|
Author
|
Topic: Selection Acting Directly on Genes
|
warren_bergerson
Member
Member # 262
|
posted 26. September 2002 14:53
Reply to Frances,
I can make no sense of your comments. The stationary population assumptions I used represent a commonly used, widely validated, and reasonable approach. If you think the result produced using standard analytical techniques you need to go argue with a couple of hundred years of validated actuarial/mathematical traditions.
Reply to Charlie,
Quote: Not clear what you mean. All drift models (indeed, all evolutionary models) of course account for mutation frequencies.
Really. Then why don’t all such ‘selection neutral’ simulations produce divergence rather than drift? The use of ‘blatantly and intentionally’ unrealistic assumptions appears, IMO, to be standard practice.
Quote: . Simulations of evolutionary change ‘assume’ an unlimited amount of time and lives available to produce result X.
You are right, I miss stated this item. I should have said evolutionary models make ‘unrealistic’ assumptions regarding the time available for change. The next item should have been models make unrealistic assumptions regarding the power and speed of systems using only ‘phenotype effect selection’. As you yourself demonstrated, even simple evolutionary change can not be simulated using realistic assumptions.
One day people argue that a major genetic factor like ‘gene direct selection’ can’t possibly exist and no analysis can possible measure the force of such a process. The next day the very same people are claiming that the factors which couldn’t have existed have now been properly reflected in all the analysis performed to date.
Obviously, the soundness of assumptions used can only be determined by looking at the assumptions used in individual studies. IMO, the blanket claim that realistic assumptions are used in testing TOE, is suspect.
Response to RBH,
Quote: Well, actually that's not the case. In the GAs my firm builds we have to go to a good deal of work and complication (think developmental processes) to provide phenotype-level selection. We do that because it is phenotypes that have to go out into the real world to control real processes. It is much easier in GAs to implement gene-level selection.
I agree that in most practical applications of GAs, the use of phenotype selection is appropriate (and complicated). I also agree that the term GA or evolutionary algorithm is so broad it covers any type of projection model. I also agree it is possible to simulate gene-level selection and that such simulations are probably much easier to simulate than phenotype simulation. Having been corrected let me restate my comments.
The analysis performed suggests that gene direct selection is a strong force in genetic change processes.[I also recognize that gene direct selection as a strong force, does not necessarily mean it has a material impact on changes produced. It could be noise. ] In attempting to simulate, test or explain evolutionary change with GA models, it is appropriate either to reflect gene direct selection or to demonstrate that ignoring this strong force will not distort the results of any tests performed. The validity of any simulation/test performed in the past which did not appropriately recognize the strong gene direct selection force may be suspect.
Do you like that better?
Quote: As for your 5 "hidden assumptions," charlie d's comments cover them very well. I'll remark only that in my description of the EAs my firm builds and deploys I mentioned that we pipe real selective environments from the outside world into the programs, so whatever properties those real environments have, including their stability or lack thereof, is available to our GAs.
The question here is not the soundness of practical applications of EAs and GAs. The issue here is the soundness of EAs and GAs as models of evolutionary processes or as models for testing GA models. No one, as far as I know would question that a GAs can find optimal or near optimal solutions to moderately complex problems.
But GA solutions to even moderately complex problems require very complex set up routines(to define the phenotype selection criteria). These algorithms are very slow requiring the equivalent of many, many lives and many, many generations.
Do GAs or RM&NS really reflect how biological systems evolve? The fact that gene direct selection, the major force of selection in genetic systems, is not even addressed in most GAs or RM&NS models is one of a number of factors which suggests that GAs and RM&NS may not provide adequate models, theories or explanations of evolutionary change. At the very least, GSH shows there is a major force/factor in genetic change which has not been explicitly recognized.
IP: Logged
|
|
peroxisome
unregistered
|
posted 27. September 2002 07:40
Hi Warren
any time you wish to address the issues I raised about point 5 of your clear and unambiguous evidence for GSH...
quote:
It therefore seems very likely that you, and many individuals with a background in evolutionary biology, are going to have a hard time understanding and accepting a ‘simple technique for measuring selection rates’, because you have been taught that measuring these rates is impossible or impractical.
