Annoying creationists

[kjkent1]

A trick he learned from Socrates. Too bad we can't teach him the trick the Athenians finally figured out for dealing with it.

I wish no one harm -- just an opportunity to learn.

Socrates, used his method to narrow the argument to a reasoned conclusion, rather than to extend it ad infinitum. And, that makes all the difference.

 

[articulett]

If there be a poster elsewhere with 16,000+ posts it isn't me.

Here a couple years ago I lost about 3000 courtesy of a management fracas and resulting db re-org. As to spending way too much time on this forum, I plead guilty. Sue me.

I don't want to sue you. I just would like you to learn some of the coolest information humans have had the privilege of knowing. You don't even have to give up your intelligent designer or become a "fundamentalist materialist" or "closet dualist" to understand speciation and how we know what we know. You've heard of Francis Collins, right? The same technology that allows us to do forensic testing and paternity testing, is also useful for showing how closely species are related. It's really really interesting and cool. It's even pretty easy to understand. You see this information as some sort of threat to a paradigm, but it's just information that's really interesting to know--people here want to share it with you. It's information you can "decipher" for yourself--prove it or disprove it to yourself. But you seem to have a knee-jerk reaction that protects you from even examining the evidence, and I suspect that it's because you think your salvation is at stake. But Francis Collins is an evangelical Christian who heads up the official human genome project. You can at least be on the same page as everyone else regarding speciation which will lend credence to what you have to say. Read some of what he writes because you really do sound like Myriad's example. Really.

 

[Unnamed]

I haven't decided, though I'm leaning to the minimum absolute difference of just the binding sites. The total absolute difference probably makes more sense though, doesn't it?

Paul (and Myriad),

I confess that I didn't follow the last few days of the thread, but I just made this small change to Evj (in the scoreChromosome method), and I thought I would let you know.

!    sortValue = mistakes = spuriousHits = 0;
     for (int p = 0; p <= chromosome.length - width; ++p) {
       if ((valuation[p] = siteValuation(p, width)) >= threshold) {
         if (!siteInd[p]) {
           if (p < availablePos) mistakes += gene; else mistakes += nongene;
!           sortValue += Math.abs(valuation[p] - threshold);
           ++spuriousHits;
         }
       } else {
         if (siteInd[p]) {
           mistakes += missed;
!          sortValue += Math.abs(valuation[p] - threshold);
         } 
       }
     }

!    //sortValue = mistakes;               // We will sort on the mistake

As a result, the number of generations to perfect creature went down by a factor of 3.0 in my tests (7 runs for each version, pop. 128, sites 256, default for everything else). I ran a few tests with 512 sites and the effect is even more impressive.

What I did was, for each mistake, to add the distance between the valuation and threshold, and use the result as the sorting order. This is a finer measure than the raw number of mistakes, but it still goes to zero when mistakes = 0, so it converges to the same place.

I didn't touch anything that had to do with ties, but I guess that using sortValue instead of mistakes would help even more.

ETA: I ran examples with genome length 1024 and 2048, and the difference is just astonishing.

 

[Paul C. Anagnostopoulos]

Very cool, Unnamed! Thanks for trying that variation.

Would you be willing to run a set of experiments with your variant of Evj? If so, use these parameters:

population 32
binding sites 8
weight width 9
site width 10
1 mutation per 512 bases

Then vary the genome size from 1024 by factors of 2 and record the number of generations to perfect creature. I've run this model up to a genome size of 128K.

~~ Paul

 

[Kotatsu]

Each Kingdom has it's own challenges in interpretation, although Plantae doesn't get much play.
Your added comments will be welcomed.

I haven't had time to look at it today, but it is Song et al. 1995 (PNAS 92, 7719-7723) I referred to. I'll bring it home with me tonight and read it later. Reading through the abstract, though, it seems I must have imagined the "phenotypical" part (1).

I'd previously provide a link with a plethora of 'species definitions' that was compiled in 1998. I don't understand the 'barcoding' comment.

Barcoding is, briefly (2), an idea put forward by Paul Hebert and colleagues, which is based on the assumption that while the within-species differences of the COI gene will be quite low (0-5%, I think), the between-species differences will be much higher (10-20%; with a saturation at about 23% in oligochaetes at least). Between these two differences, there will be a "barcoding gap", which will make it possible to determine to which species (3) a given sample belongs, regardless of phylum.

[EDIT] To clarify: If comparing closely related species in, say, a family, there should in theory be no overlap between the curve that shows the distribution of within-species differences within the family an dthat describing the distribution of between-species differences.If the higher Within-species difference in 5% and the lowest between-species difference is 10%, addition of a new species of the same family should not result in any differences between this species and any of the other which is in the range of 5-10%. The actual ranges seem to differ widely between different groups of animals.[/EDIT]

In oligochaetes, this works quite well, actually. I've been working with several families (4), and while saturation is reached in most cases (i.e. division into subfamilies and higher taxonomical ranks is rare, ambiguous, and sometimes wrong), morphologically identified species fall in the same clades, implying that they are fairly homogenous. In some cases, unidentified samples have been placed in a certain clade, and this placement has been confirmed by morphological studies, so the idea seems to hold water. However, if anything, our work has revealed that there is likely a lot of undescribed oligochaetes out there, just waiting to be found; several "common" species are most likely not monophyletic, which will prove dangerous, as several of them are used as model organisms.

