspring per mother. This formula simplifies to offspering_per_mother = e^u. Although this unrealisticlly assumes all death is due to selection, even though it’s not. For u=10, that means each mother would need at least 22 thousand offspring. Dr. Moran even says “if the deleterious mutation rate is too high, the species will go extinct… It should be no more than 1 or 2 deleterious mutations per generation”. Now in my reply to Dr. F. I argue that the limit is a little higher than these numbers. But when Mendel shows declining fitness for u=10 (the value they use in their papers), it’s untrustworthy?

    Tom, in your view (or anyone else), what percent of the genome is functional (subject to del. mutations), and what is the highest number of mutations per generation can be tolerated w/o declining fitness?

  29. Hey Tom,

    If genomes are going to pot as rapidly as Sanford et al. say, then life has not existed for long on Earth.

    I responded to this same argument earlier today. See here.

    I have to believe that they tried repeatedly, and failed.

    Do you have a source for this? In the Biologic Information volume, the half dozen or so papers that used Mendel passed peer review at Springer. But the journal was threatened with boycott by people who had never read the papers so they weren’t published. They then passed peer review again by World Scientific.

    I also don’t understand why Mendel is suspect for showing declining fitness when mutation rates are high? Take a look at the formulagiven earlier by Dr. F:

    Using the usual mutational load calculation, if the total number of mutations per haploid genome was 1 per generation, the mutational load should be 1 – e^-1 or about 0,63. The reproductive excess needed to survive this would be about 1.718

    A reproductive excess of 1.78 means 2.78 offspring per mother. This formula simplifies to offspering_per_mother = e^u. Although this unrealisticlly assumes all death is due to selection, even though it’s not. For u=10, that means each mother would need at least 22 thousand offspring. Dr. Moran even says “if the deleterious mutation rate is too high, the species will go extinct… It should be no more than 1 or 2 deleterious mutations per generation”. Now in my reply to Dr. F. I argue that the limit is a little higher than these numbers. But when Mendel shows declining fitness for u=10 (the value they use in their papers), it’s untrustworthy?

    Tom, in your view (or anyone else), what percent of the genome is functional (subject to del. mutations), and what is the highest number of mutations per generation can be tolerated w/o declining fitness?

  30. I wrote a reply to Tom and I tried submitting it more than once, but it’s not showing up. I don’t see it in the queue of comments. I’m new here so it’s probably some part of the system I don’t know about, or some kind of spam filter?

  31. Hey Tom,

    If genomes are going to pot as rapidly as Sanford et al. say, then life has not existed for long on Earth.

    I responded to this same argument earlier today. See here.

    I have to believe that they tried repeatedly, and failed.

    Do you have a source for this? In the Biologic Information volume, the half dozen or so papers that used Mendel passed peer review at Springer. But the journal was threatened with boycott by people who had never read the papers so they weren’t published. They then passed peer review again by World Scientific.

    I also don’t understand why Mendel is suspect for showing declining fitness when mutation rates are high? Take a look at the formulagiven earlier by Dr. F:

    Using the usual mutational load calculation, if the total number of mutations per haploid genome was 1 per generation, the mutational load should be 1 – e^-1 or about 0.63. The reproductive excess needed to survive this would be about 1.718

    A reproductive excess of 1.78 means 2.78 offspring per mother. This formula simplifies to offspering_per_mother = e^u. Although this unrealisticlly assumes all death is due to selection, even though it’s not. For u=10, that means each mother would need at least 22 thousand offspring. Dr. Moran even says “if the deleterious mutation rate is too high, the species will go extinct… It should be no more than 1 or 2 deleterious mutations per generation”. Now in my reply to Dr. F. I argue that the limit is a little higher than these numbers. But when Mendel shows declining fitness for u=10 (the value they use in their papers), it’s untrustworthy?

    Tom, in your view (or anyone else), what percent of the genome is functional (subject to del. mutations), and what is the highest number of mutations per generation can be tolerated w/o declining fitness?

  32. Hey Tom,

    If genomes are going to pot as rapidly as Sanford et al. say, then life has not existed for long on Earth.

    I responded to this same argument earlier today. See here.

    I have to believe that they tried repeatedly, and failed.

    Do you have a source for this? In the Biologic Information volume, the half dozen or so papers that used Mendel passed peer review at Springer. But the journal was threatened with boycott by people who had never read the papers so they weren’t published. They then passed peer review again by World Scientific.