I think you make unfounded, and untrue, assumptions about what people have been taught.
quote:
The analysis does not require references to any specific supporting studies.
so many of the comments identify serious defects about how your verbose mathematical constructs contact with reality. I believe this to be little short of a declaration of the principles of philistinism.
quote:
Note, if you claim that I have relied on incorrect information and that reliance has distorted the results, you can provide evidence supporting your claim.
quite. However, you will find that scientific method involves setting out your premises and assumptions, so that people can critically evaluate what you have done. Science thrives on open and honest discussion.
yours
per
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 27. September 2002 10:34
As this discussion winds down, it may be useful to summarize some of the key features of the Genetic Selection Hypothesis and the analysis used to develop it.
TWO FORMS OF SELECTION
GSH asserts there are two identifiable, measurable, categories of selection processes operating in gene change processes. One class of selection is the well known ‘Darwinian natural selection’ or ‘selection based on phenotype’. The second type or class of selection process is ‘selection acting directly on genes’. For convenience the two are labeled here as phenotype selection and genotype selection.
The estimates developed here suggest that in genetic change processes the ‘forces of genotype selection’ are at least 100 times greater than the ‘forces of phenotype selection’. Although beyond the scope of this thread, there are suggestions that in many situations the forces of genotype selection may be millions of times greater than the forces of phenotype selection. Whatever the numbers involved, genotype selection appears to be a very much stronger force than the more publicized phenotype selection.
The best known examples of genotype selection involve what is called ‘error correction’. These are processes which ‘select out’ many types of mutations without impacting the survival or future reproductive ability of the organism.
The known and recognized forms of genotype selection all appear to act to prevent or slow down change. As far as I am aware, there is currently no direct evidence of genotype selection operating to produce positive adaptive genetic change(where genetic change is defined as changes in alleles).
On a simple common sense basis, it seems reasonable to speculate that 1)if as appears to be the case, genetic change is a powerful and efficient process, and 2)if genotype selection is( at least at times) millions of times stronger than phenotype selection, then 3)genotype selection is likely to play a role in producing, as well as preventing, change.
It can, and I am sure will, be argued that the lack of direct evidence for ‘positive change from genotype selection’ is a strong argument suggesting that such positive change mechanism are rare. It can also be argued that a combination of bias, not knowing what to look for, and not knowing where to look, have prevented scientists from recognizing ‘positive changes produced by genotype selection. IMO, the techniques used to quantify the force of genotype selection could useful in addressing this issue.
ACTUARIAL/MARKOV MATHEMATICS
Most of the discussion/debate here has centered around the existence and effectiveness of actuarial mathematical techniques. [As an aside, let me assure you that ‘having your personally veracity attacked (even by a moderators on another sight), simply for suggesting that actuarial mathematics could be used to perform certain simple calculations’ is not a particularly pleasant experience.] I want to once again thank RBH for pointing out that not only does actuarial multi-decrement analysis exist, but it even taught in certain genetics courses under the label Markov analysis.
Having established the existence of multiple-decrement analysis, it is also useful to note that apparently very few individuals involved with EB are aware of the form of multiple decrement analysis used to justify GSH. Even fewer individuals in EB, it appears, have what could be called a working knowledge of the techniques. This explains in part why something as simple as ‘quantifying the force of genotype selection’ comes as such a surprise to so many people.
It is also useful to note that there are two ‘historical’ features of multiple decrement analysis that may help explain why the actuarial techniques being used here have not been more widely used. In the days before the widespread use of computers (I am talking prior to 1980, not prior to 1890) it was very difficult to perform actuarial calculations. Difficulty in performing calculations combined with the need to perform the calculations led to the development of numerous simple but highly effective approximation techniques. Most of the techniques used to justify GSH are based these approximation techniques.
With the widespread availability of computers, actuarial mathematics diverged into two separate paths (this is my personal view/interpretation). The one path attempted to eliminate the approximations and convert actuarial mathematics back into a continuous calculus/stochastic form similar to the formal mathematics used in physics. The other path retained the discrete/approximate nature of actuarial mathematics and applied it to the development of discrete/deterministic simulation/hypothesis testing models. The calculus/stochastic form became the ‘theoretical/academic form’ of actuarial mathematics and the discrete form became the pragmatic form used to solve practical problems. GSH is based on the discrete or pragmatic form of actuarial mathematics. I suspect, that the actuarial mathematics typically taught and used in academia is the, IMO, less useful calculus/stochastic form.
Many may find this topic boring, but IMO, it is important to be able to explain ‘how’ an obscure 200 year old form of mathematics can produce results which appear to be ‘incompatible with’, ‘surprising’ and ‘impossible’ when compared to the results produced by more ‘modern analytical techniques’.