Sadly, however, the barcoding assumptions are not universal and - some suggest - not supported by data when sampling is done over a sufficiently large area. Samples with intermediate differences regularly appear, and the papers of Hebert and colleagues (the Costa Rican butterflies, the North American birds and others) with which they gallumph so proudly through current scientific journals apparently fall apart somewhat when larger geographical areas are examined. I attended a conference recently where there was an example of European water beetles (I have forgotten which genus, and it is not very essential). In this example, the barcoding gap failed to present itself once several populations from every area was used. And, of course, COI can't be used at all in Cnidarians. Still, it may be of some use if refined.

Er, and I almost hate to say it, but, A worm is a worm.

Certainly --- depending on how you define "worm" --- but my comment was rather in connection to the species concept part of barcoding, not as a comment on speciation. Even if speciation was proven to be impossible, but barcoding gave the same results as it does today, barcoding could still work as a species concept in at least some groups. If another gene, or a set of genes, is used, it might be used universally!

---
(1) However, the study is done on three Brassica species, and unless I am mistaken, allopolyploid progeny of these do show extensive phenotypical differences from the parental species. However, I am not a botanist.
(2) And possibly not very correctly.
(3) These species would initially have to be sorted out by other means, using other species concepts; barcoding - as I understand it - works only when you already have a basic database with which to compare.
(4) Namely: Almidae, Capilloventridae, Enchytraeidae, Lumbricidae, Lumbriculidae, Megascolecidae, Opistocystidae, Phreodrilidae, Propappidae, Randiellidae, and Tubificidae (including Naididae).

 

[kleinman]

Annoying Creationists

I don’t believe you can modify the selection process and reduce the number of mutation/selection cycles by orders of magnitude. Myriad is trying to do this and he is a good mathematician and we’ll wait and see. None of the other mathematicians who inhabit this site have offered a solution. In order for this model to support the theory of evolution, you need a selection process that is orders of magnitude faster than Dr Schneider’s selection process. I don’t believe that such a selection process exists.

If you're a scientist, then you will stop using the phrase, "I don't believe..." and start actually testing your hypotheses. You are obviously a good mathematician, so you should be able to describe the mathematics of possible selection methods, OR, you should reprogram the algorithm and test your theory, empirically. Otherwise, your repeated claims about impossible evolution are exactly the same sort of "religion" as you decry in your opponents (I'll continue to repeat this until you recognize that this is a deficiency in your own analysis of the problems with ev).

I believe that Dr Schneider’s selection process shows that macroevolution is impossible by random point mutations and natural selection. You evolutionarians need to prove your own theory. It appears that some are finally trying.

I did one small series where I varied site width up to 15 bases. In that series, the rate of convergence was independent of base width. I don’t know if that behavior would remain constant to a base width of 450. When I get a chance, I’ll try running more cases with a wider width to see if such a case is possible. In addition, you don’t have a selection process.

Thanks, that would be great.

This series will take quite a bit of time to run.

Also, you didn't address my last issue, and I think it's a reasonable one. So I'll repeat it:

You seem to be suggesting that some pre-hemoglobin molecule is floating around in the Pacific Ocean, just waiting to randomly form into de novo hemoglobin. Isn't reality more like there is some complex life form with an already developed genome, which first mutates for a selective advantage -- not necessarily hemoglobin -- and then over time, continues to evolve until one day hemoglobin, or something close to it is finally defined?

If this is how evolution really works, then your example of evolving a 450 base gene from a random genome, directly to hemoglobin, is irrelevant. It's just another version of the already-discredited creationist "tornado through the junkyard creates a 747" example.

I suggest nothing of the kind. It is evolutionarians who speculate that there was a time when there were small self replicating molecules that somehow evolved to the life forms we know today. There was a time when hemoglobin did not exist. The construction of this molecule (and associate gene) had to start from some point.

I'm going to implement a simple variant of his proposal. Right now, Ev has three methods for breaking ties between creatures when it goes to kill off the worst half and replicate the best half. The idea is to add a fourth method for breaking the ties. Tie-breaking turns out to happen quite often, especially with large populations (selective sweeps occur continuously due to Ev's simplistic model).

If two creatures are tied in their number of mistakes, Evj will select the one with the lowest absolute difference between the valuations of the binding sites and the threshold. This provides a finer-grained selection process which distinguishes between two creatures with the same mistake count, selecting for the one that is, in some sense, closer to reducing its mistake count.

It will be interesting to see how this speeds up the evolution of a perfect creature, if at all. It's all a bit of a tangent, since it was not the purpose of Ev to model realistic genetic variation.

Paul, are you using the total absolute difference of all error sites? Or of all binding sites plus all error nonbinding sites? Or something else (surely not all the potential sites in the whole genome)?

I have one qualm about this experiment, which is as much an aesthetic concern as anything else. I share Dr. Schneider's reluctance to allow the internal workings of the sorting algorithm to affect the results (though in the end he didn't manage to completely eliminate all incidental effects of the sort, hence what used to be termed 'deaths not due to selection'). Tiebreaking occurs only between pairs of organisms that happen to be equidistant from the beginning and end of the list after the sort. Hence, how much "churn" occurs from generation to generation within the tied population will effect how genetic changes selected by tiebreaking can spread through the population. If the same individuals tend to stay aligned with one another from generation to generation, transfer of tiebreaker-winning genomes through the population will be different than if the tied population gets more thoroughly mixed each generation.

