|
Author
|
Topic: Simulating Self Assembly
|
warren_bergerson
Member
Member # 262
|
posted 02. February 2003 11:13
The purpose of this thread is to discuss the experimentally verifiable hypothesis that "biological development is a biological design process which can not occur without the presence of information generating ‘intelligent design processes’". The hypothesized existence of scientifically measurable and verifiable intelligent design processes as part of biological development processes would appear to contradict most of the widely held beliefs in both biology and ID. The fact that the hypothesis is unconventional should not make it an inappropriate topic for discussion here.
To start the discussion, let us consider the following concepts and definition.
ASSEMBLY AND SELF ASSEMBLY PROCESS- A series of steps, ‘state changes’, or operations which produce some type of defined transformation. An assembly process may involve an intelligent assembler such as a human who modifies or interprets the assembly instructions so that the desired end product is produced. The term self assembly is used to refer to processes that operate over a period of time without interference by an external intelligent assembler.
ASSEMBLY INSTRUCTION- An assembly process is defined and modeled by a set of assembly instructions operating in a defined environment. An assembly instruction can be characterized as a ‘change of state operation’ or a causal relationship of the form ‘stimulus or trigger S causes response R’. In a mechanical assembly mechanisms, for example, "‘temperature drops below X’ causes ‘turn on heating element’" would be an example of assembly instruction.
DESCRIPTIVE VERSUS COMPLETE OR CAUSAL INSTRUCTIONS- A set of self instructions is complete or ‘causal’ if and only if the assembly instructions, applied in defined environment can actually produce the defined result. An incomplete set of instructions describing selective key steps are descriptive instructions.
‘DYNAMIC AND TELEOLOGICAL’ VERSUS ‘FIXED’ INSTRUCTIONS - An assembly instruction is defined as ‘dynamic and teleological’ or ‘programmable’ if 1)the instruction can take many different forms, and 2)only a portion of the possible forms are compatible with the defined final assembly objective. The complexity or ratio of possible forms to teleological forms defines the amount of information needed to code an assembly instruction. An assembly instruction with a complexity of 1 is a fixed instruction. An instruction can be fixed with respect to a particular assembly process and/or fixed with respect to all assembly process that could evolve. A permanent assembly process which does not evolve would typically be considered part of the environmental conditions.
INTELLIGENT DESIGN PROCESS- As defined here, an intelligent design process is a complex set of operations capable of 1)finding the teleological form of a causal relationship such as an assembly instruction, 2) adding to the set of possible instructions by adding to the lists of recognized inputs, stimuli or causes, 3)adding to the set of possible outputs, responses or effects (capabilities 1, 2, and 3 make it possible to produce creative teleological assembly instructions), 4)creating new instructions, and 5)increasing the processing capacity of the design process. It is possible to define abstract mathematical processes(logic machines) which have all five capabilities. Such logic machines are used here to model intelligent design processes in self assembly processes.
Given the above definitions we can suggest at least four types of experimental design.
1. Evaluate the possibility of simulating complex assembly without dynamic and teleological instructions.
2. Evaluating the possibility of simulating complex assembly without dynamic and teleological instructions that take new teleological forms during the assembly process.
3. Evaluating the possibility of simulating complex assembly without intelligent design processes.
4. Evaluating the possibility of simulating the evolution of any of the identified assembly instructions using only natural selection and random mutation.
The hypothesis I proposed at the beginning predicts that 1)any clearly defined developmental processes can be modeled and simulated using intelligent design processes, 2) the evolution of any defined developmental process can be modeled and simulated using intelligent design processes, and 3) there will be clearly defined developmental processes which it will not be possible to model without intelligent design processes.
This is an admittedly quick and dirty overview of the proposed hypothesis and experimental design. I will gladly entertain any questions or comments before proceeding.
IP: Logged
|
|
Evan
Member
Member # 164
|
posted 02. February 2003 13:14
Warren begins his post by writing "The purpose of this thread is to discuss the experimentally verifiable hypothesis that "biological development ..."
and he later writes,
"Given the above definitions we can suggest at least four types of experimental design.",
He then lists the following:
quote:
1. Evaluate the possibility of simulating complex assembly without dynamic and teleological instructions.
2. Evaluating the possibility of simulating complex assembly without dynamic and teleological instructions that take new teleological forms during the assembly process.
3. Evaluating the possibility of simulating complex assembly without intelligent design processes.
4. Evaluating the possibility of simulating the evolution of any of the identified assembly instructions using only natural selection and random mutation.
I would like to note that all the items in the list involve ?evaluating the possibility of simulating? something.
It seems to me this formulation points to a problem with most ideas for ID-related research: there is no actual experiment proposed. An experiment involves making an hypothesis about the real world, gathering data about the real-word, and evaluating whether the data supports the hypothesis.
"Evaluating the possibility of simulating" something is not an experiment - it is a stretch, in my opinion, to even call it research. Actually simulating something can be a useful exercise, but even that is one step removed from actual research (as only real research can test whether the simulation was a valid model of reality.) "Evaluating the possibility of simulating" something is two large steps removed from real research.
