Annoying creationists

[rocketdodger]

OK I updated my simulation program and made it more realistic and quite a bit faster. The source code is at www.jedi-arts.com/code/jev.cpp and an example input file is at www.jedi-arts.com/code/data.txt .

The program is quite simple. It takes a population of p creatures, each with a genome of bases of size g, and exerts P pressures on it. Initially the genome is initialzed so that the targeted bases of each pressure are neutral -- the pressures do not infer any fitness change to the population at the start of generation 1.

Each generation, m mutations randomly occur in each genome. During selection, the targeted bases of the pressures are examined. If the base is the same as the pressures "match" field, f fitness is added to the creature. If the base is the same as the pressures "badMatch" field, f * badWeight fitness is subtracted from the creature. This is meant to implement the idea that a bad mutation would decrease fitness (I.E. likely kill the creature) more than a good mutation would increase fitness.

Selection is performed by summing up the fitness of the entire population, then allocating offspring according to the percentage of the total fitness each creature has. So if the total fitness is 100, and a creature has 20 fitness, 20% of the next generation is intialized with the genome of that creature.

You can play with most of the parameters, they are all documented in the source code. Under the following parameters

genome 16000
population 2048
mutation 8
fixation 0.8
fitness 1
badWeight 4
reps 4
numPressures x 5

The following data gets generated, for x = 1, 5, 10, 20, 40, and 80:

Overall rate under 1 pressures was 145.000000( 145.000000 on average per pressure ).
Overall rate under 5 pressures was 162.229996( 811.149979 on average per pressure ).
Overall rate under 10 pressures was 218.225006( 2182.250061 on average per pressure ).
Overall rate under 20 pressures was 204.338135( 4086.762695 on average per pressure ).
Overall rate under 40 pressures was 114.856873( 4594.274902 on average per pressure ).
Overall rate under 80 pressures was 56.574688( 4525.975037 on average per pressure ).

These statements mean "the overall rate under x pressures was R <fixations per generation> ( A <generations to fixation> on average per pressure ).

What does this mean? First, notice that the number of generations to fixation for a pressure monotonically increases as we add more pressures. For a single pressure it took 145 generations on average, but for 80 it took 4525 generations on average. Second, notice that the ovarall rate of fixation initially increases, then decreases. Eventually, at some point between 20 and 40 pressures, the rate of fixation passes that of a singly applied pressure.

Why? Because although each pressure takes much longer to fixate, they are all on their way to fixation at the same time -- in parallel. The factor each pressure adds to the total time to fixation is not as great as the bonus incurred by evolving against another pressure in parallel.

This demonstrates, among other things, that Kleinman is outright wrong. None of the data he has relied on goes over 4 or 5 pressures at once, so he has no way to refute this simulation (short of looking at the source code, which he won't do for reasons we all know).


Clearly, n + 1 selective pressures lead to a greater rate of fixation (evolution) than n pressures, for many values of n. Also, clearly, n pressures lead to fixation at a greater rate than a single pressure for many values of n.

Dr. Adequate is right. We are right. You are simply wrong, Kleinman. There is nothing else to say now. I did all the work of showing you, very clearly, that sorting/optimization problems are NOT always confounded by multiple sorting conditions. Particularly, when sorting happens in parallel, as it does in mutation and selection.

 

[Belz...]

Dodger, your model is not peer-reviewed. Klein can safely ignore it !

 

[rocketdodger]

wtf... I spend 4 hours writing a program that proves Kleinman is wrong, once and for all, and he disappears?

Well I guess that is a good thing...

 

[kjkent1]

wtf... I spend 4 hours writing a program that proves Kleinman is wrong, once and for all, and he disappears?

Well I guess that is a good thing...

kleinman will return -- count on it.

Far earlier in the thread, I did some studies with ev which showed that convergence of Rseq -> Rfreq appeared to form a normal distribution curve, centered on the maximum number of selection pressures applied. And, your program supports this idea -- which should be intuitively obvious, because everything in nature that happens to biological organisms seems to take on a bell-shaped curve.

Naturally, kleinman discounted my findings as nonsense.

So, allow me to congratulate you -- you're a very talented programmer.

 

[delphi_ote]

wtf... I spend 4 hours writing a program that proves Kleinman is wrong, once and for all, and he disappears?

He'll be back, and your program won't change his mind. If evidence mattered at all in his decision making process, he would've recanted his position hundreds of pages ago.

I'd really like to be wrong about this.

ETA Let me second the "talented programmer" comment. Your code is sexy!

 

[rocketdodger]

... because everything in nature that happens to biological organisms seems to take on a bell-shaped curve.

Pretty much they do, yeah. And its not just biological, its everything. The Gaussian distribution is simply the mathematical effect of lots and lots of flat probability distributions applied together. So you see it even in completely unnatural phenomena as well.

So, allow me to congratulate you -- you're a very talented programmer.

Thanks (you too delphi). I am just a newbie though... only got my degree last year. You should see some of the people in my industry!

 

[delphi_ote]

You should see some of the people in my industry!

If you work in the industry I think you work in, yea. I'm a PhD student at a top CS school, and I'm impressed by the code coming out of that industry...

 

[Belz...]

ETA Let me second the "talented programmer" comment. Your code is sexy!

Nerd.

 

[sol invictus]

You can’t change hundreds of genes simultaneously by mutation and selection. Transforming two genes by mutation and selection simultaneously is a much slower process than transforming a single gene.

Wrong. See below.

If you have two genes evolving simultaneously with each gene needing only a single mutation and the probability of 1 in a million for the probability of each mutation, the probability of any member of the population getting both mutations simultaneously is (1/1,000,000)^2. If you are evolving a similar system with a hundred genes evolving, the probability for a member of that population getting all the proper mutations simultaneously is (1/1,000,000)^100. That’s why you can’t evolve hundreds of genes simultaneously.

What you said here is technically correct, but it's the right answer to the wrong question. You don't want to ask what the probability is that all those mutations happened in a single generation - that would be extremely unlikely. But evolution happens gradually - you want to ask what the probability is that all those mutations happened after some large number of generations. Or to turn that around: how many generations it will take on average for them all to occur?

It's easy to get the answer. Suppose for a moment we only care about one gene, and the probability per generation that gene will mutate is p. Then the number of generations t it takes for the mutation to occur is on average simply t_avg = 1/p.