Myriad makes a good point here. If you design a selection method that reduces the importance of errors in the nonbinding site region, you can reduce the effect of the increasing genome length. In the limit, if you choose a selection method that completely ignores errors in the nonbinding site region, the rate of convergence should become independent of genome length when you use a mutation rate fixed to a number of bases.

What you call evidence in the fossil record is open to interpretation.

Hand waving does not a convincing argument make.

Are you talking about evolutionarian hand waving or creationist hand waving about the fossil record?

What I am saying is that the de novo evolution of a gene is macroevolution.

Genes do not evolve de novo. That is in fact a contradiction in terms.

Ok, then explain how you can take a series of microevolutionary steps from the first base of the hemoglobin gene to the fully formed gene. Don’t forget to tell us what the selection process is.

The peer reviewers and editors of Nucleic Acids Research don’t seem to think so, they published Dr Schneider’s work.

That means they thought the paper was worth publishing, not that they agreed with it.

I exchanged some email with one of the editors. I think he agrees with it.

That’s only half the story. You have to show how these genomes can transform from one to another. Dr Schneider’s model shows that this is mathematically impossible (at least for random point mutations and natural selection).

And it is clear that Dr Schneider's model doesn't handle population genetics in any way remotely resembling reality, so the results from it are not useful in forming any conclusions relating to population genetics, and phyletic evolution clearly does relate to population genetics.

I love posting these quotes from Dr Schneider’s web site.

The following quotes were taken from Dr Schneider’s blog web page: http://www.lecb.ncifcrf.gov/~toms/paper/ev/blog-ev.html

The following are Dr Schneider’s responses to a critique of his paper Evolution of biological information by Dr Stephen E Jones.

"Schneider's paper is misleadingly titled: "Evolution of biological information". But it is just a *computer* simulation. No actual *biological* materials (e.g. genomes of nucleic acids, proteins, etc) were used, nor does Schneider propose that his simulation be tested with *real* genomes or proteins

Actual biological materials were used to determine the original hypothesis. Read the literature: Schneider1986

It only becomes *real* biological information and random mutation and natural selection, when the simulation is tested in the *real* world, using *real* DNA, proteins, with *real* mutations and a *real* environment does the selecting. It is significant that Schneider does not propose this, presumably because he knows it wouldn't work.

You are very bad at reading my mind, I have considered doing this experiment. Given the right conditions, it WILL WORK. Do you have th gumption to do the experiment yourself? That's the way real science works! FURTHERMORE, if you read the literature, you will recognize that related experiments have been repeatedly done for 20 years. Look up SELEX.

In the rest of the paper he uses the single word "selection". I take this as a tacit admission that his model is not a simulation of *real* biological natural selection.

No. A rose is a rose by any other name. Selection is selection whether it be natural (generally meaning the environment of earth), breeding (by humans usually, though perhaps some ants select their fungi), SELEX or in a computer simulation. Of COURSE it is a simulation of natural selection! The paper would not be relevant to biology and would not have been published in a major scientific journal if it were not!

Schneider lets slip that there is another unrealistic element in his (and indeed all) computer simulations in that it (they) "does not correlate with time":

So? Run the program slower if you want. Make one generation per 20 minutes to match rapid bacterial growth. THIS WILL NOT CHANGE THE FINIAL RESULT!

Well, when Schneider's simulation is actually tested with *real* "life" (e.g. a bacterium), and under *real* mutation and natural selection it gains information, then, and only then, would "creationists" be favourably impressed. But if they are like me, they would already be impressed (but unfavourably) that Schneider does not mention in his paper that his simulation should now be so tested in the *real* "biological" world.

1. The simulation was of phenomena in the "real" world.
2. Dr. Jones is invited yet again to do an experiment.

The following is a response Dr Schneider made to a statement made by David Berlinski.

Where attempts to replicate Darwinian evolution on the computer have been successful, they have not used classical Darwinian principles, and where they have used such principles, they have not been successful.

The ev program disproves this statement since it uses classical Darwinian principles and was successful.

The previous statements are clear that Dr Schneider believes that ev simulates the real world.

There are lots of examples like this. These are examples of microevolution which I believe occurs. It is the de novo evolution of genes that I don’t believe occurs.

NOBODY believes that de novo evolution of genes occurs.

That’s reassuring, because it is completely illogical to believe such a thing. So where do you believe genes came from?

I don’t argue that reptiles and birds do not have genes. What I argue is that you can’t make the transition from the reptile to the bird genome by random point mutations and natural selection. The process is profoundly slow, too slow to accomplish this transition. This is demonstrated by the ev computer model.

And as has been pointed out, there is no reason to believe that the computer model in question is a plausible representation of reality.

You might find that Dr Schneider doesn’t agree with your view. I am sure that I don’t.

Where did these existing genes come from?

They are existing genes. What sort of question is that?

Unless you believe that genes are eternal, they had to come into being somehow.

I’ll let Dr Schneider statement about his model answer this:

A good simulation does not attempt to simulate everything; only the essential components are modeled. For the issue at hand, the form of the genetic code is not relevant; information measured by Shannon's method is more general than that.

The form of the genetic code is absolutely relevant to the transmission of genetic modifications and to the evolutionary process; Dr Schneider's assertions to the contrary are disingenuous at best.