MODELING ISSUES
GSH is based on a highly validated actuarial techniques which makes it possible to measure the forces of selection applicable to potential alleles. The existence of this technique, and the results of applying this technique have three ‘interesting implications’ for modeling genetic change.
1. PRACTICALITY OF BUILDING ‘ENTIRE GENOME SIMULATIONS’
The popular/traditional view is that it is not practical/possible to construct meaningful models/simulations of genomic change. However, if you have reasonable estimates of 1)the possible or likely mutations, 2)the rates of increment (mutation), and 3)rates of decrement(selection) then full genomic simulations are logically possible, and with modern computers practical. [While it may not be readily apparent what such simulations would be used for, it should be apparent that such models are possible/practical.]
2. REALISM OF GA MODELS
In one form or another, the claim/suggestion is made that GA models can be used to demonstrate/test genetic change in biological systems. The GSH analysis suggests that selection in genetic systems is 99%+ genotype selection. Most GA simulations involve 99%+ phenotype selection. There is, therefore, at least a ‘reasonable’ basis for questioning the validity of GA models as representations of genetic change.
3. VALIDITY OF RM&NS CLAIMS
At least in the popular literature, there are numerous statements and claims of the form "RM&NS explains…" or "Darwinian theory explains ….". There are also claims of the general form "It has been demonstrated that ….". GSH at least suggests that the soundness and validity of some of these claims should/could be investigated.
AC-POP VERSUS REAL SCIENCE POSITIONS
In discussing TOE, it is, IMO, useful to distinguish between ‘academic/popular’ (AC-POP) positions and ‘Real science’ positions. GSH suggests that 99%+ of genetic selection involves genotype selection. GSH clearly contradicts the Ac-Pop positions of ‘RM&NS explains’ and ‘Natural selection explains’.
GSH, however, is not necessarily incompatible with the ‘current real science TOE’. We can argue from a philosophical perspective whether current TOE deserves to be called a real ‘hard science’ theory, but if we accept what is generally referred to as TOE, then we accept that even a ratio of 1 million to one for genotype to phenotype selection is not incompatible with current TOE.
The problem/issue, IMO, is that far too many academics can’t or won’t differentiate between ac-pop and real science positions. IMO, GSH illustrates the gap that exists today in EB between ac-pop and real science positions. How the issue can/should be addressed is still unresolved, IMO.
Response to per:
IMO, it is time to put an end to these "I am the expert and since you don’t agree with me you are an idiot’ arguments. Many individuals, IMO, were legitimately surprised that some unfamiliar form of analysis could produce results so different from the results produced by more conventional forms of analysis. However, at this time, we have established that 1)actuarial multiple decrement techniques do exist, 2)they can and have been applied to measuring selection rates, and 3)relatively few individuals within EB have a working knowledge of the analytical techniques being used.
Given what we have learned, it is time to move from the ‘its impossible’ mode to the ‘does it actually work’ mode. Given what we know, it is appropriate to move from the ‘attack/discredit mode’ to the ‘analysis/understand mode’. Anyone with any real experience in science or business knows and recognizes the differences between the two types of behavior.
In introducing design science techniques based on actuarial techniques I have, and will continue to, make a variety of claims. I do not expect or desire that these claims be expected on faith. I do, however, expect that these claims will be addressed from the analysis mode, not the attack mode. (‘Expect’ is probably too optimistic a term.)
I will be happy to address the issue you raise, but only if you make the appropriate adjustments in the form of your request.
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 27. September 2002 11:00
quote:
Really. Then why don’t all such ‘selection neutral’ simulations produce divergence rather than drift? The use of ‘blatantly and intentionally’ unrealistic assumptions appears, IMO, to be standard practice.
I am utterly confused by your use of language. Drift is what you observe in one population, divergence is what you observe when you compare more than one population. Two independently drifting populations eventually diverge (provided population sizes and absence of selection favor drift, of course). That's why our pseudogenes differ from chimps' more than our genes do: pseudogenes drift, hence they diverge.
quote:
You are right, I miss stated this item. I should have said evolutionary models make ‘unrealistic’ assumptions regarding the time available for change. The next item should have been models make unrealistic assumptions regarding the power and speed of systems using only ‘phenotype effect selection’. As you yourself demonstrated, even simple evolutionary change can not be simulated using realistic assumptions.
I did of course no such thing. I just showed a simplified model of an otherwise untreatable evolutionary problem (the evolution of longer wings) which for some reason you thought was particularly meaningful, just to show the principle of how evolution of a quantitative trait could occur. For some reason, despite my repeated explanations, you insist claiming I was trying to show how wings actually evolved: that was what you were trying to do.