If Warren wants to claim that his ideas are "experimentally verifiable," then he needs to describe an experiment that might verify them. Abstract attempts to "evaluate the possibility of simulating" certain processes does not verify anything.
[ 02. February 2003, 13:16: Message edited by: Evan ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 03. February 2003 08:11
Evan,
Quote: If Warren wants to claim that his ideas are "experimentally verifiable," then he needs to describe an experiment that might verify them
What I an describing is a very rigorous ‘standard’, ‘hypothesis testing’, scientific, experimental paradigm. You may not recognize this experimental paradigm because it comes from engineering and does not appear to be used widely in biology.
The engineering or reverse engineering paradigm is a techniques for evaluating the validity of mathematical models or theories of the operation of complex causal processes or machines. In order to apply the paradigm you need
1. A complex real world phenomena- in the application here biological self assembly
2. Mathematical techniques for precisely modeling the complex phenomena- as outlined, we have techniques for modeling, a)fixed assembly instructions, b)dynamic and teleological assembly instructions, c)intelligent design processes, and d)environmental conditions.
3. The ability to artificially simulate the process being analyzed and modeled. (There is extensive experience with simulating assembly or manufacturing processes both with and with the use of intelligent design processes). and
4. A objective, verifiable standard for evaluating and ranking the validity of models- in the example here, robustness of the simulation.
It will be noted that the experimental design being proposed is a multi-discipline approach as it draws on substantial existing bodies of knowledge - a)biology for knowledge of biological assembly processes, b)computer science- for techniques for modeling complex assembly and c)manufacturing/engineering for knowledge of assembly processes and the engineering paradigm. The feature of the proposed approach that might be considered ‘new’ is the recognition of the importance in assembly processes of mechanical intelligent design processes to produce dynamic and teleological assembly instructions.
To repeat, the ‘experimental design’ being discussed here is a standard, rigorous, scientific, engineering paradigm. In may not be readily recognized because the experimental design being proposed in based on techniques from several disciplines and because it involves techniques and concepts not routinely used in biology.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 03. February 2003 10:42
Given the nature of Evan’s comments, it will probably be useful to review some of the basic concepts of the experimental paradigm being discussed.
ROBUSTNESS- As used here, an assembly model and a simulation produced from a model is robust if the assembly being studied is actually produced by the simulation. Rex and I had a difference of opinion on constituted a causal model of an assembly process. Robustness provides an objective, verifiable, and rigorous criteria for differentiating a descriptive and a causal model. The specific criteria used is that an model or theory is valid is the simulation is at least as robust as the real world assembly process. Note that robustness can also be used to compare the relative strength or validity of different models and simulations.
ENVIRONEMENTAL CONDITIONS- As is fairly obvious, robustness varies with environmental conditions. A relatively simple set of assembly instructions can be robust under very rigidly controlled conditions. In general, the more volatile the environment, the greater the need for dynamic and teleological assembly instructions. As should be obvious, a simulation is robust or completely robust only if successful assembly is achieved under the same range of environmental conditions applicable to biological self assembly.
REDUCTIONISM- As should be obvious, our current knowledge of biological development processes and our current simulation technologies are inadequate to simulate the assembly of an entire organism. However, given current technology it is possible to isolate, model, and simulate relatively simple sub-components of a complex self assembly process. Based on the results obtained from analyzing sub-components, we can then use computers simulations to analyze the behavior of very complex assembly processes.
TELEOLOGY AND INTELLIGENT DESIGN PROCESSES- In biology, teleology and intelligent design processes seem to have taken on almost mystical or at least metaphysical meaning. Although the terminology may be different, in complex computer modeling and the analysis of assembly processes, scientific teleology and scientific intelligent design processes are real working concepts. An assembly instruction is dynamic and teleological if it can be modified into a form which increases the likelihood of a successful assembly. An intelligent design process is a process which modifies an assembly process in order to increase the likelihood of a successful assembly.
Intelligent design processes can be extremely complex. There are, however, some relatively simple, well known and highly effective intelligent design processes. The best known is the feed back loop which maintains equilibrium. An assembly instructions such as ‘turn on heat’ or ‘turn off heat’ can be continuously modified by a simple feed back loop that measures temperature.
A common problem in the analysis of biological design processes, is that observers often incorrectly assume that complex information processing must involve complex physical mechanisms. If, it is incorrectly assumed, a physical mechanisms is simple then the volume of information processing must be small. This is particularly true when talking about a physical mechanism everybody is familiar with. One of the advantages of defined techniques for measuring volumes of information and information generating capacity, is having objective standards for evaluating ‘simple mechanisms’.
I hope this background material will be helpful in understanding the experimental design being discussed.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 03. February 2003 13:04
warren,
Having read your posts in this thread and elsewhere, I have seen you make many claims about what a proper simulation of evolutionary processes would and would not show. I also see some specifications, vague though they sometimes are, for what you would consider a proper simulation model that would enable you to demonstrate that your "design science" alternative is preferable. Your readers have not actually seen such a simulation that actually exists in code and runs on real machines, though, and that's a serious problem for you. Absent an actual simulation model running on a machine somewhere, available for replicated tests of your claims, it is difficult or impossible to accept some of the claims you make.