Do you think genetic code is eternal?

How would you know, have you even looked at the model?

Yes.

Since ev is a mathematical model, you should be able to enumerate what you think is wrong about the model. Why don’t you tell us what is wrong with this simulation of random point mutations and natural selection?

Paul wrote an online version of the model. You should study it. You would get an interesting lesson in the mathematics of random point mutation and natural selection. And as you have already said the following about recombination and natural selection:

But it is impossible to get phyletic evolution (or any major change, really) this way.

You are correct, you can not increase the information in the gene pool by recombination without error, but you can lose alleles (information in the gene pool) by recombination with natural selection.

Which explains nothing. We know that phyletic evolution cannot occur through recombination. We know that phyletic evolution occurs. We know that there is a mechanism for this. Your entire argument against this mechanism is a single computer model which we know is deeply flawed precisely in the way it models this mechanism. Therefore we have no reason at all to consider the argument valid.

What I am trying to explain here is that recombination and natural selection can give rapid phenotypic changes in a species. It is this rapid phenomena that Darwin inappropriately extrapolated to mutation and natural selection when he was interpreting the differences between finch beaks in isolated populations. Gould did the same when he postulated his concept of punctuated equilibrium. This concept is applicable to recombination and natural selection but not to mutation and natural selection.

The form of the genetic code is absolutely relevant to the transmission of genetic modifications and to the evolutionary process; Dr Schneider's assertions to the contrary are disingenuous at best.

The purpose of Ev is not to model genetic propagation in nature, but to demonstrate that information can evolve in a genome. For this purpose, the exact encoding of the information is irrelevant.

Please note: Most of the detailed behavior of Ev that we have been discussing has nothing to do with Schneider's original intent. He did not claim to model the complete evolutionary landscape, and certainly not phyletic evolution.

The fact that Kleinman has decided to co-opt Ev as the be-all-end-all of evolutionary modeling is his problem, not Dr. Schneider's.*

*Disclosure: I have only an informal working relationship with Dr. Schneider. I am not employed by him, nor have I been paid by him.

I tend to agree that this discussion has nothing to do with Dr Schneider’s original intent for ev, however Dr Schneider has gone far beyond his original intent. Consider what Dr Schneider has proposed on his FAQ page at:
http://www.lecb.ncifcrf.gov/~toms/paper/ev/faq-for-ev.html

Why don't you do a real biological experiment instead of just a computer model?

The primary reason is that we don't have infinite resources and time. If you have the resources (a molecular biology lab), are interested in doing an experiment, and would like to discuss it please contact me.

Paul, what part of the evolutionary landscape is Dr Schneider trying to model and then do a real biological experiment?

Don’t forget, I also co-opted the micro/macroevolution concept as well. There is so little else in the theory of evolution that is worth co-opting.

We know that phyletic evolution cannot occur through recombination.

Hmm. I wonder if it's possible that control regions could change substantially enough to cause a phyletic event?

Paul, where did you get that idea?

And if it did, then all you need to do is find it as factor in the fossil record to demonstrate phyletic gradualism is meaningful.

Consider that phylum are simply a categorization scheme that can be defined any way you want. Since Gould hypothesis of punctuated equilibrium fits more logically with recombination and natural selection, it is less likely to find evidence of gradualism (on the macroevolutionary scale) by the mechanism of recombination and natural selection.

The purpose of Ev is not to model genetic propagation in nature, but to demonstrate that information can evolve in a genome. For this purpose, the exact encoding of the information is irrelevant.

For that purpose, yes, you're quite right.

Too bad the author of the computer model says something different. And this is a peer reviewed and published model. Dr Schneider wants to do experiments verifying his model.

Please note: Most of the detailed behavior of Ev that we have been discussing has nothing to do with Schneider's original intent. He did not claim to model the complete evolutionary landscape, and certainly not phyletic evolution.

Ah. Sorry. I looked at some information about it and saw that it was hardly a comprehensive model, and did not support many claims made for it by third parties, but given your explanation it would appear that the model is adequate for its intended purpose, and the fault lies with the inappropriate claims and not the model.

Paul, continue to diminish the value of ev. Why did you waste your time writing the java version of the program?

The fact that Kleinman has decided to co-opt Ev as the be-all-end-all of evolutionary modeling is his problem, not Dr. Schneider's.

Quite, and I apologize to Dr Schneider for any offense my remarks may have caused.

As I told Dr Schneider months ago when we were discussing his model, when the full mathematical behavior of the model became known to evolutionarians, they would discredit his model. Now his own programmer has devalued the model. Paul, does ev have any scientific merit?

However, as my buddy drkitten likes to point out I'm not even scientifically semi-literate. Perhaps you can educate me. Once cladistics identifies the proposed common ancestor of those different species, can a genotype for that ancestor be proposed? If so, if not now, soon ... build one; let's see if it's a viable phenotype.

In general, I'm not sure how you can work backward completely accurately, given that the mutations that led to the different species are more or less random. And how would you ever know about features of the common ancestor that have disappeared?

Paul! You used the word random. So you can’t work backwards and ev shows you can’t work forward, which way are you going from here?

There is no way to explain mathematically all the new genes required to evolve reptiles to birds.

Oooh, you thought of a new lie!

Of course, you can't substantiate your halfwitted lie, 'cos it's a lie, but it does seem to be one you haven't told before.

