Annoying creationists

[delphi_ote]

For you Delphi, any time. The use of multiple drugs (triple antiviral medications) to prevent selection and evolution of drug resistant strains of HIV is extensively documented in the guidelines for the treatment of this disease.

Thanks for the 121 page document you haven't read. Another smokescreen. Why am I not suprised?

Adding multiple drugs... each of which would cause selection... prevents selection? Are you really typing these things?

Also, I note that you've dodged away from your original position...

In fact, ev will not converge at all in many cases when all the selection processes are acting simultaneously. This is also observed in reality.

You said multiple selection pressures stop evolution, and that this has been observed in nature. You've repeated this fiction several times. That's obviously not the case here, as HIV and AIDS are still a problem. The only way to "stop" evolution is to put the entire population under too stringent a selection pressure (i.e. kill them all off.)

 

[Dr Adequate]

Lie #1.

Truth #1.

 

[Taffer]

Truth #1.

 

[kleinman]

Annoying Creationists

Why don’t you explain why ev stops converging as you lengthen the genome? And then when you eliminate two of the three selection pressures, ev then will converge with all other parameters held fixed.

Your first question relates to no part of my experiment or my conclusion, so your raising it here is irrelevant.

Once again, you miss the point. The reason why ev stops converging is that the multiple competing selection process stop evolution. Why don’t you perform some relevant experiments?

Re your second question, until you show me the specific evidence supporting your conclusion, I have no comment.

The evidence couldn’t be simpler. Choose any case you want and lengthen the genome until ev no longer converges with each of the weight factors set to 1. Then take any two of the three weight factors and set them to zero and behold, ev converges with a single selection condition. It is the multiple selection conditions that prevents ev from converging.

While you're at it, why don't you explain why my experimental results directly contradict your conclusion that multiple selective forces slow/halt evolution.

Your explanation is incomprehensible. Post your experimental results and I’ll comment on them.

http://aidsinfo.nih.gov/contentfiles/AdultandAdolescentGL.pdf

Did you actually read your source? If you did, then perhaps you can point it out to me where it says that multiple drugs are used to stop resistance developing in the HIV virus. Because I could not find it.

Take that paper and do a word search on the words “resistant” and “resistance” and you will find hundreds of occurrences of these words in this guideline. The main point of this guideline for the treatment of HIV is to prevent the emergence drug resistant strains. Here is a typical quote from the guidelines:

Simultaneously stopping all drugs in a regimen containing these agents may result in functional monotherapy with the NNRTIs, because their half-lives are longer than other agents. This may increase the risk of selection of NNRTI-resistant mutations.

Ev does stop converging. Paul is trying to attribute this to his Rcapacity factor but it is easily seen that the failure of ev to converge is due to the multiple selection factors. Read the link I provide to the guidelines for treatment of HIV and you will see the treatment strategy is identical to what ev is demonstrating. If you can ever describe a selection pressure to evolve a gene from the beginning, imagine what would happen with a multitude of selection pressures evolving genes from the beginning would have to overcome.

*Sigh* You obviously do not understand evolution in the slightest, or you would know this is complete bollocks. List, kleinman, this is very important. Selective pressures are not positive.

So let’s hear your convoluted logic of how genes arose.

And please demonstrate how what happens in the ev model translates to what happens in the real world? Even if ev shows the stopping of evolution, this does not mean it actually happens in real life. Since evolutioni is an observable fact it obviously does not stop, does it?

Be glad to do that for you Taffer. Now you other evolutionists don’t complain that I am repeating myself because Taffer has asked.

Ev has three selection conditions; they are the identification of binding sites where they should occur, the prevention of spurious binding sites in the gene region and the prevention of spurious binding sites outside of the gene. As one lengthens the genome length parameter in ev, the generations for convergence becomes greater and greater until ev fails to converge at all. If you take that case that fails to converge, and remove any two of the three selection pressures, ev will successfully evolve that sequence based on that selection condition which would not evolve when applying all three selection conditions. This is exactly analogous to the strategy used in the treatment of HIV. By using multiple drug regimens, you apply multiple selection pressures and reduce the ability of the virus to evolve. Monotherapy treatment markedly increases the likelihood of evolving resistant strains of the virus.

Your continued claim that there are no selection pressures to evolve a gene "from the beginning" is a false dichotomy. A functional gene need not be the lowest form of life.

Genes came from somewhere, explain to us how an original gene arose.

Lastly, please demonstrate that the failure in ev for convergence is due to multiple selective pressures, and not Rcap.

Take Dr Schneider’s baseline case and keep doubling the genome length until ev fails to converge to a perfect creature. Then go to the advanced features of ev and set any two of the three weight factors for the selection conditions to 0 and you will see that ev very rapidly converges to a perfect creature. Setting two of the three weight factors to 0 is equivalent to eliminating two of the three selection pressures. This is exactly analogous to what we see with the use of multi-drug regimens for the treatment of HIV. Single drug regimens (single selection pressures) quickly lead to resistant strains of the virus while multi-drug regimens (multiple selection pressures) markedly reduce the emergence of drug resistant strains of HIV.

Ev does accurately model mutation and selection.

It accurately enough models mutation and selection of a single binding site, with a 'perfect' creature already established. It does not accurately model evolution in the real world. Get over it.

Dr Schneider has very effectively captured the mathematics of multiple selection pressures. What his model demonstrates due to the multiple selection pressure accurately reflects what happens in reality. Now his selection conditions do not accurately model the evolution of a gene from the beginning but that is because there is no such selection condition.

It shows that multiple selection pressures slow and ultimately stop evolution...

No, it doesn't. You just claim that it does.

Run that example I’ve described above and give yourself a lesson in the mathematics of mutation and selection with single and multiple selection pressures.

...and that principle is used to treat HIV.