Now suppose there are N mutations we want to keep track of, and for simplicity assume that each has that same probability p to occur in any given generation. It's true that the chance for them all to occur in a single generation is very small: p^N. But what's the average time it takes for all N mutations to occur? That's slightly more complicated, but the answer is the following: the chance that after t generations a given mutation has not occurred is (1-p)^t. So that chance it has occurred is 1-(1-p)^t. Then the chance that all N have occurred is (1-(1-p)^t)^N. We can use that to compute the average time exactly, or make an estimate.

For N large, the answer is approximately t_avg ~ (1/p)(d + Log[N]), where d is a number of order 1. So the time grows very slowly with N at large N, and the average time per mutation decreases at large N as Log[N]/(pN). In your example, where p = 1/1,000,000 and N = 100, the average number of generations for all 100 mutations to occur is only 5.2 10^6, or 5.2/p.

This is a simpler version of what rocketdodger did above.

Executive summary: increasing the number of mutations you're tracking does not significantly increase the typical time required for them all to occur.

 

[kleinman]

Annoying Creationists

Paul who has the bicycle with the backwards facing seat because he has back peddled so much on what he claims about ev asks me to make up my mind. Well Paul, if you want to cling to the ridiculous notion that ev evolves something with genetic function, go for it. I’ll cling to my notion that all the ev sorting algorithm does is evolve sequences of bases which satisfy the selection conditions written for the model. My, my Paul, you really have some weird ideas about what ev does but that goes well with your belief in the theory of evolution.

Ev is a model, Alan. It evolves a function that models something in real life. It is an abstraction of real life. It obviously does not evolve an actual biological function, as in biology.

You must have a particular meaning of "genetic function" in mind when you make statements like the one I highlighted. Unfortunately, I don't know what that meaning is.

Of course ev does not evolve any actual biological function. What ev demonstrates is how the mutation and selection sorting/optimization process is profoundly slowed when trying to sort on multiple selection conditions. The reason why you don’t know the meaning of the highlighted text is there are no selection conditions that would evolve a genetic function de novo. You tried to suggest that the presence of oxygen led to the evolution of hemoglobin. To you want to try to put some details on your speculation or are vague speculations all you need to call your theory “scientific”.

Did you get your choice of colors for the paint which you use to paint yourself into a corner? So how do you get the evolution of a gene de novo?

By starting with something that has a very simple function. You can interpret that statement however you want, since I have no idea what you think a "biological function" is.

Just what is this “simple function”? What is your example of a “simple” form of hemoglobin? You are the one who is claiming that ev is an “abstraction of real life” and that ev is evolving a “function that models something in real life”. So, you tell us how this “abstraction of real life” is evolving a “function that models something in real life”.

What I claim is that ev is an idealized model of mutation and selection and what it shows is how profoundly slow the sorting/optimization process become when trying to sort by multiple selection conditions simultaneously. This is exactly what happens in all real, measurable and repeatable examples of mutation and selection. This is what ev demonstrates. So Paul, have fun trying to tell us what this “abstraction of real life” is evolving a “function that models something in real life”.

That’s what so much fun for me in this discussion, I don’t have to present the math; you and Dr Schneider have already done it for me.

Wow, we're good, huh? Okay, then could you please show me where in Evj we proved: hugenumberofgenerations = toolong

From the Evolutionisdead forum:

Extrapolating your 1 mutation per genome data to a lousy 1 million population gives 254 generations. To a measly 10 million population gives 106 generations. Even with a ratio of 2000:1, that's a mere 212,000 generations.

Even if I accepted your extrapolations as accurate, recall this series is based on a G=1000. Thus far, ev has shown that genome length is the dominant parameter in determining the number of generations for convergence. If your “primitive” microorganism has G=100,000 rather than 1000, and the number of generations is linearly related, the mere 212,000 generations becomes 21,200,000 generations to evolve the 16 binding sites.

Heck, make is 210,000,000. At one generation per day, that's an insignificant 575,000 years.

So Paul your estimate for the number of generations to evolve 96 loci in the ev model for a population of 1 million, G=100,000 and mutation rate=1/genome/generation is 210,000,000 generations. If you assume the generation time for this population is one generation per day, you get a mere 575,000 years. Paul, do you want to do the calculation to estimate the number of years required to evolve those 96 loci for a population that reproduces once a year? Perhaps you would like to use your extrapolation technique and estimate the number of generations for a population of 1 million but with a G=1 billion instead of 100,000? This is another demonstration of why the theory of evolution by mutation and selection is mathematically impossible. Mutation and selection is simply far to slow a process.

Do you want to go back over your estimate of how many hundreds of millions of generations it would take to evolve your function on 96 loci for a population of 100k genomes

I don't know which experiments you're referring to. What is "96 loci" and do you mean a population of 100K or a genome size of 100K? How many generations did I estimate?

Don’t you recall the G=1000 series we did, varying populations? Do you want me to repost that data? Paul, don’t you recall that all but the tiniest genomes take huge numbers of generations to evolve “function” in ev? And why does it take huge numbers of generations to evolve “function” in ev? It takes huge numbers of generations to evolve “function” in ev because the sorting algorithm becomes profoundly slow for the multiple sorting conditions for all but the tiniest search spaces. And Paul, that’s exactly what the empirical data for mutation and selection shows, this sorting/optimization process is profoundly slow for all but single selection conditions.

So the function evolved in ev is an abstraction of a transcription factor. Once again you put your bicycle in reverse gear. Paul, you are the only person I know who has multiple reverse gears on your bicycle. Paul, tell us what the function of an abstraction of a transcription factor does.

If you think I was ever saying that Evj evolves an actual transcription factor, then clearly one or the other of us is stark raving mad. The abstraction of the transcription factor matches the binding sites on the genome, but no other positions on it. I think I've said that numerous times.

How does your abstraction relate to reality? Why does it take such huge numbers of generations to evolve your “abstraction of transcription factor matches” on all but the tiniest genomes? Even Dr Schneider knows that it would take years of CPU time to evolve your “abstraction of transcription factor matched” for a realistic G value for a free living organism. Would you like a pair of matching pants for your straight jacket?

There you go Paul, explain to us how a single selection pressure in ev has no function yet the three selection pressures somehow do.

I said it had a different function. Pay attention.