Hey, look everyone, kleinman thought of a new lie!

Scatequate, why would I lie to you? You are my favorite annoyee. I would much rather use logic on you. Ev is my substantiation. You have heard of ev, haven’t you? Ev is a computer simulation of random point mutation and natural selection and when you use realistic parameters in the model, the rate of evolution is profoundly slow, too slow to evolve any gene de novo. Of course, when you get to infinity, everything happens in one generation.

Scatequate, we can try to get a cubicle for you next to kjkent1 so you can learn the mathematics of ev together.

Back near topic, do no responses to my question on definitions mean y'all haven't made up a new one since 1998? Paulie, where are 'ya? You're the guy who ceaselessly whines for definitions.

I ceaselessly whine for definitions when people insist on chit-chatting about concepts that appear to be undefined: god, free will, macroevolution. If you don't make claims about these things, I won't whine for definitions.

What about macroevolution=de novo evolution of a gene? Of course PixyMisa believes genes are eternal.

Circular arguments, where one begins by assuming the answer, bother me when presented as 'proof' the original assumption was correct.

Darwin started with the physical evidence and developed a theory to explain it. This theory has been vindicated many times over by multiple independent lines of evidence. There's no circular argument coming from the science camp. Implying that scientists need to agree on a perfect definition of species before they can explain the origin of species is as absurd as demanding they have a mathematically precise definition of "red" and "pink" before they can explain the origin of colors.

Darwin inappropriately extrapolated the rapid changes that can occur with recombination and natural selection to the profoundly slow mechanism of mutation and natural selection. Evolutionarians continue to make this same erroneous extrapolation. With respects to mathematically precise definitions of colors, that has existed for years. It is called the frequency spectrum of light.

Yup. I agree. Now all you need is a demonstrable macro-ev event.

Are you willing to define exactly what you mean by demonstrable here?

Macroevolution=de novo evolution of a gene.

I haven't decided, though I'm leaning to the minimum absolute difference of just the binding sites. The total absolute difference probably makes more sense though, doesn't it?

Paul (and Myriad),

I confess that I didn't follow the last few days of the thread, but I just made this small change to Evj (in the scoreChromosome method), and I thought I would let you know.

Finally, an evolutionarian is doing some mathematics. Unnamed, since you are modifying the java code, perhaps you or Paul would post the names of the corresponding variables and functions in the Pascal version of the program. Are you ignoring the “spuriousHits” and only using “mistakes” when you doing your sort?

 

[hammegk]

To an illiterate dolt such as myself Kotatsu's comments add to the complexity of defining 'species'.

Perhaps y'alls old standby talkorigins needs updating?

Their most significant summary of speciation is:

Confirmation:
Speciation of numerous plants, both angiosperms and ferns (such as hemp nettle, primrose, radish and cabbage, and various fern species) has been seen via hybridization and polyploidization since the early 20th century. Several speciation events in plants have been observed that did not involve hybridization or polyploidization (such as maize and S. malheurensis).

Some of the most studied organisms in all of genetics are the Drosophila species, which are commonly known as fruitflies. Many Drosophila speciation events have been extensively documented since the seventies. Speciation in Drosophila has occurred by spatial separation, by habitat specialization in the same location, by change in courtship behavior, by disruptive natural selection, and by bottlenecking populations (founder-flush experiments), among other mechanisms.

Several speciation events have also been seen in laboratory populations of houseflies, gall former flies, apple maggot flies, flour beetles, Nereis acuminata (a worm), mosquitoes, and various other insects. Green algae and bacteria have been classified as speciated due to change from unicellularity to multicellularity and due to morphological changes from short rods to long rods, all the result of selection pressures.

Speciation has also been observed in mammals. Six instances of speciation in house mice on Madeira within the past 500 years have been the consequence of only geographic isolation, genetic drift, and chromosomal fusions. A single chromosomal fusion is the sole major genomic difference between humans and chimps, and some of these Madeiran mice have survived nine fusions in the past 500 years (Britton-Davidian et al. 2000).

Need I say Mice is Mice?

 

[cyborg]

Every time I hear you bang on about somebody else doing mathematics I cannot help but recall your own misgivings about it kleinman.

And the fact I see precious little of it coming from you.

I see someone else's work, a program, and you, running a program (or not as the case may be when doing so would be inconvenient to you). I don't see that as YOU doing any work at all.

Unbelievably conceited.

 

[kleinman]

Annoying Creationists

Every time I hear you bang on about somebody else doing mathematics I cannot help but recall your own misgivings about it kleinman.

And the fact I see precious little of it coming from you.

I see someone else's work, a program, and you, running a program (or not as the case may be when doing so would be inconvenient to you). I don't see that as YOU doing any work at all.

Unbelievably conceited.

In case you haven’t notice, I did the parametric study with ev and posted these results which show the mathematical behavior of ev. I have run hundreds of cases with this mathematical model. So stop whining because you don’t like the results from this evolutionarian model which disproves your own theory. You need to fix your own mathematical models (if they can be fixed).

I know I am not saying much. All I am saying is the ev computer model shows that macroevolution by random point mutations and natural selection is mathematically impossible when realistic genome lengths and mutation rates are used in the model, that the model contradicts Gould’s hypothesis of punctuated equilibrium and that huge populations do not reduce the generations for convergence sufficiently to contradict either of the two propositions.