No, it isn't. Multiple antiretroviral agents are used because restance arrises. Using more then one does not slow the evolution of the resistance of any single antiretroviral agent, but rather prolongs the course of treatment before the HIV virus becomes resistant to all three. You need to understand, kleinman, that not all selective pressures are do or die.

Which do you think happens more quickly, evolving a strain of HIV resistant to three drugs when the drugs are administered serially or administering the drugs simultaneously?

Not every mutation in ev causes a creature to die, yet ev demonstrates that multiple selection pressures slow and ultimately stop evolution.

You have no evidence that evolution has stopped, so stop claiming that it has.

Ask Paul whether evolution stops in ev. Paul knows it stops; his problem is that he has concocted a convoluted explanation with his Rcapacity concept. How unusual that an evolutionist comes up with a concocted, convoluted explanation.

Ev is showing what happens mathematically with multiple selection pressures, which is evolution slows down and ultimately stops. Ev is not doing this as a matter of black-and-white. If you think that ev is doing this as a matter of black-and-white, correct the model and show us how multiple selection pressures work.

I don't have to show how multiple selective pressures work, because I have observed them working myself. Remember those bacteria I've created? They were multiple resistant bacteria. Three, to be exact. We spontaniously generated resistant bacteria from non-resistant bacteral cultures. Wow, evolution! Oh, but that's right, it's "mathematically impossible".

Tell us, what occurs more quickly, creating multi-drug resistant bacteria using one drug at a time or using all the drugs simultaneously?

Very few people have studied ev and what it shows due to its multiple selection pressures. This is why ev fails to converge and this is the reason triple antiviral drugs are used to treat HIV. I doubt any evolutionist ever considered what happens when multiple selection pressures occur simultaneously and how it affects the rate of evolution.

You are a pompous git. We already have models for multiple selective pressures!

Then why don’t you understand your own models? If you did, you would understand that multiple selection pressures slow down the evolutionary process. If you are so smart, explain to Dr Schneider and Paul your selection pressure models and fix what ev shows.

Now that ev is available and its behavior has been studied, it is becoming apparent why the theory of evolution is mathematically impossible. It is not just that there are no selection pressures that can evolve a gene from the beginning; it is also that multiple selection pressures slow and then stop evolution.

And anyone who disagrees with you is wrong. Because you are right. Not the people who actually understand the material. You. Right.

You are wrong if you think that multiple selection pressures does not slow down the evolutionary process. Why don’t you give us an example where multiple selection pressures which speed up evolution. I’ve already shown how ev demonstrates that multiple selection pressures slow and ultimately stop evolution and I have given a real example of this phenomena. So give us some math and a real example of your convoluted thinking.

Taffer, multiple antibacterial resistant bacteria occurs much more quickly when the bacteria are subjected to the antibiotics in a serial manner, one antibiotic at a time. If the bacteria are subject to the antibiotics in parallel, they are much less likely to evolve. So what are the selection pressures that evolved the hundreds of genes necessary for the simplest free living organism and how did this happen in parallel?

You do not understand at all. Selection for antibacterial resistance occurs at the organism level. Take three loci, all of which give a different antibacterial reistance. Put bacteria on media in series, and you will evolve triple-resistant bacteria. Put the bacteria on media which contains all three antibiotics, and you will evolve triple-resistant bacteria. I have done this myself. So your claims that multiple selection pressures stop evolution are stupid, ignorant, and above all, false. The only reason why it seems to be harder to evolve triple-resistant bacteria in "parallel", is because their resistance has to evolve at once. But the probability is the same. 1+1+1 is the same as 3X1.

That’s right, when you apply your selection pressures simultaneously; it slows the evolutionary process because you must get all three random mutations simultaneously. This is why evolution slows with multiple selection pressures. Ev demonstrates this and it is the reason why this model shows that evolution by random point mutation and natural selection is profoundly slow, so slow that the theory of evolution is mathematically impossible. You may be able evolve three loci to get drug resistant bacteria but you have no selection condition that would evolve the thousands of bases to make a gene, especially when other selection conditions would interfere with this process.

Paul, in case you haven’t noticed, ev can easily evolve binding sites despite your Rcapacity condition when you set the weight factors for spurious binding sites to 0.

Alan, you simply refuse to pay any attention, don't you? I'm going to say this one more time and then give up:

Rcapacity is the number of bits of information required to distinguish binding sites from the rest of the genome. If you are not distinguishing binding sites from the rest of the genome, Rcapacity is irrelevant.

Paul, you are in denial. Ev easily recognizes binding sites no matter how long the genome is. Set any two of the three selection conditions to 0 and you will evolve the binding condition specified by that third condition. It doesn’t matter whether ev is locating the desired binding sites or preventing the spurious binding sites. Ev can do each no matter how long the genome is. Ev just can’t evolve all three selection conditions simultaneously when the genome gets too long. It is the competing selection pressures that stop evolution in ev. I listen to your explanation at it is convoluted and confused.

Oh really, why? Paul, ev can easily evolve all three of the selection conditions when done separately on a genome where your Rcapacity value equals Rfrequency. The failure of ev to converge when using all three selection conditions simultaneously is due to the competition of these conditions. It has nothing to do with being able to identify the binding sites. Binding sites are easily identified. This is shown by setting the weight factors for spurious bindings to zero.

Binding sites are easily identified because the program knows where they are. Locating the binding sites is not interesting. What is interesting is distinguishing binding sites from other sites using the transcription factor modeled by the weight matrix and threshold.

And ev can’t do all three selection conditions when the genome becomes too long.

Not only does setting two of the three selection conditions to zero affect the number of generations to evolve a perfect creature, it is the only way to evolve a perfect creature when you lengthen the genome in ev.

You do not know this, because you have never run an experiment where evolution "stops dead" when Rcapacity is not an issue. And apparently you never will, because you do not believe that the width of the binding sites puts a limit on Rsequence. I'm not sure I've ever had a conversation with someone who installed a belief system on top of mathematics.