Paul, you are so cute when you squirm. Why don’t you tell us what these different functions that ev is evolving when evolving only a single selection condition at a time.

Hey, your fellow evolutionist rocketdodger has a model of mutation and selection where the mutations are random and the selection pressures are random and he claim to show that n+1 selection pressures evolve faster than n selection pressures. Paul, perhaps you could tell us what function his model evolves or do only some selection pressures evolve genes with some time of function?

See, here's the problem, Alan. It doesn't matter what function his model evolves, as long as it shows that more selection pressures can sometimes evolve faster than fewer. This acts as a counterexample to your glib proof to the contrary. Now you must address his counterexample in your proof and prove that it has nothing to do with real life.

When rocketdodger was asked to describe his selection pressures in his model he said this:

The pressures are completely random, and each of them targets a single mutation. You can even set the relative fitness bonus per pressure if you wish, but that gets tedious pretty quick, so I added the ability to just set them all to a given fitness bonus.

Both Adequate and rocketdodger are claiming that n+1 selection pressures evolve more rapidly than n selection pressures. Neither has been able to present a real example of this nor has any other evolutionist been able to present such an example. Both Adequate’s and rocketdodger’s assertions have no basis in reality. I’ll continue to post real examples which contradict their baseless assertions, these examples demonstrates the mathematics which your own model shows, that is combination selection pressures profoundly slow the mutation/selection, sorting/optimization process.

But you can't do that yet. First you have to prove that Ev encompasses all possible real-life scenarios and that at least some of those scenarios did not have enough time to evolve.

Paul, I’ll keep posting scenarios of how mutation and selection actually works for as long as you want. Every real measurable and repeatable example of mutation and selection shows that combination selection pressures profoundly slow the evolutionary process and we will all wait for you evolutionists to post a single real example which contradicts these findings. Paul, I almost pity you in this debate, you have no mathematical basis for your beliefs, your own model contradicts your theory and you have no empirical evidence to support your scientifically baseless theory. The only real question in this debate is which is dumb and which is dumber, abiogenesis or the theory of evolution.

Sometimes you gotta do stuff in a particular order, don't you know.

Oh, do you mean sequential selection pressures?

So you want to draw a correlation between an intelligently designed computer program and a cell you claim came about by mutation and selection.

What’s your point?

That, in order to state with certainty that one thing is more complex than another thing, we need a precise definition of complexity.

I suspect you are going to have some difficulty coming up with a mathematical definition for complexity because this has subjectivity associated with it. You can come up with a mathematical definition for order or disorder but something that is complex or intricate for one person may be easy to understand for another. Complexity is a qualitative property of a system. You might be able to obtain a quantitative measure for complexity by using the number of lines of computer code necessary to describe they system. Do you want to try to write a computer model for a cell? Why not try to write a computer simulation of a mitochondrion?

So I guess you don’t want to take up Delphi’s banner that a temperature change was the selection pressure that transformed reptiles into birds. That’s what my original comment that you quoted was about.

It is up to you to present your evidence that Evolution is wrong. You are the one who argues against the scientific evidence.

What mathematical or scientific evidence have you presented? I have presented the mathematical data from a peer reviewed and published mathematical model of random point mutations and natural selection which shows that combination selection pressures profoundly slow the evolutionary process and I have presented hundreds of real empirical examples of mutation and selection, all which show that mutation and selection is profoundly slowed when you have more than a single selection pressure.

Temperature changes certainly change the conformation of enzymes and their catalytic properties. The reason why hundreds of genes can not change simultaneously is that the mutation and selection process is a sorting/optimization problem. You can’t change hundreds of genes simultaneously by mutation and selection. Transforming two genes by mutation and selection simultaneously is a much slower process than transforming a single gene. If you have two genes evolving simultaneously with each gene needing only a single mutation and the probability of 1 in a million for the probability of each mutation, the probability of any member of the population getting both mutations simultaneously is (1/1,000,000)^2.

If you are evolving a similar system with a hundred genes evolving, the probability for a member of that population getting all the proper mutations simultaneously is (1/1,000,000)^100. That’s why you can’t evolve hundreds of genes simultaneously.

Why do mutations necessarily need to be the same?

Mutations don’t have to be the same; in fact there is no reason for us to believe that mutations will be the same. When a biological system is subjected to multiple selection conditions simultaneously, it becomes profoundly difficult for the population to sort beneficial and detrimental mutations in order to increase the frequency of beneficial sequences of bases. It is this false assumption that multiple genes can evolve simultaneously to multiple different selection conditions that forms the foundation for the theory of evolution.

CFLarsen, here are a couple of examples which demonstrate empirically what happens to the mutation selection process when you have more than a single selection pressure.

http://en.wikipedia.org/wiki/Pesticide_resistance

http://www.absw.org.uk/Briefings/insecticide_resistance.htm

Combination selection pressures targeting more than a single gene profoundly slow the evolutionary process. That’s what the mathematics shows and that’s what the empirical evidence shows.

Ehhh....hello? Your examples are evidence of just how fast species can evolve when there is a lot of changes in a short time.

The problem with some pesticides quickly becoming less effective is precisely evidence that contradicts your argument.

Wake up and smell the coffee CF, these articles show that rapid evolution only occurs with the use of a single pesticide. Using a second pesticide in combination with the first profoundly slows the ability of these populations to evolve against both pesticides simultaneously. It works the same for HIV, HBV, HCV, influenza, malaria, cancer, weeds,… All these populations can evolve rapidly to a single selection pressure but add additional selection pressures simultaneously and these populations can no longer evolve quickly. Mutation and selection is simply a sorting/optimization process. The process can occur quickly when only a single gene is evolving in a population, but force a population to evolve multiple genes simultaneously and this becomes a profoundly slow process. That is what the mathematical and empirical evidence of mutation and selection shows. The theory of evolution by mutation and selection is mathematically and empirically impossible.

Notice how Kleinman completely ignores my posts, now ? He doesn't even bother with his "clever" scarcasm!

How could I ever be sarcastic beggaminases?

Actually three degrees, BS, MS and PhD.

I'm sure I know a few high-schoolers who have a better grasp of these things than you do.

Get one of those high-schoolers into the discussion, perhaps they could explain to us what a beggaminases is.

I have posted hundreds of citations which show how mutation and selection actually works

Actually, you have. You just don't realise it.