Anyway, I like co-opting evolutionarian ideas. The problem is there are so few worth co-opting.

 

[kjkent1]

Paul (and Myriad),

I confess that I didn't follow the last few days of the thread, but I just made this small change to Evj (in the scoreChromosome method), and I thought I would let you know.

!    sortValue = mistakes = spuriousHits = 0;
     for (int p = 0; p <= chromosome.length - width; ++p) {
       if ((valuation[p] = siteValuation(p, width)) >= threshold) {
         if (!siteInd[p]) {
           if (p < availablePos) mistakes += gene; else mistakes += nongene;
!           sortValue += Math.abs(valuation[p] - threshold);
           ++spuriousHits;
         }
       } else {
         if (siteInd[p]) {
           mistakes += missed;
!          sortValue += Math.abs(valuation[p] - threshold);
         } 
       }
     }
 
!    //sortValue = mistakes;               // We will sort on the mistake

As a result, the number of generations to perfect creature went down by a factor of 3.0 in my tests (7 runs for each version, pop. 128, sites 256, default for everything else). I ran a few tests with 512 sites and the effect is even more impressive.

What I did was, for each mistake, to add the distance between the valuation and threshold, and use the result as the sorting order. This is a finer measure than the raw number of mistakes, but it still goes to zero when mistakes = 0, so it converges to the same place.

I didn't touch anything that had to do with ties, but I guess that using sortValue instead of mistakes would help even more.

ETA: I ran examples with genome length 1024 and 2048, and the difference is just astonishing.

Unnamed (or anyone else who understands what's going on here),

Would you mind explaining to us mere mortals how your programming change relates to real-world biological behaviors. We wouldn't want some creationist, such as Dr. Kleinman to claim (1) that your modifications do not/cannot actually appear in nature, and/or (2) that you are "smuggling" an "intelligent design" into the process of evolution.

 

[cyborg]

In case you haven’t notice, I did the parametric study with ev and posted these results which show the mathematical behavior of ev.

Yes, I noticed. The point still remains - I don't actually see what you have done apart from run someone else's program and post the results.

So stop whining because you don’t like the results from this evolutionarian model which disproves your own theory. You need to fix your own mathematical models (if they can be fixed).

As has been pointed out so many times before:

1) ev is not the totality of evolution. It's not even close.
2) It was written to show one objective, which it shows. You have unreasonably extrapolated its predictive abilities far beyond its capabilities.

You might as well just go ahead and tell me that Einsteinian physics is wrong because the Newtonian mathematics of Quake doesn't demonstrate any of its effects.

I know I am not saying much. All I am saying is the ev computer model shows that macroevolution by random point mutations and natural selection is mathematically impossible

Yes you say that but still have failed to tell us what macroevolution is - MATHEMATICALLY. This is incredibly important because if you cannot quantify it then it cannot possibly be mathematically demonstrated as such. You simply throw out the assertion time and time again.

You cannot show macroevolution is impossible in ev if you cannot say what constitutes macroevolution - genome lengths and mutation rates be damned. They don't matter until you can tell us that.

 

[Paul C. Anagnostopoulos]

Would you mind explaining to us mere mortals how your programming change relates to real-world biological behaviors. We wouldn't want some creationist, such as Dr. Kleinman to claim (1) that your modifications do not/cannot actually appear in nature, and/or (2) that you are "smuggling" an "intelligent design" into the process of evolution.

The smuggling issue has already been addressed:

http://www.lecb.ncifcrf.gov/~toms/paper/ev/dembski/rebuttal.html

Ev's normal selection method is quite discrete: A creature with fewer binding mistakes is selected over one with more. These new schemes make a finer distinction between creatures with the same number of mistakes. It is as if stronger binding is favored over weaker binding, even when the number of sites bound is the same. The idea of weak binding appears in the literature:

http://www.pnas.org/cgi/content/abstract/102/20/7097

That said, a person can certainly argue that Ev is not modeling the precise kinds of selection that occur in nature. But Ev's point is simply to show that information can evolve in a genome.

~~ Paul

 

[Kotatsu]

To an illiterate dolt such as myself Kotatsu's comments add to the complexity of defining 'species'.

Please, if I am at some point too technical, confusing or vague, state which point that is, so that I may elucidate. Barcoding is nothing very difficult, but failure to communicate what I mean may have resulted, as English is not my first language.

However: I have now read Song's et al. article again, and they do mention phenotypic divergence as well, though only briefly:

"Phenotypic variation among F5 plants within each polyploid provided additional evidence for genetic divergence. Fertilities among F5 plants ranged from 0% to 24.9% for the AB/BA synthetics and from 0% to 100% for the AB/BA synthetics [Should probably read "AC/CA"]. F5 plants also varied for morphological traits, such as leaf shape and color, branching patterns, and number of side shoots."

A = Brassica rapa
B = B. nigra
C = B. oleracea
AB and BA corresponds to the natural polyploid hybrid B. juncea, and AC and CA to the natural polyploid hybrid B. napus. BC and CB would --- if I understand things correctly --- correspond to the natural polyploid hybrid B. carinata, but don't take my word for it.

F5 is the fifth generation (self-pollinated) of the various synthetic polyploid hybrids. Interestingly, the genetic variation between the F5 plants and their diploid parents are 9.6% for AB, 8.2% for BA, 4.1% for AC, and 3.7% for CA --- all just in five generations!