Yes I have Paul, I’ll post the series again here.
The case I used consists of G=16384, binding sites=6, weight width=5, Rfreq=10, Paul’s infamous Rcapacity=2*site width=10. With the weights for each mistake, (missed site, spurious binding in gene, spurious binding outside gene), set to 1, this case fails to converge. But for varying weights you get the following results:
missed site/spurious binding-gene/spurious binding outside gene/gens to perfect creature
1/0/0/7
0/1/0/223
0/0/1/233
So, a case with three different selection conditions will not converge when all three conditions must be satisfied simultaneously. While if you select based on one condition at a time it takes only 7 generations to evolve the binding sites, 223 generations to eliminate spurious binding within the gene and only 233 generations to eliminate spurious binding outside the gene. These cases take so few generations to evolve, it is instructive to step through the evolution and watch the mistakes go to zero.
Choose any case you want that doesn’t converge and try the same concept. Set any two of the three selection conditions to 0 and watch ev evolve the remaining condition.

While you're at it, why don't you explain why my experimental results directly contradict your conclusion that multiple selective forces slow/halt evolution.

Because your genome is not large enough, obviously! Increase its size so that Rfrequency > Rsequence and try again. You silly goose.

Paul, any length genome will evolve in ev if you use only one of the three selection conditions.

For you Delphi, any time. The use of multiple drugs (triple antiviral medications) to prevent selection and evolution of drug resistant strains of HIV is extensively documented in the guidelines for the treatment of this disease.

Thanks for the 121 page document you haven't read. Another smokescreen. Why am I not suprised?

Adding multiple drugs... each of which would cause selection... prevents selection? Are you really typing these things?

The whole guide is about the prevention of the evolution of drug resistant strains of HIV. I’ll again post a quote from the guide:

Simultaneously stopping all drugs in a regimen containing these agents may result in functional monotherapy with the NNRTIs, because their half-lives are longer than other agents. This may increase the risk of selection of NNRTI-resistant mutations.

I guess you were the one who didn’t read the guide.
Also, I note that you've dodged away from your original position...

Also, I note that you've dodged away from your original position...

Are you whining about moving goal posts again? My position has always been that ev shows that the theory of evolution is mathematically impossible. The reason why ev shows this is that the competing selection conditions slow and ultimately stop evolution.

In fact, ev will not converge at all in many cases when all the selection processes are acting simultaneously. This is also observed in reality.

You said multiple selection pressures stop evolution, and that this has been observed in nature. You've repeated this fiction several times. That's obviously not the case here, as HIV and AIDS are still a problem. The only way to "stop" evolution is to put the entire population under too stringent a selection pressure (i.e. kill them all off.)

What I said is that ev shows that with increasing genome lengths, the multiple selection pressures slow and then ultimately stop the evolutionary process. Multiple selection pressures on the HIV virus slow the evolution of resistant strains sufficiently to allow people to live for years with the virus.

 

[Paul C. Anagnostopoulos]

Paul, you are in denial. Ev easily recognizes binding sites no matter how long the genome is. Set any two of the three selection conditions to 0 and you will evolve the binding condition specified by that third condition. It doesn’t matter whether ev is locating the desired binding sites or preventing the spurious binding sites. Ev can do each no matter how long the genome is. Ev just can’t evolve all three selection conditions simultaneously when the genome gets too long. It is the competing selection pressures that stop evolution in ev. I listen to your explanation at it is convoluted and confused.

You are the poster child for denial, Alan. Ev recognizes binding sites no matter what parameters you specify, because it knows precisely where they are. But if you set any of the mistake counts to zero, then the gene does not distinguish the binding sites from the rest of the genome. Do you see the difference? Please answer yes or no.

And ev can’t do all three selection conditions when the genome becomes too long.

How long would that be?

Yes I have Paul, I’ll post the series again here.
The case I used consists of G=16384, binding sites=6, weight width=5, Rfreq=10, Paul’s infamous Rcapacity=2*site width=10. With the weights for each mistake, (missed site, spurious binding in gene, spurious binding outside gene), set to 1, this case fails to converge. But for varying weights you get the following results:
missed site/spurious binding-gene/spurious binding outside gene/gens to perfect creature
1/0/0/7
0/1/0/223
0/0/1/233

But obviously I was not talking about the scenario where any of the mistake counts are zero, so your answer is irrelevant. Please run an experiment with mistake counts of 1, where Rcapacity is not an issue, and evolution "stops dead."

~~ Paul

 

[kjkent1]

***Deleted***

I'll leave Paul's comments undisturbed. Alan doesn't seem to comprehend that setting mistake weights to zero yields meaningless results.

 

[kleinman]

Annoying Creationists

Paul, you are in denial. Ev easily recognizes binding sites no matter how long the genome is. Set any two of the three selection conditions to 0 and you will evolve the binding condition specified by that third condition. It doesn’t matter whether ev is locating the desired binding sites or preventing the spurious binding sites. Ev can do each no matter how long the genome is. Ev just can’t evolve all three selection conditions simultaneously when the genome gets too long. It is the competing selection pressures that stop evolution in ev. I listen to your explanation at it is convoluted and confused.

You are the poster child for denial, Alan. Ev recognizes binding sites no matter parameters you specify, because it know precisely where they are. But if you set any of the mistake counts to zero, then the gene does not distinguish the binding sites from the rest of the genome. Do you see the difference? Please answer yes or no.

Paul, I’m not the one whose evaluation of ev has gone from it represents reality, to it represents a tiny fraction of the evolutionary landscape to it represents a stylized model of mutation and natural selection.