Sure I realize what they show and so do you; they are cherry picked to show that combination selection pressures profoundly slow the evolutionary process. Too bad there are no cherries on your trees, nothing but some dry rot on your evolutionary tree.

Indeed, we are doing something like this already. Different evolution simulations are as we speak competing in the Annoying Creationist fitness landscape, mutating and reproducing under various selection pressures.

We all look forward to you posting a real example of your simulations while I continue to post real examples of what Dr Schneider’s peer reviewed and published ev model shows which is combination selection pressures profoundly slow the evolutionary process.

Evolution has been shown to have happened. The Theory of Evolution is the statement of how it happens. If the theory were to be 'thrown into the dustbin', as you put it, it doesn't mean that evolution does not exist. The theory would be revised to show the new data. That is, after all, how science works.

Sure Shalamar, microevolution exists, Dr Schneider’s computer simulation shows how it works and the hundreds of citations which I have posted substantiates what Dr Schneider’s model shows. The concept of common descent by mutation and selection is a mathematical and empirical impossibility. Mutation and selection simply doesn’t work that way.

If you are evolving a similar system with a hundred genes evolving, the probability for a member of that population getting all the proper mutations simultaneously is (1/1,000,000)^100. That’s why you can’t evolve hundreds of genes simultaneously.

Your statement here pretty much explains why you can't "see" the obvious.

You are fixated on the idea that evolution must reach a predetermined goal. Your use of the phrase "proper genes" demonstrates this conclusively.

When did I use the phrase “proper genes”? There is a huge amount of evidence that particular selection pressures lead to particular mutations increasing in frequency in a population. This phenomenon is used to identify resistance to drugs, pesticides and herbicides. If evolution does not have a goal, why do the same mutations show up over and over to particular selection pressures?

You think that the change from a dinosaur form to a bird form was the result of something other than just an accumulation of various unpredictable biological changes which just happened to have provided their owners a fitness advantage at a particular historical point.

Evolution is not deterministic, Alan. Two trials of an experiment using identical selection pressures will not necessarily produce the same evolutionary outcome, because the "mutation" part of the evolutionary process cannot be predicted in advance.

Too bad that the evidence contradicts your speculations; for example, the use of particular drugs for treatment of HIV leads to increase in the frequency of particular mutations appearing in the viral population. This effect is used to identify when resistance to a particular drug has evolved.

OK I updated my simulation program and made it more realistic and quite a bit faster.

and

Clearly, n + 1 selective pressures lead to a greater rate of fixation (evolution) than n pressures, for many values of n. Also, clearly, n pressures lead to fixation at a greater rate than a single pressure for many values of n.

Dr. Adequate is right. We are right. You are simply wrong, Kleinman. There is nothing else to say now. I did all the work of showing you, very clearly, that sorting/optimization problems are NOT always confounded by multiple sorting conditions. Particularly, when sorting happens in parallel, as it does in mutation and selection.

Ok rocketdodger, here is your opportunity to give us a real example of your model. If you need help with your response, I’ll remind you what Adequate said.

So far as I know, no-one has done the experiment.

and

and too bad you don’t have any empirical examples of your silly graph ...

As I have explained to you, I produced the model because I've not heard of this precise experiment being done.

wtf... I spend 4 hours writing a program that proves Kleinman is wrong, once and for all, and he disappears?

Well I guess that is a good thing...

You missed me rocketdodger, how touching.

kleinman will return -- count on it.

and

He'll be back, and your program won't change his mind. If evidence mattered at all in his decision making process, he would've recanted his position hundreds of pages ago.

I'd really like to be wrong about this.

Kjkent1, for once you are correct, Delphi, you haven’t been correct on the concept of mutation and selection since you posted your citation to Wikipedia and the fitness landscape.

ETA Let me second the "talented programmer" comment. Your code is sexy!

Delphi, for your own good, lay off the sterno.

If you have two genes evolving simultaneously with each gene needing only a single mutation and the probability of 1 in a million for the probability of each mutation, the probability of any member of the population getting both mutations simultaneously is (1/1,000,000)^2. If you are evolving a similar system with a hundred genes evolving, the probability for a member of that population getting all the proper mutations simultaneously is (1/1,000,000)^100. That’s why you can’t evolve hundreds of genes simultaneously.

What you said here is technically correct, but it's the right answer to the wrong question. You don't want to ask what the probability is that all those mutations happened in a single generation - that would be extremely unlikely. But evolution happens gradually - you want to ask what the probability is that all those mutations happened after some large number of generations. Or to turn that around: how many generations it will take on average for them all to occur?

Welcome to the discussion sol. So let’s see what you are arguing.

It's easy to get the answer. Suppose for a moment we only care about one gene, and the probability per generation that gene will mutate is p. Then the number of generations t it takes for the mutation to occur is on average simply t_avg = 1/p.

Now suppose there are N mutations we want to keep track of, and for simplicity assume that each has that same probability p to occur in any given generation. It's true that the chance for them all to occur in a single generation is very small: p^N. But what's the average time it takes for all N mutations to occur? That's slightly more complicated, but the answer is the following: the chance that after t generations a given mutation has not occurred is (1-p)^t. So that chance it has occurred is 1-(1-p)^t. Then the chance that all N have occurred is (1-(1-p)^t)^N. We can use that to compute the average time exactly, or make an estimate.

For N large, the answer is approximately t_avg ~ (1/p)(d + Log[N]), where d is a number of order 1. So the time grows very slowly with N at large N, and the average time per mutation decreases at large N as Log[N]/(pN). In your example, where p = 1/1,000,000 and N = 100, the average number of generations for all 100 mutations to occur is only 5.2 10^6, or 5.2/p.

Sol, you have got your questions wrong. It is not how many generations it takes for all 100 mutations to occur in a population. It is how many generations it takes for all 100 mutations to show up in a single individual in the population. Then that individual has evolved to all 100 selection pressures.