As an aside, I noticed in Soltis & Soltis, 1995 (PNAS 92, pp 8089-8091) a quotation to a study by Soltis et al., 1995 (Am J Bot 82, says "in press" in my paper, why I don't know the page numbers) wherein is shown that Tragopogon mirus and Tragopogon miscellus may have formed as many as 9 to 21 times just in the last 50 years (now 60 years) due to allopolyploidization. I shall see if I can get that article tomorrow.

Need I say Mice is Mice?

So you are looking for an example of speciation in which the resulting taxa is not only a new species, genus, or tribe but a new family or even new order?

 

[kleinman]

Annoying Creationists

In case you haven’t notice, I did the parametric study with ev and posted these results which show the mathematical behavior of ev.

Yes, I noticed. The point still remains - I don't actually see what you have done apart from run someone else's program and post the results.

That about sums it up. So why are you evolutionarians so annoyed?

So stop whining because you don’t like the results from this evolutionarian model which disproves your own theory. You need to fix your own mathematical models (if they can be fixed).

As has been pointed out so many times before:

1) ev is not the totality of evolution. It's not even close.
2) It was written to show one objective, which it shows. You have unreasonably extrapolated its predictive abilities far beyond its capabilities.

You might as well just go ahead and tell me that Einsteinian physics is wrong because the Newtonian mathematics of Quake doesn't demonstrate any of its effects.

So what is the relativistic effect in the theory of evolution that will fix the mathematics that ev reveals for random point mutations and natural selection?

I know I am not saying much. All I am saying is the ev computer model shows that macroevolution by random point mutations and natural selection is mathematically impossible

Yes you say that but still have failed to tell us what macroevolution is - MATHEMATICALLY. This is incredibly important because if you cannot quantify it then it cannot possibly be mathematically demonstrated as such. You simply throw out the assertion time and time again.

You cannot show macroevolution is impossible in ev if you cannot say what constitutes macroevolution - genome lengths and mutation rates be damned. They don't matter until you can tell us that.

# of generations to accomplish de novo evolution of a gene by random point mutations and natural selection > time available

As an example of this equation, Paul has estimated that it would take 65,000,000 generations to evolve 80 loci on a 100k genome with a population of 1,000,000. This is done with a highly idealized very accurate selection process on a genome that is still at least a factor of 5 shorter genome than any known free living organism, yet is still much, much slower that recombination and natural selection.

Would you mind explaining to us mere mortals how your programming change relates to real-world biological behaviors. We wouldn't want some creationist, such as Dr. Kleinman to claim (1) that your modifications do not/cannot actually appear in nature, and/or (2) that you are "smuggling" an "intelligent design" into the process of evolution.

The smuggling issue has already been addressed:

http://www.lecb.ncifcrf.gov/~toms/pa.../rebuttal.html

[/quote]
The code fragment Unnamed posted appears to be weighting errors in the binding and non-binding site regions. What is the value assigned to the variables “gene” and “nongene”. If you set nongene=0 then you are ignoring errors in the nonbinding site region of the genome. That is equivalent to ignoring harmful mutations.

Ev's normal selection method is quite discrete: A creature with fewer binding mistakes is selected over one with more. These new schemes make a finer distinction between creatures with the same number of mistakes. It is as if stronger binding is favored over weaker binding, even when the number of sites bound is the same. The idea of weak binding appears in the literature:

http://www.pnas.org/cgi/content/abstract/102/20/7097

That modification that Unnamed appears to be changing much more than the threshold.

That said, a person can certainly argue that Ev is not modeling the precise kinds of selection that occur in nature. But Ev's point is simply to show that information can evolve in a genome.

Now that’s a laugh. What kind of precise selection process that occur in nature would evolve a gene de novo? You must have drunk more than a few beers after celebrating that issue of agreement with hammegk in order to come up with this idea.

Sure ev shows that information can evolve in a genome. It just takes more time than the history of the earth to get a gene de novo.

 

[cyborg]

That about sums it up. So why are you evolutionarians so annoyed?

Annoyed? Hardly. It's more that it's staggering how arrogant you are with your call to 'evolutionists' to do the maths when you're basically saying that they have! You're the one who has done bugger all.

So what is the relativistic effect in the theory of evolution that will fix the mathematics that ev reveals for random point mutations and natural selection?

Way to go Mr Does-Not-Get-The-Point.

You using ev to disprove evolution is the equivalent of using Quake to disprove Einsteinian physics.

Do you get that Quake is an entirely insufficient representation of real world physics or not?

Do you get that ev is an entirely insufficient representation of real world evolution or not?

# of generations to accomplish de novo evolution of a gene by random point mutations and natural selection > time available

That does not answer the question.

That is your conclusion. I require the mathematical definition of macroevolution. That may require something like algebra. I would like a logical analysis of the ev program please. I would like to see the logical conditions for macroevolution occurring - an assessment of the change of state of the genome represented in the ev environment. An example. Anything tangible.

If you cannot tell us what a macroevolutionary event would be in ev I fail to see how you can tell us that ev shows it is impossible - irrelevant of how accurately it captures the real world processes.

 

[Unnamed]

Would you be willing to run a set of experiments with your variant of Evj? If so, use these parameters:

population 32 / binding sites 8 / weight width 9 / site width 10 / 1 mutation per 512 bases

Then vary the genome size from 1024 by factors of 2 and record the number of generations to perfect creature. I've run this model up to a genome size of 128K.