  33. Hey Tom,

    Let’s get to the meat of the matter: You seem to think Mendel is suspect because it shows declining fitness when mutation rates are high? Take a look at the formula given earlier by Dr. F:

    Using the usual mutational load calculation, if the total number of mutations per haploid genome was 1 per generation, the mutational load should be 1 – e^-1 or about 0,63. The reproductive excess needed to survive this would be about 1.718

    A reproductive excess of 1.718 means 2.718 offspring per person, or 5.436 per mother. This formula simplifies to offspering_per_mother = 2e^u. Although this unrealistically assumes all death is due to selection, even though it’s not. For u=10, that means each person would need at least 44 thousand offspring. Dr. Moran even says “if the deleterious mutation rate is too high, the species will go extinct… It should be no more than 1 or 2 deleterious mutations per generation”. Now in my reply to Dr. F. I argue that the limit is a little higher than these numbers. But when Mendel shows declining fitness for u=10 (the value they use in their papers), it’s untrustworthy?

    Tom, in your view (or anyone else), what percent of the genome is functional (subject to del. mutations), and what is the highest number of mutations per generation can be tolerated w/o declining fitness?

  34. If genomes are going to pot as rapidly as Sanford et al. say, then life has not existed for long on Earth.

    I responded to this same argument earlier today. See here.

  35. I have to believe that they tried repeatedly, and failed.

    Do you have a source for this? In the Biologic Information volume, the half dozen or so papers that used Mendel passed peer review at Springer. But the journal was threatened with boycott by people who had never read the papers so they weren’t published. They then passed peer review again at World Scientific.

  36. There are six orders of magnitude more nucleotides than genes in the human genome. Of course the total amount of point mutations is going to be higher even if the RATE of duplications is larger than the RATE of pont mutations.

  37. Was he right?

    No, he was a total clown. He was asking stuff about whether mendel modeled haploid and diploiods. It does, he presumed it didn’t. And when Rob Carter pointed it out to him, he got that dazed look on his face.

    He asked about modelling of linkage. It does. He assumed it didn’t. He asked about mirgration, and stirring. It models it.

    But there are some outputs of Mendel where it agrees with accepted models. For example, a very simple test, does it give the right answer to kimuras famous equation of fixation rate = mutation rate

    https://en.wikipedia.org/wiki/Fixation_(population_genetics)#Probability_of_fixation

    That’s at least better than weasel, avida, tierra combined as far as fidelity that it models this accepted equation correctly.

    It also confirms the fixation rate of near neutral slightly deleterious.

    There are other classic predictions in population genetics. We can put it on the Mendel Engine and test if it agrees with predicted results. How about the Hardy Weinberg equilibrium.

    There are tests I’d suggest, like the Muller limit which Larry Moran, myself, and practically the whole pop gen community agrees is in the range of about 1 mutation per generation per individual that the human population can tolerate.

    The Muller limit, independent of Mendel is already accepted. Mendel is stating what even Captain Obvious should know (the pop gen version of Captain Obvious, that is). That’s already the heart of the genetic entropy argument. Weasel, Avida, Tierra disagree with accepted pop gen models, Mendel agrees. That’s why I don’t delve too much into Mendel. The important stuff, like the Muller limit is already accepted and as far as the Muller Limit, it’s results are like Captain Obvious.

    See:
    http://www.pnas.org/content/107/3/961.abstract

    Finally, a consideration of the long-term consequences of current human behavior for deleterious-mutation accumulation leads to the conclusion that a substantial reduction in human fitness can be expected over the next few centuries in industrialized societies unless novel means of genetic intervention are developed.

    Novel means? Like what, stopping the natural mutation rate that has been in existence for supposedly millions of years? Having couples produce 700 kids each and we kill off all but the 2 fittest? Genetic re-engineering (aka intelligent design)?

    There’s also lots of nutty stuff in evolutionary explanations of transposable elements. How do Alu’s magically invade the primate line, and then stop transposing for millions of years, and then one new Alu emerges and starts invading the genomes again? Variable transposition mutations that are punctuated for like a few generations and then go silent for 20 million years. How does one model punctuated mutation rates like that? We don’t have population genetic models to describe these scenarios. All the continuous time math models sort of fall apart with cataclysmic insertions of tens of thousands of Alu insertion mutations at once.

    Oh that’s the other thing, these ancient Alu (the J family), are they really ancient? Has anyone bothered to measure the sequence divergences in the Alu J family? Does it agree with the molecular clock predictions, or do they have the same sequence divergences as the more recent Alus (the Alu Ya5 Yb8, etc)? No one looks at this stuff. If they aren’t as diverged within the J family, then evolutionary claims would look kind of bogus (yet again).

Leave a Reply