Here is an example of how this computation is properly done.
http://www.billingpreis.mpg.de/hbp04/mybil.pdf

Despite the approval of almost 20 antiretroviral drugs and the use of combination therapy, successful treatment of HIV-infections is hampered by the emergence of drug-resistant genetic variants in response to therapy. Finding a new potent drug combination after treatment failure is considered challenging, because most accumulated mutations confer resistance to multiple drugs. We present three computational tools for the analysis and simulation of viral genomic sequences, phenotypic drug resistance, and clinical outcomes. Mtreemix is a software package for estimating and using mixture models of trees that describe probabilistically the evolution of drug resistance. Geno2pheno is a web-based system for the prediction of phenotypic resistance from viral genotypes. It also implements normalization methods that make these predictions comparable between different drugs. Finally, theo predicts virological response within a patient from the infecting viral strain and the selected drug combination. Together these models and programs provide a quantitative picture of the evolution of drug resistance and support the design of individualized antiviral therapies.

and

We mention a clinical study from De Luca, Cozzi-Lepri, Perno, Balotta, Di Giambenedetto, Orani, Mussini, Toti & d’Arminio Monforte (2003) to demonstrate the benefit of predicting phenotypic resistance for the selection of new antiretroviral regimens. These researchers analyze therapy changes accompanied by a genotypic resistance tests in 332 previously untreated patients. Phenotypes are predicted with geno2pheno for the components of the combination therapies, and each drug is scored as active if the virus is predicted susceptible to it. Using a Cox proportional hazards model, they show that patients with a combination therapy consisting of ≤ 2 active drugs have a significantly higher risk of virological failure (sustained virus load increase) than patients receiving ≥ 3 active drugs (p < 0.004). The authors also compare the performance of 11 rules-based interpretation systems and our data-driven approaches. Genotypic scoring based on the SVM predictions as implemented in geno2pheno turns out to be the only interpretation system that provides significant predictions of virological failure after 24 weeks of treatment (De Luca, Cingolani, Di Giambenedetto, Trotta, Baldini, Rizzo, Bertoli, Liuzzi,Narciso, Murri, Ammassari, Perno & Antinori 2003

These authors are answering the correct question Sol.
Here is another citation on the same topic which shows mathematically how mutation and selection works.
http://bioinformatics.oxfordjournals.org/cgi/content/full/21/21/3943

Summary: The development of drug resistance is a major obstacle to successful treatment of HIV infection. The extraordinary replication dynamics of HIV facilitates its escape from selective pressure exerted by the human immune system and by combination drug therapy. We have developed several computational methods whose combined use can support the design of optimal antiretroviral therapies based on viral genomic data.

and

Suppose we have estimated a mutagenetic tree model for the development of resistance to a certain drug. In particular, this model can be used to compute transition probabilities between mutational patterns. As described in the previous section we can predict the resistance phenotype from the genotype. Using a classifier restricted to the set of n mutations we predict each mutational pattern to be either susceptible or resistant. Now, for a given virus we may ask what the transition probability to any resistant state is. In fact, this question is crucial for minimizing the risk of resistance development with the next regimen. We refer to the genetic barrier as the probability of not reaching any resistant state after a fixed time period under therapy. This quantity can be calculated as the sum of the probabilities of all mutational patterns predicted as susceptible. Thus, a higher genetic barrier indicates that the virus is less likely to become resistant.

For example, Table 2 shows the genetic barriers to both low level and high level zidovudine resistance of the wild type virus under three different regimens, namely zidovudine monotherapy, double therapy with zidovudine plus lamivudine, and double therapy with zidovudine plus didanosine. The underlying mutagenetic tree model is the tree displayed in Figure 1 scaled to a mean sampling time of 96 weeks. As expected, the genetic barrier to zidovudine is always higher under the combination of zidovudine plus lamivudine than under zidovudine alone, because these drugs do not share any resistance mutations. More surprisingly, we find that zidovudine resistance appears to develop faster under zidovudine plus didanosine than under zidovudine monotherapy. This effect may be explained by the stronger selective pressure exerted by the double therapy and the cross-resistance profile of zidovudine and didanosine (Beerenwinkel et al., 2005b; Brun-Vezinet et al., 1997). Thus, the genetic barrier is a useful concept for designing effective treatment strategies.

and

In order to support clinical decision making on the basis of viral genomic data, we have developed and applied several computational methods and tools. Specifically, we have addressed data integration and management (Arevir database), prediction of drug resistance and co-receptor usage from genotypes (geno2pheno), modeling of the evolution of drug resistance and the genetic barrier by mutagenetic trees (mtreemix), and selection of optimal drug combinations (theo). The integration of various types of genomic, phenotypic and clinical data as well as the coupling of different computational models yields predictive models of therapy outcome that may support the design of combination therapies.

I post the above citation for two reasons. First I post this citation because of the highlighted text in the second paragraph. This is the first citation I have seen which indicates that combination therapy leads to faster evolution than monotherapy and the second point is these authors still are advocates of combination therapy. Belz, I think you are going to find the above quote is not a cherry in your bowl, only the pits but let’s see if you can extrapolate this finding to support common descent.

The nonsensical notion that n+1 selection pressures evolve more quickly than n selection pressures would lead one to believe that if you continued to add more drugs to the treatment combinations of HIV that the virus will evolve more quickly. This is mathematical nonsense which Adequate and rocketdodger a putting out.

 

[sol invictus]

Sol, you have got your questions wrong. It is not how many generations it takes for all 100 mutations to occur in a population. It is how many generations it takes for all 100 mutations to show up in a single individual in the population. Then that individual has evolved to all 100 selection pressures.

Nope - that's exactly what I calculated. To re-state, I answered the following question:

Suppose we start with a population of un-mutated individuals, and let's assume for simplicity that they're immortal and don't reproduce. How long do we have to wait before, say, 90% of the individuals have each undergone all N mutations, if the chance per unit time for any given mutation to occur in an individual is p?

Or asked another way, how long do we have to wait before each individual has a 90% chance of having undergone all N mutations?

Answer: t ~ log(N)/p (valid for p<<1 and N>>1).


It's mathematically the same question as the following: suppose there are N lotteries. I can buy tickets, and each ticket has a chance p of winning each of the lotteries (so one ticket might win nothing, or it might win one particular lottery (with chance p), or even all N (with chance p^N)). Question: how many tickets do I have to buy to have a 90% chance of winning ALL the lotteries?

Same answer as above: log(N)/p.

Again, to summarize: it takes hardly any longer for an individual to mutate N times than it does for it to mutate once. For example, 1,000,000 mutations will happen to one individual in only about 14 times as long as a single mutation.

 

[Paul C. Anagnostopoulos]

Paul, you are so cute when you squirm. Why don’t you tell us what these different functions that ev is evolving when evolving only a single selection condition at a time.