Paul, you have just changed you position on what ev does. I have said that the weight matrix always recognizes a binding site no matter how long the genome. When you set any of the weight counts to zero, all you are doing is not including that selection condition in the evolutionary process. Paul, there is no gene in ev, there is just a region of the genome that you set aside to evolve binding sites. If you set the weight condition for missed sites to 1, and spurious sites inside and outside the binding site region to 0, you will evolve the binding sites where they should be according to the model but still have spurious bindings elsewhere on the genome. Here is another example: G=100,000, binding site width=4, weight width is 3, Rfrequency is 13.61, Rcapacity=8, gamma=8, mutation rate=1 per genome per generation.
Missed sites/spurious gene/spurious non-gene/Gens convergence PC
1/0/0/1
0/1/0/54
0/0/1/2192
Here is a case where Rfrequency is much greater than Rcapacity. Ev has no problem locating the binding sites where they should be when that weight factor is set to 1, ev has no problem eliminating spurious binding sites in the gene region when that weight condition is set to 1, and ev has no problem eliminating spurious binding sites outside the gene region when that weight factor is set to 1. This is all done in a case with Rfrequency is almost twice the value of your Rcapacity value. The ev computation cycles through a mutation/selection process. When you have too many selection conditions on too long of a gene, evolution stops. Eliminate 2 of the 3 selection conditions and ev starts evolving that portion of the genome.

When you say “gene does not distinguish the binding sites from the rest of the genome”, what are you talking about? If you are talking about the weight matrix which represents the binding protein, the above example once again shows that the weight matrix has no trouble locating binding sites whether they are missed sites where they should have occurred or spurious sites where they should not occur no matter how large the genome is.

And ev can’t do all three selection conditions when the genome becomes too long.

How long would that be?

I just did an example above with a genome length of 100,000 and a binding site width of only 4 bases. Even the elimination of the spurious binding sites in the non-binding site region of the genome took only a couple of thousand cycles. I believe you can do a genome length of 3 gigabases and get convergence if you only have a single selection condition, but more than one selection condition and you can only evolve tiny nonrealistic length genomes.

Yes I have Paul, I’ll post the series again here.
The case I used consists of G=16384, binding sites=6, weight width=5, Rfreq=10, Paul’s infamous Rcapacity=2*site width=10. With the weights for each mistake, (missed site, spurious binding in gene, spurious binding outside gene), set to 1, this case fails to converge. But for varying weights you get the following results:
missed site/spurious binding-gene/spurious binding outside gene/gens to perfect creature
1/0/0/7
0/1/0/223
0/0/1/233

But obviously I was not talking about the scenario where any of the mistake counts are zero, so your answer is irrelevant. Please run an experiment with mistake counts of 1, where Rcapacity is not an issue, and evolution "stops dead."

Paul, it is not Rcapacity which stops ev from converging, it is the conflicting selection processes that stops evolution. This mathematical effect that ev is demonstrating is the exact principle that is used for treating HIV. This is totally relevant because this is the reason why ev takes huge numbers of generations to converge (if it can converge at all) when you try to evolve binding sites and eliminate spurious binding sites on longer genomes. When your model has this much trouble evolving binding sites and eliminating spurious binding, what do you think will happen when you try to model the evolution of an entire gene (if you could ever model a selection process that would do this) when you have “millions” of other selection processes acting at the same time?

Your Rcapacity concept is useless for estimating which cases will converge unless you include all three selection conditions in ev.

I'll leave Paul's comments undisturbed. Alan doesn't seem to comprehend that setting mistake weights to zero yields meaningless results.

You know more about your string cheese theory of evolution than you do about ev. Setting a weight to zero is simply eliminating that particular selection process.

 

[Paul C. Anagnostopoulos]

Paul, I’m not the one whose evaluation of ev has gone from it represents reality, to it represents a tiny fraction of the evolutionary landscape to it represents a stylized model of mutation and natural selection.

That is neither a yes nor a no.

Paul, you have just changed you position on what ev does. I have said that the weight matrix always recognizes a binding site no matter how long the genome.

If by "recognizes" you mean matches, that is not correct.

When you set any of the weight counts to zero, all you are doing is not including that selection condition in the evolutionary process.

I presume you mean mistake counts. That is correct, setting a mistake count to zero removes that selection pressure, rendering Rcapacity et al moot.

Paul, there is no gene in ev, there is just a region of the genome that you set aside to evolve binding sites.

The weight matrix and threshold are referred to as the gene.

If you set the weight condition for missed sites to 1, and spurious sites inside and outside the binding site region to 0, you will evolve the binding sites where they should be according to the model but still have spurious bindings elsewhere on the genome.

I presume you mean mistake count. You are correct.

Here is another example: G=100,000, binding site width=4, weight width is 3, Rfrequency is 13.61, Rcapacity=8, gamma=8, mutation rate=1 per genome per generation.
Missed sites/spurious gene/spurious non-gene/Gens convergence PC
1/0/0/1
0/1/0/54
0/0/1/2192
Here is a case where Rfrequency is much greater than Rcapacity.

Rfrequency and Rcapacity are irrelevant here. They only matter when attempting to evolve creatures who distinguish binding sites from other sites, that is, when all three mistake counts are positive. Let me try to make this simple:

Measures of information capacity, requirement, and current state are only relevant when distinguishing binding sites from other sites. If you are not trying to distinguish them, then the amount of information required to do so is irrelevant.

Ev has no problem locating the binding sites where they should be when that weight factor is set to 1, ev has no problem eliminating spurious binding sites in the gene region when that weight condition is set to 1, and ev has no problem eliminating spurious binding sites outside the gene region when that weight factor is set to 1.

Correct.

This is all done in a case with Rfrequency is almost twice the value of your Rcapacity value.

See above.

The ev computation cycles through a mutation/selection process. When you have too many selection conditions on too long of a gene, evolution stops.

You have not demonstrated this.

Eliminate 2 of the 3 selection conditions and ev starts evolving that portion of the genome.

Correct.

When you say “gene does not distinguish the binding sites from the rest of the genome”, what are you talking about?

See above.

If you are talking about the weight matrix which represents the binding protein, the above example once again shows that the weight matrix has no trouble locating binding sites whether they are missed sites where they should have occurred or spurious sites where they should not occur no matter how large the genome is.