In models of realistically treatable evolutionary situations, evolutionary simulations do in fact an excellent job, that's why they are used in a number of non-biological situations when our designing brains are stumped or simply not creative enough (such as RBH's economic models).
quote:
One day people argue that a major genetic factor like ‘gene direct selection’ can’t possibly exist and no analysis can possible measure the force of such a process. The next day the very same people are claiming that the factors which couldn’t have existed have now been properly reflected in all the analysis performed to date.
Hard to say what you are referring here. Selection acting on a DNA sequence (such as "selfish" DNA), with no effect on phenotype can be measured by standrd darwinian models. What I and other people were asking for, was an explanation of what exactly your "gene-level selection" was. Just a basic, simple description of how the process works, and what effects it. All we got was more comments on actuarial mathematics and the inadequacy of darwinian models. Without an actual explanation, your claim that "gene-direct selection" (according to warren, not the run-of-the-mill dawkinsian one) is "the major force of selection in genetic systems" is simply pulled out of nowhere.
Finally, not directed to me, but anyway: quote:
It therefore seems very likely that you, and many individuals with a background in evolutionary biology, are going to have a hard time understanding and accepting a ‘simple technique for measuring selection rates’, because you have been taught that measuring these rates is impossible or impractical.
Actually, measuring selection coefficients is one of the first things you learn in basic population genetics and evolutionary biology. Open any textbook and take a look.
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 27. September 2002 11:06
Dear Warren,
How can you make the assumption of steady state and at the same time you assume that mutations happen and selection plays a role. Under such assumptions your enforce a 100% mortality rate for all mutations that happen.
I have shown why your assumptions are inappropriate to represent reality.
GSH seems to be dead as far as a realistic model of selection and mutation.
Thus your conclusions about phenotype and genotype selections seem to be totally unwarranted by your model for two simple reasons: 1) your model is erroneous 2) your model does not even suggest this.
Your 'model' is a simplistic variant of Markov models with a steady state assumption enforcing 100% extinction rates for mutations. Thus the results are not incompatible with scientific findings, they just show what goes wrong when you make a poor model.
I have investigated your GSH related claims and found them to be without any merrit beyond an educational level of what can go wrong with modeling. The analysis is simple and straightforward, if you allow for mutations to happen but at the same time enforce steady state then mutations are all 100% detrimental.
QED
BTW I would love to see you refer some of these "blatantly and intentionally" unrealistic assumptions you seem to refer to. Could you give some explicit examples rather than vague and somewhat insulting accusations?
IP: Logged
|
|
peroxisome
unregistered
|
posted 27. September 2002 19:43
Hi Warren
quote:
IMO, it is time to put an end to these "I am the expert and since you don’t agree with me you are an idiot’ arguments.
errr, like "i am the expert in actuarial mathematics and all evolutionary biologists are idiots" type of statement ?
I agree with you absolutely Warren. That is why you raised 7 points of 'clear and unambiguous' evidence for GSH. In order to address the substance of your argument, I identified your point 5 as apparently being totally without foundation, and raised the issue with you. I have asked three times- pretty please. So far, it appears to me that point 5, and hence 6/7, are just unfounded gibberish.
quote:
The GSH analysis suggests that selection in genetic systems is 99%+ genotype selection. Most GA simulations involve 99%+ phenotype selection.
I have to say that it appears that you routinely make claims which appear to be wholly unsubstantiated. Why GSH suggests "99%+ genotype selection" is a complete mystery to me, because you have not set out what you have done to arrive at your conclusion. There appears to be negligible contact with the reality known as "biology" in these discussions, and this is a problem.
yours
per
[ 27. September 2002, 19:54: Message edited by: peroxisome ]
IP: Logged
|
|
Moderator
Administrator
Member # 1
|
posted 27. September 2002 22:40
This thread is being closed for various reasons. Primary among them is that fact that people are talking past each other and the discussion is digressing rather than progressing.
There are other reasons. Labeling someone's argument as gibberish is not appropriate. If you think that someone is making a fool of him/herself(speaking gibberish), the best way to make that statement is to ignore the argument. Beating on someone whose argument you find vacuous is "out of bounds."
The moral of the story...if you don't think someone is worth responding to, or you think someone is fundamentally wrong, say it with silence not ridicule.
IP: Logged
|
|
|
Topic Closed
Printer-friendly view of this topic