Well, let's see. If you only count mistakes points when a binding site is missed, then you are evolving a function that matches binding sites and probably every other position as well. If you only count mistake points when a spurious binding occurs, then you are evolving a function that does not match spurious positions and probably also does not match anywhere else.

Notice how those two functions are different from matching the binding sites but nowhere else. Simple, really.

~~Paul

 

[kleinman]

Annoying Creationists

Sol, you have got your questions wrong. It is not how many generations it takes for all 100 mutations to occur in a population. It is how many generations it takes for all 100 mutations to show up in a single individual in the population. Then that individual has evolved to all 100 selection pressures.

Nope - that's exactly what I calculated. To re-state, I answered the following question:

Suppose we start with a population of un-mutated individuals, and let's assume for simplicity that they're immortal and don't reproduce. How long do we have to wait before, say, 90% of the individuals have each undergone all N mutations, if the chance per unit time for any given mutation to occur in an individual is p?

This is why you are asking the wrong question. This is not simply a probability problem; it is a sorting and optimization problem as well. Only those most fit members of the population can accumulate the beneficial mutations. The harmful mutations are causing individuals to be selected out.

But for the sake of discussion, let’s consider your approach but instead of 90% of the individuals undergoing all N mutations, consider a single individual undergoing all N (100) mutations. In the first generation that individual has a (1/(1,000,000))*100 probability of getting a beneficial mutation, if that individual is gets one of those beneficial mutations, in the next generation, that individual has a (1/(1,000,000))*99 probability of getting a second beneficial mutation or a total probability of (1/(1,000,000))*100*(1/(1,000,000))*99 and so on for each additional possible beneficial mutation. The probability of a particular individual getting all one hundred beneficial mutations is (100/1,000,000)*(99/1,000,000)* … *(1/1,000,000). Sol, I’ll let you run the numbers for the probability of any individual getting all 100 beneficial mutations in 100 generations, let alone 90% of the population getting all 100 beneficial mutations. The point you are missing is that without selection, you can not improve the frequencies of genetic sequences in the population. You only have a few basic ways of improving these probabilities. You can increase population, number of generations, mutation rates (within limits) or you can reduce down the number of selection pressures.

Or asked another way, how long do we have to wait before each individual has a 90% chance of having undergone all N mutations?

Answer: t ~ log(N)/p (valid for p<<1 and N>>1).

It's mathematically the same question as the following: suppose there are N lotteries. I can buy tickets, and each ticket has a chance p of winning each of the lotteries (so one ticket might win nothing, or it might win one particular lottery (with chance p), or even all N (with chance p^N)). Question: how many tickets do I have to buy to have a 90% chance of winning ALL the lotteries?

Same answer as above: log(N)/p.

Again, to summarize: it takes hardly any longer for an individual to mutate N times than it does for it to mutate once. For example, 1,000,000 mutations will happen to one individual in only about 14 times as long as a single mutation.

So Sol, you are telling us that it only takes 14 times longer to win 100 lotteries as it takes to win 1 lottery? After all, that’s what an individual in the population has to do to get the 100 beneficial mutations and for that individual, the probability of winning one of the lotteries is decreasing with each beneficial mutation the individual wins.

Paul, you are so cute when you squirm. Why don’t you tell us what these different functions that ev is evolving when evolving only a single selection condition at a time.

Well, let's see. If you only count mistakes points when a binding site is missed, then you are evolving a function that matches binding sites and probably every other position as well. If you only count mistake points when a spurious binding occurs, then you are evolving a function that does not match spurious positions and probably also does not match anywhere else.

Notice how those two functions are different from matching the binding sites but nowhere else. Simple, really.

And now you can tell us what these different functions serve other than they give sequences of bases which either match or don’t match the weight matrix. After all, that’s all that your sorting algorithm really does, simple, really.

 

[Paul C. Anagnostopoulos]

And now you can tell us what these different functions serve other than they give sequences of bases which either match or don’t match the weight matrix.

Your sentence is a bit muddled. The function of the modeled gene, when all three selections are present, is to match binding sites but no other positions. This is modeled using a weight matrix and a threshold. So ultimately it's all a question of math. In life, it's all a question of chemistry. Either you think modeling chemistry with math is okay, or you think it's bogus. But regardless, the math is different when all three selections are present as opposed to only one or two of them. And so therefore is the modeled function.

~~Paul

 

[kleinman]

Annoying Creationists

And now you can tell us what these different functions serve other than they give sequences of bases which either match or don’t match the weight matrix.

Your sentence is a bit muddled. The function of the modeled gene, when all three selections are present, is to match binding sites but no other positions. This is modeled using a weight matrix and a threshold. So ultimately it's all a question of math. In life, it's all a question of chemistry. Either you think modeling chemistry with math is okay, or you think it's bogus. But regardless, the math is different when all three selections are present as opposed to only one or two of them. And so therefore is the modeled function.

I’m not sure what you are having trouble understanding in my statement. Each of the selection conditions in ev does a part of the sorting requirement of the genomes. In order to get an arrangement of bases that satisfy all three sorting conditions simultaneously, it is an extremely slow process for all but the tiniest genomes. This is your so-called function. However, any individual portion of the sorting process can be done much more quickly if done independently of the other sorting conditions. A portion of your total function can be achieved by applying only a single sorting condition.

Paul, I think that modeling chemistry with mathematics is more than okay, it allows for accurate application of physical laws. That said, ev is modeling how mutation and selection sorts and optimizes in a biochemical system and doing parts of the sort individually proceeds far more rapidly than doing the total sorting conditions simultaneously. That is the only function ev demonstrates.

 

[rocketdodger]

Ok rocketdodger, here is your opportunity to give us a real example of your model. If you need help with your response, I’ll remind you what Adequate said.

Why don't you give us a real example of the model used in ev?

You want a war of examples? Well guess what Kleinman, none of your examples show that multiple selection pressures slow evolution in general. All they show is that many species of organisms fail to develop resistance before their population is destroyed when a few human generated pressures are applied to them. You haven't shown a single study where the pressures satisfy any of the following conditions:

1) Greater in number than 3 or 4.
2) Natural as opposed to human applied.
3) Not intended to kill the organism outright.

Thus, your little theory will hold water if and only if those three conditions hold for every population in the history of Earth. Do you think they do?