That is not what I am talking about.

And ev can’t do all three selection conditions when the genome becomes too long.
...
I just did an example above with a genome length of 100,000 and a binding site width of only 4 bases.

But you did it with two of the selection pressures turned off, so your experiment has nothing to do with the first statement you made here.

Even the elimination of the spurious binding sites in the non-binding site region of the genome took only a couple of thousand cycles. I believe you can do a genome length of 3 gigabases and get convergence if you only have a single selection condition, but more than one selection condition and you can only evolve tiny nonrealistic length genomes.

Please demonstrate this claim.

Paul, it is not Rcapacity which stops ev from converging, it is the conflicting selection processes that stops evolution.

Please demonstrate this.

Your Rcapacity concept is useless for estimating which cases will converge unless you include all three selection conditions in ev.

No kidding, Sherlock. That is because Rcapacity is completely irrelevant unless all three selection pressures are present.

You know more about your string cheese theory of evolution than you do about ev. Setting a weight to zero is simply eliminating that particular selection process.

Which renders the information measures meaningless, as Kjkent said.

~~ Paul

 

[kleinman]

Annoying Creationists

Paul, I’m not the one whose evaluation of ev has gone from it represents reality, to it represents a tiny fraction of the evolutionary landscape to it represents a stylized model of mutation and natural selection.

That is neither a yes nor a no.

The answer to your question is the gene distinguishes nothing and your Rcapacity concept is hogwash. The reason why ev does not converge with longer genomes is that the competing selection processes in the model stop the evolution. Eliminate 2 of the three selection conditions and ev will start evolving again.

Paul, you have just changed you position on what ev does. I have said that the weight matrix always recognizes a binding site no matter how long the genome.

If by "recognizes" you mean matches, that is not correct.

Exceeds the threshold Mr Rcapacity. The weight matrix can do this no matter how long the genome is.

When you set any of the weight counts to zero, all you are doing is not including that selection condition in the evolutionary process.

I presume you mean mistake counts. That is correct, setting a mistake count to zero removes that selection pressure, rendering Rcapacity et al moot.

Are you paying attention kjkent1?

Paul, there is no gene in ev, there is just a region of the genome that you set aside to evolve binding sites.

The weight matrix and threshold are referred to as the gene.

And that weight matrix has no trouble locating binding sites which exceed the threshold no matter how long the genome is. The reason ev fails to converge is conflicting selection conditions.

If you set the weight condition for missed sites to 1, and spurious sites inside and outside the binding site region to 0, you will evolve the binding sites where they should be according to the model but still have spurious bindings elsewhere on the genome.

I presume you mean mistake count. You are correct.

Again, are you paying attention kjkent1?

Here is another example: G=100,000, binding site width=4, weight width is 3, Rfrequency is 13.61, Rcapacity=8, gamma=8, mutation rate=1 per genome per generation.
Missed sites/spurious gene/spurious non-gene/Gens convergence PC
1/0/0/1
0/1/0/54
0/0/1/2192
Here is a case where Rfrequency is much greater than Rcapacity.

Rfrequency and Rcapacity are irrelevant here. They only matter when attempting to evolve creatures who distinguish binding sites from other sites, that is, when all three mistake counts are positive. Let me try to make this simple:

Measures of information capacity, requirement, and current state are only relevant when distinguishing binding sites from other sites. If you are not trying to distinguish them, then the amount of information required to do so is irrelevant.

Paul, Dr Schneider’s definition for Rfrequency is:
Rfrequency = -log2(gamma/G)
Rfrequency depends only on the number of binding sites and the length of the genome. The selection process(es) does not change the computed value of Rfrequency. Your definition for Rcapacity is:
Rcapacity = 2*binding site width
Your definition for Rcapacity has no relationship to selection process(es). It is an interesting coincidence that your equation gives a fairly good estimate when ev will stop converging when including all three selection conditions in ev but your equation is useless for estimating when ev will fail to converge for different selection conditions. In particular, if you have only a single selection condition, ev will evolve sequences of bases that will satisfy that condition no matter how long the genome is. It is the multiple selection conditions that stop ev from converging.

Ev has no problem locating the binding sites where they should be when that weight factor is set to 1, ev has no problem eliminating spurious binding sites in the gene region when that weight condition is set to 1, and ev has no problem eliminating spurious binding sites outside the gene region when that weight factor is set to 1.

Correct.

You paying attention kjkent1?

This is all done in a case with Rfrequency is almost twice the value of your Rcapacity value.

See above.

Paul, your Rcapacity value is meaningless. See above, see the results from ev, see the reality revealed by the treatment of HIV using multiple drugs (multiple selection pressures).

The ev computation cycles through a mutation/selection process. When you have too many selection conditions on too long of a gene, evolution stops.

You have not demonstrated this.

Paul, I have already done two cases, one with a G > 16,000 and another with G = 100,000 and both with Rfrequency > Rcapacity which won’t converge with all three of ev’s selection conditions imposed but will converge for all three conditions if done one at a time. If you want, I will start doing hundreds of cases to further demonstrate this.

Eliminate 2 of the 3 selection conditions and ev starts evolving that portion of the genome.

Correct.

Paying attention kjkent1?

When you say “gene does not distinguish the binding sites from the rest of the genome”, what are you talking about?

See above.

We know that the weight matrix easily finds matches which exceeds the threshold no matter how long the genome is.

If you are talking about the weight matrix which represents the binding protein, the above example once again shows that the weight matrix has no trouble locating binding sites whether they are missed sites where they should have occurred or spurious sites where they should not occur no matter how large the genome is.

That is not what I am talking about.

Paul, I looked above, and the weight matrix has no trouble locating sites which exceed the threshold no matter how long the gene is.

And ev can’t do all three selection conditions when the genome becomes too long.
...
I just did an example above with a genome length of 100,000 and a binding site width of only 4 bases.