These are my results so far (genome size, mutations, run 1, run 2, run 3):
1024, 2, 190, 197, 206
2048, 4, 219, 210, 262
4096, 8, 256, 256, 268
8192, 16, 322, 357, 332
16384, 32, 394, 375, 366
32768, 32, 474, 474, 430
65536, 32, (running)
I am using the seeds 0, 1 and 2. The cases where the generations to perfect creature are the same have different seeds, so it's a coincidence or the software didn't update the seed.

Starting at 32768, the number of mutations was fixed at 32, because the application won't let me increase it and I didn't notice that until now.

I changed the tie-breaking algorithm to use sortValue (the sum of the strengths of the mistakes) instead of the number of mistakes. This helps a bit, but not much. I also added some overflow protection.

Finally, an evolutionarian is doing some mathematics. Unnamed, since you are modifying the java code, perhaps you or Paul would post the names of the corresponding variables and functions in the Pascal version of the program. Are you ignoring the “spuriousHits” and only using “mistakes” when you doing your sort?

I am not familiar with the Pascal version. I couldn't get it to run on my computer and never looked at the code. I am adding all kinds of mistakes.

Would you mind explaining to us mere mortals how your programming change relates to real-world biological behaviors. We wouldn't want some creationist, such as Dr. Kleinman to claim (1) that your modifications do not/cannot actually appear in nature, and/or (2) that you are "smuggling" an "intelligent design" into the process of evolution.

The smuggling issue has already been addressed:

http://www.lecb.ncifcrf.gov/~toms/paper/ev/dembski/rebuttal.html

Ev's normal selection method is quite discrete: A creature with fewer binding mistakes is selected over one with more. These new schemes make a finer distinction between creatures with the same number of mistakes. It is as if stronger binding is favored over weaker binding, even when the number of sites bound is the same. The idea of weak binding appears in the literature:

http://www.pnas.org/cgi/content/abstract/102/20/7097

My idea (actually a misinterpretation of Myriad's idea) was to replace the number of mistakes, which is very coarse, with the sum of the strength of the mistakes, which is a finer grained measure. The selection step can then distinguish between "weak" mistakes and "strong" mistakes. Actually, my version prefers creatures with many weak mistakes over ones with fewer but stronger mistakes.

The code fragment Unnamed posted appears to be weighting errors in the binding and non-binding site regions. What is the value assigned to the variables “gene” and “nongene”. If you set nongene=0 then you are ignoring errors in the nonbinding site region of the genome. That is equivalent to ignoring harmful mutations.

That modification that Unnamed appears to be changing much more than the threshold.

I don't use the weight variables "gene" and "nongene", and in any case I was running with them set to 1 all the time.

My modification was to replace "mistakes" with "sum of strength of mistakes". That was all. I didn't change what a mistake was, I didn't ignore spurious hits and I didn't ignore non-gene mistakes.

Now, I can't find the old results because the thread is huge. Does this version scale better than the original with genome size?

 

[kjkent1]

What kind of precise selection process that occur in nature would evolve a gene de novo? You must have drunk more than a few beers after celebrating that issue of agreement with hammegk in order to come up with this idea.

Sure ev shows that information can evolve in a genome. It just takes more time than the history of the earth to get a gene de novo.

Your argument is subtly "evolving" (or maybe you're just adding a little more argument and subtly moving the goalpost a tad).

Until very recently you were content to argue that ev was flawed because it could not evolve a gene fast enough. Now you seem to be suggesting that even if the selection mechanism were modify to make ev work fast enough, this wouldn't satisfy you, because your new position is that the gene must evolve without anything more than completely random selection.

Please explain why a a genome cannot have a selective advantage to an organism at a relatively simple stage of its evolution, which is substantially different from the selective advantage present at some relatively later and more complex point.

As you have already conceded, ev is very fast for short genomes, so there would be no barrier to a quick evolution without selection to some very simple selective advantage, followed by more environmentally channeled selection speeding the evolutionary process toward a more complex advantage.

 

[Unnamed]

Now that the 128K run is finished, I'll republish the results:

Here is my data so far:

population 32
binding sites 8
weight width 9
site width 10
1 mutation / 512 bases

genome size, generations

1024, 8000
2048, 20000
4096, 34000
8192, 37000
16384, 76000
32768, 272000
65536, 392000
128000, 991000

I just got this result:
65536, 32 mutations/generation (should be 128), 827 generations to perfect creature.

The difference is just too big, so I am posting a screenshot of the options dialog.

 

[Paul C. Anagnostopoulos]

That modification that Unnamed appears to be changing much more than the threshold.

He explained what he did:

What I did was, for each mistake, to add the distance between the valuation and threshold, and use the result as the sorting order. This is a finer measure than the raw number of mistakes, but it still goes to zero when mistakes = 0, so it converges to the same place.

~~ Paul

 

[Paul C. Anagnostopoulos]

I just got this result:
65536, 32 mutations/generation (should be 128), 827 generations to perfect creature.

Whoa, Nelly! Is that possible?

The difference is just too big, so I am posting a screenshot of the options dialog.

You have an old version of Evj that doesn't let you set the mutations per base. Can you grab the latest code and make your changes again, then run with the parameters I specified? We're now on version 2.38.

~~ Paul