You need to use mathematics to show anything else, because of how narrow your citations are. That is why you keep running home to your gross misinterpretation of ev. On your own, you haven't even given us an equation, not one single equation, that makes any sense, to work with -- at first I suspected it was because you know we will blow it apart, but now I think its because you are a fraud and in fact know nothing about math beyond what your elementry level bible school has taught you. But you know what? It doesn't matter, because I do.

The algorithm I used in my model is exactly the algorithm used in mutation and selection. The only thing it was intended to show is that a sorting/optimization problem is not always confounded by more optimization conditions -- and it does show that, plain and simple. My program sorts by fitness and in every possible set of conditions adding more optimization conditions eventually leads to faster sorting overall. It is a mathematical fact Kleinman -- you can try to argue against this but you are just wrong.

If you dispute anything, all you have to do is take a look at the source code (which others will attest is easy to follow) and tell me where I went wrong. Please, be my guest. Surely, with your three degrees and fancy professional background, you should be able to decimate the code that I, with only a lowly bachelor's degree, came up with from scratch.

The nonsensical notion that n+1 selection pressures evolve more quickly than n selection pressures would lead one to believe that if you continued to add more drugs to the treatment combinations of HIV that the virus will evolve more quickly. This is mathematical nonsense which Adequate and rocketdodger a putting out.

And the virus would, if the population survived. It is a mathematical fact. I have provided equations to show why. You have not provided equations to show why not, nor even done so much as critique our equations and show why they might be wrong. You are just full of sh-- Kleinman, there is no other way to put it.

 

[kleinman]

Annoying Creationists

Ok rocketdodger, here is your opportunity to give us a real example of your model. If you need help with your response, I’ll remind you what Adequate said.

Why don't you give us a real example of the model used in ev?

You want a war of examples? Well guess what Kleinman, none of your examples show that multiple selection pressures slow evolution in general. All they show is that many species of organisms fail to develop resistance before their population is destroyed when a few human generated pressures are applied to them. You haven't shown a single study where the pressures satisfy any of the following conditions:

1) Greater in number than 3 or 4.
2) Natural as opposed to human applied.
3) Not intended to kill the organism outright.

Rocketdodger, do you think a war of examples would really be fair. All you evolutionists with your speculations, extrapolations and hunches against one annoying creationist with a peer reviewed and published model of random point mutations and natural selection and hundreds of empirical examples which show that combination selection pressures profoundly slow the evolutionary process. Ok, fire your best speculation since you don’t have any examples.

So, now you have a new list of conditions. So let’s consider these new conditions. 1) Greater in number than 3 or 4, I believe that if you look at some of the citations for TB, you will find that sometimes more than 4 drugs are used. If you want, I’ll go back and find these citations. In addition, there are studies already appearing with the use of 4 drug combination therapy for HIV. It wouldn’t surprise me if 5 drug therapy may be tried. Stay tuned, I’ll look for examples of this. 2) Natural as opposed to human applied; does that mean that any experiment or measurement that has human involvement no longer has validity? Rocketdodger, are you going to start supplying examples of combination natural selection pressures accelerating evolution? 3) Not intended to kill the organism outright, rocketdodger, perhaps you aren’t aware that there are no viral treatments that kill the viruses; they only impair the viruses’ ability to reproduce. Many antibacterial agents are also bacteriostatic not bacteriocidal. All selection pressures are required to do is impair the population’s ability to reproduce.

Thus, your little theory will hold water if and only if those three conditions hold for every population in the history of Earth. Do you think they do?

I think there is enough water to sink the theory of evolution. Your theory has hit an iceberg. The theory of evolution is a Titanic theory.

You need to use mathematics to show anything else, because of how narrow your citations are. That is why you keep running home to your gross misinterpretation of ev. On your own, you haven't even given us an equation, not one single equation, that makes any sense, to work with -- at first I suspected it was because you know we will blow it apart, but now I think its because you are a fraud and in fact know nothing about math beyond what your elementry level bible school has taught you. But you know what? It doesn't matter, because I have.

Oh, ev does show other things about mutation and selection, for example huge populations do not compensate as much as evolutionists like to claim. Ev shows the accelerating effect on the sort process done by mutation and selection from increasing population decreases rapidly with increasing population. You don’t have to do much interpretation here, just plot the data. I don’t have to present my own mathematics; Dr Schneider and Paul have done this already. All I have had to do is present the data from their mathematics. Why should I reinvent the wheel when Dr Schneider has properly modeled the mathematics of mutation and selection and the model has already gone through the peer review process.

The algorithm I used in my model is exactly the algorithm used in mutation and selection. The only thing it was intended to show is that a sorting/optimization problem is not always confounded by more optimization conditions -- and it does show that, plain and simple. My program sorts by fitness and in every possible set of conditions adding more optimization conditions eventually leads to faster sorting overall. It is a mathematical fact Kleinman -- you can try to argue against this but you are just wrong.

If your algorithm properly models mutation and selection, give us a real example of your model. Tell us what these random selection pressures are that you use in your model and how they accelerate the evolutionary process. I’m not asking much of you.

If you dispute anything, all you have to do is take a look at the source code (which others will attest is easy to follow) and tell me where I went wrong. Please, be my guest. Surely, with your three degrees and fancy professional background, you should be able to decimate the code that I, with only a lowly bachelor's degree, came up with from scratch.

Hey, perhaps you and Adequate have the mutation and selection process correctly modeled and n+1 selection pressures evolve more rapidly than n selection pressures so it should be easy for you to produce a real example of this. I won’t even require that you limit your example to “natural” processes, you can even use a human designed experiment.

The nonsensical notion that n+1 selection pressures evolve more quickly than n selection pressures would lead one to believe that if you continued to add more drugs to the treatment combinations of HIV that the virus will evolve more quickly. This is mathematical nonsense which Adequate and rocketdodger a putting out.

And the virus would, if the population survived. It is a mathematical fact. I have provided equations to show why. You have not provided equations to show why not, nor even done so much as critique our equations and show why they might be wrong. You are just full of sh-- Kleinman, there is no other way to put it.