But you did it with two of the selection pressures turned off, so your experiment has nothing to do with the first statement you made here.

Paul, that’s the point. Ev fails to converge with all three selection conditions imposed in the model. When you turn off two of the three selection conditions, ev will evolve one selection condition at a time without any problem. Multiple selection conditions slow and ultimately stop the evolutionary process in ev. What ev is demonstrating here is a real phenomena. I’ll go around with you on this issue as many times as you want because Dr Schneider got this effect right.

Even the elimination of the spurious binding sites in the non-binding site region of the genome took only a couple of thousand cycles. I believe you can do a genome length of 3 gigabases and get convergence if you only have a single selection condition, but more than one selection condition and you can only evolve tiny nonrealistic length genomes.

Please demonstrate this claim.

You want me to do the 3 gigabase case? Hey kjkent1, you ready to get out of your cubicle and get your company’s parallel processor system crunching on this case. We’ll get that one done. Hey, what is the Rcapacity value for this case?

Paul, it is not Rcapacity which stops ev from converging, it is the conflicting selection processes that stops evolution.

Please demonstrate this.

Paul, I’ve already demonstrated this on a 16k genome and a 100k genome. I’ll start posting hundreds of cases that demonstrate this. This thread is going to be longer than an ev run with Rfrequency greater than Rcapacity and all three selection conditions turned on.

Your Rcapacity concept is useless for estimating which cases will converge unless you include all three selection conditions in ev.

No kidding, Sherlock. That is because Rcapacity is completely irrelevant unless all three selection pressures are present.

Your Rcapacity definition is also useless with Unnamed’s selection process.

You know more about your string cheese theory of evolution than you do about ev. Setting a weight to zero is simply eliminating that particular selection process.

Which renders the information measures meaningless, as Kjkent said.

Kjkent1 is now your expert on ev? He’s also an expert on the string cheese theory of evolution.

Paul, I really like Dr Schneider’s computer model and you have done a wonderful job on programming the online version of the model. It is interesting studying the mathematics of mutation and selection.

 

[delphi_ote]

Simultaneously stopping all drugs in a regimen containing these agents may result in functional monotherapy with the NNRTIs, because their half-lives are longer than other agents. This may increase the risk of selection of NNRTI-resistant mutations.

This is evidence that if you ease overall selection pressure, the probability of mutants adapting to the remaining selection pressure increases. If more of the virus can survive, obviously there's a greater danger that one of them will develop resistance.

This is not evidence that multiple selection pressures "stop evolution". It has nothing to do with the number of selection mechanisms. It's simply a matter of the population size! Multiple selection pressures do not make evolution impossible. As I stated before (and you compeltely ignored) if that were the case, HIV and AIDS would be cured for good by these drugs. Two of them would be enough to guarantee that HIV never developed resistance. Simply carrying through your own logic on your own example shows you're wrong. Christ eating babyback ribs, kleinman. THINK FOR A CHANGE!

 

[kjkent1]

Kjkent1 is now your expert on ev? He’s also an expert on the string cheese theory of evolution.

Evidently, I understand ev a hell of a lot better than you do.

If a "perfectly ordered" deck of cards is 1 through 13, hearts, clubs diamonds and spades, and you try to sort the deck without considering numerical order, how long do you think it will take to "converge" on, or produce a "perfectly ordered" deck?

And if you don't consider suits, how long do you think it will take to sort it to perfection?

The answer is that it will take less time to sort for either of the above conditions, because you're ignoring the other condition.

In the first case you will have four suits with random numbers. In the second you will have ordered numbers and random suits.

So, the RESULT of your sort is crap, because in neither circumstance do you produce a perfectly ordered deck.

By turning off the mistake weights, your experiment is not evolving anything close to a perfect creature. Ev is, however, reporting a perfect creature, because it doesn't recognize that it can't sort correctly due to the zeroed mistake weight.

 

[Paul C. Anagnostopoulos]

The answer to your question is the gene distinguishes nothing and your Rcapacity concept is hogwash.

So once a perfect creature has evolved, the gene is not distinguishing the binding sites from the rest of the genome? Wow, that's interesting. That means that Ev does absolutely nothing.

The reason why ev does not converge with longer genomes is that the competing selection processes in the model stop the evolution.

Please demonstrate this.

Exceeds the threshold Mr Rcapacity. The weight matrix can do this no matter how long the genome is.

But you said "the weight matrix always recognizes a binding site." That is not true until a perfect creature evolves.

Are you paying attention kjkent1?

Wait, are you agreeing with my entire statement: "That is correct, setting a mistake count to zero removes that selection pressure, rendering Rcapacity et al moot."

And that weight matrix has no trouble locating binding sites which exceed the threshold no matter how long the genome is. The reason ev fails to converge is conflicting selection conditions.

Please demonstrate this.

Paul, Dr Schneider’s definition for Rfrequency is:
Rfrequency = -log2(gamma/G)
Rfrequency depends only on the number of binding sites and the length of the genome. The selection process(es) does not change the computed value of Rfrequency. Your definition for Rcapacity is:
Rcapacity = 2*binding site width
Your definition for Rcapacity has no relationship to selection process(es).

It has a relationship. You mean the math does not refer to the selection process. But that does not mean that Rfrequency and Rcapacity are relevant to all those selection combinations, does it? In fact, I have on my to-do list an item to flag the R values when they are not relevant to the chosen selection process.

It is an interesting coincidence that your equation gives a fairly good estimate when ev will stop converging when including all three selection conditions in ev but your equation is useless for estimating when ev will fail to converge for different selection conditions.

Well, at least you're beginning to understand something. Now, if Rcapacity has nothing to do with the information capacity of binding sites, what do you think would happen if a binding site width of 1 is specified? Do you think this could allow a gene that distinguishes the binding sites from all other sites?

Paul, I have already done two cases, one with a G > 16,000 and another with G = 100,000 and both with Rfrequency > Rcapacity which won’t converge with all three of ev’s selection conditions imposed but will converge for all three conditions if done one at a time. If you want, I will start doing hundreds of cases to further demonstrate this.

But that is not your claim, is it? Your claim is that evolution stops when all three mistake counts are positive and Rcapacity is not an issue. Please demonstrate this.

You want me to do the 3 gigabase case? Hey kjkent1, you ready to get out of your cubicle and get your company’s parallel processor system crunching on this case. We’ll get that one done. Hey, what is the Rcapacity value for this case?

Thank you for acknowledging that you have not demonstrated your claim.

Paul, I’ve already demonstrated this on a 16k genome and a 100k genome.

You're absolutely sure that evolution had stopped in those experiments and that Rcapacity was not an issue? If so, present the parameters.

Your Rcapacity definition is also useless with Unnamed’s selection process.

Thank you for acknowledging that you were previously applying Rcapacity to irrelevant cases. However, Rcapacity is relevant to Unnamed's selection process.

~~ Paul

 

[kleinman]

Annoying Creationists

Kjkent1 is now your expert on ev? He’s also an expert on the string cheese theory of evolution.

Evidently, I understand ev a hell of a lot better than you do.

You do understand the string cheese theory of evolution far better than I do.

By turning off the mistake weights, your experiment is not evolving anything close to a perfect creature. Ev is, however, reporting a perfect creature, because it doesn't recognize that it can't sort correctly due to the zeroed mistake weight.

Kjkent1, you didn’t pay attention to Paul’s answers.

Anyway, let’s run that 3 gigabase genome case with a single selection condition on your company’s computer. Maybe then you and Paul will understand that it is the competing selection conditions that slows then ultimately stops the evolutionary process.

 

[Paul C. Anagnostopoulos]

By turning off the mistake weights, your experiment is not evolving anything close to a perfect creature. Ev is, however, reporting a perfect creature, because it doesn't recognize that it can't sort correctly due to the zeroed mistake weight.

Indeed, and so now on my to-do list are items to clarify these issues in the user interface.

~~ Paul

 

[Paul C. Anagnostopoulos]

Kjkent1, you didn’t pay attention to Paul’s answers.

Why do you think so?

~~ Paul

 

[kjkent1]

You do understand the string cheese theory of evolution far better than I do.

A little more humility and the ability to learn from your mistakes would serve you well.

 

[kleinman]

Annoying Creationists

The answer to your question is the gene distinguishes nothing and your Rcapacity concept is hogwash.

So once a perfect creature has evolved, the gene is not distinguishing the binding sites from the rest of the genome? Wow, that's interesting. That means that Ev does absolutely nothing.

When you have one selection condition, ev evolves for that one condition, when you have two or more conditions, ev will evolve for those selection conditions if the genome is short enough. What is interesting is that you cling on to this Rcapacity concept when you admit it only has application to Dr Schneider’s selection process. Why don’t you define Rcapacity for Unnamed’s selection process.

The reason why ev does not converge with longer genomes is that the competing selection processes in the model stop the evolution.

Please demonstrate this.

Already have done this for you twice.

Exceeds the threshold Mr Rcapacity. The weight matrix can do this no matter how long the genome is.

But you said "the weight matrix always recognizes a binding site." That is not true until a perfect creature evolves.

The weight matrix finds many matches for binding sites which exceed the threshold long before the perfect creature evolves. You call them spurious binding sites.

Are you paying attention kjkent1?

Wait, are you agreeing with my entire statement" "That is correct, setting a mistake count to zero removes that selection pressure, rendering Rcapacity et al moot."

Paul, your observation that with Dr Schneider’s selection conditions lead to a point where ev stops converging when Rfrequency approaches 2*binding site width is an interesting coincidence but does not explain why ev stops converging. I doubt you will find any simple algebraic expression which allows you to estimate when ev will converge based on the number of selection conditions. If you modified ev to evolve two sets of binding sites rather than the one set as done now, I doubt you could apply your Rcapacity equation to that situation.

And that weight matrix has no trouble locating binding sites which exceed the threshold no matter how long the genome is. The reason ev fails to converge is conflicting selection conditions.

Please demonstrate this.

I’ve already done two cases that demonstrate this. I’m checking with kjkent1 to see if we can use his company’s computer to do the 3 gigabase case.

Paul, Dr Schneider’s definition for Rfrequency is:
Rfrequency = -log2(gamma/G)
Rfrequency depends only on the number of binding sites and the length of the genome. The selection process(es) does not change the computed value of Rfrequency. Your definition for Rcapacity is:
Rcapacity = 2*binding site width
Your definition for Rcapacity has no relationship to selection process(es).

Correct, but that does not mean that Rfrequency is relevant to all those selection combinations, does it? In fact, I have on my to-do list an item to flag the R values when they are not relevant to the chosen selection process.

Paul, Rfrequency is dependent on gamma and G. gamma and G are independent of the selection process. Unless Dr Schneider improperly defined Rfrequency, this value is independent of the selection process.

It is an interesting coincidence that your equation gives a fairly good estimate when ev will stop converging when including all three selection conditions in ev but your equation is useless for estimating when ev will fail to converge for different selection conditions.

Well, at least you're beginning to understand something. Now, if Rcapacity has nothing to do with the information capacity of binding sites, what do you think would happen if a binding site width of 1 is specified? Do you think this could allow a gene that distinguishes the binding sites from all other sites?

A binding site width of 1 means that particular base will always be recognized as a binding site. What ev should do is evolve so that all binding sites will be of that base in the binding site region and eliminate that base from causing spurious binding inside and outside the gene.

Paul, I have already done two cases, one with a G > 16,000 and another with G = 100,000 and both with Rfrequency > Rcapacity which won’t converge with all three of ev’s selection conditions imposed but will converge for all three conditions if done one at a time. If you want, I will start doing hundreds of cases to further demonstrate this.