Too bad when you were studying for your bachelors degree, nobody told you that antiviral drugs do not kill the virus, they just impair the reproduction of the virus. Of course, if you knew this, you could present empirical examples of how mutation and selection actually works, like this:
http://qjmed.oxfordjournals.org/cgi/reprint/91/9/593.pdf

In order to support clinical decision making on the basis of viral genomic data, we have developed and applied several computational methods and tools. Specifically, we have addressed data integration and management (Arevir database), prediction of drug resistance and co-receptor usage from genotypes (geno2pheno), modeling of the evolution of drug resistance and the genetic barrier by mutagenetic trees (mtreemix), and selection of optimal drug combinations (theo). The integration of various types of genomic, phenotypic and clinical data as well as the coupling of different computational models yields predictive models of therapy outcome that may support the design of combination therapies.

Or how about this one rocketdodger which show that “natural” bacterial toxins when used in combination slow the evolution of mosquito larvae resistance.
http://aem.asm.org/cgi/content/abstract/73/19/6066

Two mosquitocidal toxins (Mtx) of Bacillus sphaericus, which are produced during vegetative growth, were investigated for their potential to increase toxicity and reduce the expression of insecticide resistance through their interactions with other mosquitocidal proteins. Mtx-1 and Mtx-2 were fused with glutathione S-transferase and produced in Escherichia coli, after which lyophilized powders of these fusions were assayed against Culex quinquefasciatus larvae. Both Mtx proteins showed a high level of activity against susceptible C. quinquefasciatus mosquitoes, with 50% lethal concentrations (LC50) of Mtx-1 and Mtx-2 of 0.246 and 4.13 µg/ml, respectively. The LC50s were 0.406 to 0.430 µg/ml when Mtx-1 or Mtx-2 was mixed with B. sphaericus, and synergy improved activity and reduced resistance levels. When the proteins were combined with a recombinant Bacillus thuringiensis strain that produces Cry11Aa, the mixtures were highly active against Cry11A-resistant larvae and resistance was also reduced. The mixture of two Mtx toxins and B. sphaericus was 10 times more active against susceptible mosquitoes than B. sphaericus alone, demonstrating the influence of relatively low concentrations of these toxins. These results show that, similar to Cyt toxins from B. thuringiensis subsp. israelensis, Mtx toxins can increase the toxicity of other mosquitocidal proteins and may be useful for both increasing the activity of commercial bacterial larvicides and managing potential resistance to these substances among mosquito populations.

Rocketdodger, go ahead and fire some more of your strange speculations as to how mutation and selection works and I’ll fire back some real examples as to how mutation and selection works and they don’t show that n+1 selection pressures evolve more rapidly than n selection pressures.

 

[Paul C. Anagnostopoulos]

I’m not sure what you are having trouble understanding in my statement. Each of the selection conditions in ev does a part of the sorting requirement of the genomes. In order to get an arrangement of bases that satisfy all three sorting conditions simultaneously, it is an extremely slow process for all but the tiniest genomes. This is your so-called function. However, any individual portion of the sorting process can be done much more quickly if done independently of the other sorting conditions. A portion of your total function can be achieved by applying only a single sorting condition.

But a portion of a whole function is not the same function as the whole function, is it? Lots of things in life are made easier by only doing part of the whole task.

~~Paul

 

[Olowkow]

Reply to annoying ID'ers

Just gotta love this reply by P.Z. Meyers to an inane email from a Mr. Wood, which can be seen in its entirety on Pharyngula (scienceblogs.com/pharyngula/2007/08/i_get_email_5.php#more):

Somebody is seriously overcompensating, aren't they? That's some piece of twisty-turny logic couched in arch and overwrought language. Just a suggestion, Mr Wood: you can't fill a vacuum with pedantry, no matter how much you try to shovel in.

Let me help. I get this argument all the time: "you wouldn't be so angry if the Designists/Creationists/Illuminati/Holocaust Deniers/Second Gunmen/Flat Earthers weren't right!" It's a very silly rationale, and no, writing it in a more longwinded style doesn't help.

There's a simple reason why biologists get pissed off with creationists, and it has nothing to do with a "first person ontology" — it's that we have the hard work, the data, the experiments, the whole dang enchilada of the "objective facts of the matter," and pretentious pissants like Mr Wood think nothing of overlooking their own self-admitted ignorance of evolution to pronounce a verdict based entirely on their half-assed psychoanalysis of the universe. We can see quite clearly (especially in this instance) what it is that drives a person to oppose Darwin (as if ol' Chuck had anything to do with the issue at this point): it is the arrogance of incompetence, the self-satisfied smugness of preening ********, the sanctimony of pious lackwits, the insufferable stupidity of pompous windbags who think they can rationalize their superstitions by seeking justification in a kind of gasified cold reading.

Your bubble-headed ******** doesn't bamboozle me, Mr Wood — I think the only person your verbose drivel might persuade is another superficial drone who mistakes diarrhea for depth.

 

[Foster Zygote]

Just gotta love this reply by P.Z. Meyers to an inane email from a Mr. Wood, which can be seen in its entirety on Pharyngula (scienceblogs.com/pharyngula/2007/08/i_get_email_5.php#more):

Somebody is seriously overcompensating, aren't they? That's some piece of twisty-turny logic couched in arch and overwrought language. Just a suggestion, Mr Wood: you can't fill a vacuum with pedantry, no matter how much you try to shovel in.

Let me help. I get this argument all the time: "you wouldn't be so angry if the Designists/Creationists/Illuminati/Holocaust Deniers/Second Gunmen/Flat Earthers weren't right!" It's a very silly rationale, and no, writing it in a more longwinded style doesn't help.

There's a simple reason why biologists get pissed off with creationists, and it has nothing to do with a "first person ontology" — it's that we have the hard work, the data, the experiments, the whole dang enchilada of the "objective facts of the matter," and pretentious pissants like Mr Wood think nothing of overlooking their own self-admitted ignorance of evolution to pronounce a verdict based entirely on their half-assed psychoanalysis of the universe. We can see quite clearly (especially in this instance) what it is that drives a person to oppose Darwin (as if ol' Chuck had anything to do with the issue at this point): it is the arrogance of incompetence, the self-satisfied smugness of preening ********, the sanctimony of pious lackwits, the insufferable stupidity of pompous windbags who think they can rationalize their superstitions by seeking justification in a kind of gasified cold reading.

Your bubble-headed ******** doesn't bamboozle me, Mr Wood — I think the only person your verbose drivel might persuade is another superficial drone who mistakes diarrhea for depth.

Welcome. And thanks for the chuckles.