But that is not your claim, is it? Your claim is that evolution stops when all three mistake counts are positive and Rcapacity is not an issue. Please demonstrate this.

Paul, that’s what these examples do. These examples do not evolve when all three selection conditions are turned on yet when you set two of the three to 0, the non-zero condition evolves sequences that satisfy the remaining selection condition.

You want me to do the 3 gigabase case? Hey kjkent1, you ready to get out of your cubicle and get your company’s parallel processor system crunching on this case. We’ll get that one done. Hey, what is the Rcapacity value for this case?

Thank you for acknowledging that you have not demonstrated your claim.

Paul, I never said I demonstrated the 3 gigabase case. What I said is that I believe that with a single selection condition, the 3 gigabase case will converge. Now if we can get kjkent1 to get his company’s computer in action, I could show the results for this case. As it stands, we’ll have to live with the results of the 100k case which converges for all three selection conditions in less than three thousand cycles.

Paul, I’ve already demonstrated this on a 16k genome and a 100k genome.

You're absolutely sure that evolution had stoppped in those experiments and that Rcapacity was not an issue?

The 16k case was one from a series I did last year and it did not converge. The 100k case, I chose parameters much worse than many of the other cases which failed to converge. I chose a small site width of 4 and a larger genome length than many of the other cases which failed to converge. Feel free to burn up some clock cycles on your computer on this 100k case and prove me wrong and that this case will converge with all three selection conditions acting.

Your Rcapacity definition is also useless with Unnamed’s selection process.

Thank you for acknowledging that you were previously applying Rcapacity to irrelevant cases. However, Rcapacity is relevant to Unnamed's selection process.

Just what good is your Rcapacity concept? Let’s see you apply it to Unnamed’s selection process.

Kjkent1, you didn’t pay attention to Paul’s answers.

Why do you think so?

Because he doesn’t realize that setting a mistake weight to 0 is equivalent to turning off that particular selection condition

You do understand the string cheese theory of evolution far better than I do.

A little more humility and the ability to learn from your mistakes would serve you well.

Evidently, I understand ev a hell of a lot better than you do.

 

[kjkent1]

Because he doesn’t realize that setting a mistake weight to 0 is equivalent to turning off that particular selection condition.

What you don't recognize is that an ev "perfect creature" is defined as a genome substantially absent missed or spurious bindings. So, when you turn off a selection condition, ev reports a perfect creature when no perfect creature has yet evovled according to the ev definition of perfection. Thus, the report is an error -- as is your conclusion based on the error.

 

[Taffer]

kleinman, I can honestly not be bothered wading through your response to me anymore. Either clean up your formatting, or I'm not going to bother responding to specific points. Perhaps later, when I'm not in "a mood".

One specific point, though. The time to create a triple-resistant mutant is the same in both cases you provided. If you disagree, provide evidence and citations. Until you do, I will trust my own studies, thank you.

 

[Paul C. Anagnostopoulos]

When you have one selection condition, ev evolves for that one condition, when you have two or more conditions, ev will evolve for those selection conditions if the genome is short enough. What is interesting is that you cling on to this Rcapacity concept when you admit it only has application to Dr Schneider’s selection process. Why don’t you define Rcapacity for Unnamed’s selection process.

I did not say it was relevant only to Ev's selection process. I said it was only relevant when the width of the binding sites are fixed, as they are in Ev. Rcapacity works the same with Unnamed's selection process, assuming that the mistake counts are all positive.

The weight matrix finds many matches for binding sites which exceed the threshold long before the perfect creature evolves. You call them spurious binding sites.

Ah, okay. You called them "binding sites," which confused me.

Paul, your observation that with Dr Schneider’s selection conditions lead to a point where ev stops converging when Rfrequency approaches 2*binding site width is an interesting coincidence but does not explain why ev stops converging.

It explains why it has stopped converging in all the experiments I've seen. You claim to have two experiments that stop converging even when Rcapacity is not an issue. Please present the parameters of those experiments.

I’ve already done two cases that demonstrate this. I’m checking with kjkent1 to see if we can use his company’s computer to do the 3 gigabase case.

Please present the parameters of those experiments.

Paul, Rfrequency is dependent on gamma and G. gamma and G are independent of the selection process. Unless Dr Schneider improperly defined Rfrequency, this value is independent of the selection process.

Alan, what is Rfrequency? Could you define it for us?

A binding site width of 1 means that particular base will always be recognized as a binding site. What ev should do is evolve so that all binding sites will be of that base in the binding site region and eliminate that base from causing spurious binding inside and outside the gene.

And so as Rfrequency climbs above Rcapacity (2 in this case), the number of bits required to distinguish binding sites from other sites exceeds the available number of bits in the binding sites. This happens quickly when Rcapacity = 2.

Paul, that’s what these examples do. These examples do not evolve when all three selection conditions are turned on yet when you set two of the three to 0, the non-zero condition evolves sequences that satisfy the remaining selection condition.

Which examples? Please present the parameters.

Paul, I never said I demonstrated the 3 gigabase case.

I know you haven't. What I want to see are the two examples you claim "stopped."

The 16k case was one from a series I did last year and it did not converge. The 100k case, I chose parameters much worse than many of the other cases which failed to converge. I chose a small site width of 4 and a larger genome length than many of the other cases which failed to converge. Feel free to burn up some clock cycles on your computer on this 100k case and prove me wrong and that this case will converge with all three selection conditions acting.

I can't burn any cycles until you specify the parameters.

Just what good is your Rcapacity concept? Let’s see you apply it to Unnamed’s selection process.

It isn't any good! It just explains why certain Ev experiments will not converge. You claim evolution stops for a different reason. I just want you to demonstrate your claim. Is this really so hard to understand?

Because he doesn’t realize that setting a mistake weight to 0 is equivalent to turning off that particular selection condition.

He doesn't?

~~ Paul