As it happens, I have recently been looking over a program designed for research in artificial life that appears to have the capability of doing what you want a simulation to do. It is Avida, and it is freely available on the Web. I am not an expert on the program - I have been looking at it fairly seriously for only a few days, those days being interspersed with my regular work with GAs. But it appears to have the kind of capabilities that would allow you to demonstrate some of your claims about the limitations of 'orthodox' evolutionary processes.
Avida's "creatures" are small assembly language programs that self-reproduce. The assembly language of Avida is Turing complete, so the basic "physics and chemistry" of the program are quite flexible. (You can, of course, alter the instruction set to impose whatever limitations you deem appropriate.) The program distinguishes between genotypes (the list of assembly language instructions defining the genes of an individual creature) and phenotypes (the results of running the code embodied in the gene string) so you can study the mapping from genes to phenes.
One can run Avida in auto-adaptive mode or one can specify one or another sort of extrinsic fitness function so the program would allow you to create and impose fitness functions that meet the conditions that you say standard evolutionary processes can't cope with. It allows variable length creatures, so there is not a limitation there. It allows a range of kinds of mutations - point mutations, insertion and deletion mutations occurring as dividing errors (the creatures reproduce by fission), and copying errors - so the creatures are buffeted about in various ways as they attempt to reproduce themselves. Avida enables local interactions among creatures, both intra-species and extra-species, and also allows parasitism, all of which would allow you to test the effects of competition and co-evolution on your claims about information processing in evolution.
Apropos of information processing, the program automatically records the changing (through generations) Shannon entropy at the population, species, and creature levels, so you can directly track changes in information over cycles/generations at several levels of analysis. In addition, you can optionally require the program to write various other characteristics of the population to disk (all the way down to recording the composition of every genome that occurs) in great detail at regular intervals, so post-run analyses can be performed on virtually any property of the evolving population.
I strongly suggest that rather than talking about what a simulation would or would not show, you obtain Avida and actually do the demonstrations you have discussed only in the abstract. That would settle the issues you raise with some definitiveness, and would allow your critics to actually see your points demonstrated (and they could actually replicate your research) in a real simulation as opposed to an imaginary one. You would be providing operational definitions of your claims embodied in a real running simulation model, as opposed to hypothetical claims based on an imaginary simulation model. Several of the people who have discussed these matters with you actually work with real evolutionary systems (i.e., they are biologists) and/or they work with evolutionary algorithms that incorporate and model biological processes. We often find it very hard to believe your claims about what evolutionary processes can and cannot do, so it's up to you to provide some data, some actual results from real experiments run in an implemented model, to convince us that your approach is worth pursuing.
RBH
[ 03. February 2003, 13:15: Message edited by: RBH ]
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 03. February 2003 16:51
quote:
A set of self instructions is complete or ‘causal’ if and only if the assembly instructions, applied in defined environment can actually produce the defined result.
What do you mean by "defined result"? Green fluorescent protein is greenly fluorescent regardless of whether the second codon is GTA, GTC, GTG, or GTT, since they all code for valine.
It is certainly not the case that you get GFP iff you have a single DNA sequence.
quote:
An assembly instruction is defined as ‘dynamic and teleological’ or ‘programmable’ if 1)the instruction can take many different forms, and 2)only a portion of the possible forms are compatible with the defined final assembly objective.
I'm not quite sure how one would rigorously define the "assembly objective". But, anyway, there are 64 possible codons, of which only 4 code for valine, so trivially protein translation is 'dynamic and teleological'.
I'm not sure this is at all instructive, though. I was well aware of this before. Throwing 'teleological' in just makes me think of philosophers, and adds nothing to my understanding of the science or mechanism.
quote:
As defined here, an intelligent design process is a complex set of operations capable of 1)finding the teleological form of a causal relationship such as an assembly instruction, 2) adding to the set of possible instructions by adding to the lists of recognized inputs, stimuli or causes, 3)adding to the set of possible outputs, responses or effects (capabilities 1, 2, and 3 make it possible to produce creative teleological assembly instructions), 4)creating new instructions, and 5)increasing the processing capacity of the design process.
I have no idea what (1) means.
(2) is easily done via RM&NS--antibodies do this all the time.
(3) is easily done via RM&NS--any alteration in regulatory control fits this description.
(4) if a protein counts as an instruction, this is trivially done by RM&NS (specifically via gene duplication).
(5) I'm not sure what you mean here. Does it count if you make more ribosomes so you can make proteins faster? Or if you make more organisms, so you have more opportunties for mutation and selection?
I'm afraid I am having trouble distinguishing your intelligent design process from the known properties of genetic material, transcription, and translation.
If your mapping was not bijective but instead was only onto, and the claim was that f(x)->{y} out of which very few z in {y} were "good", yet those z were reliably achieved anyway, then there might be a distinction. I don't think there is any evidence that this happens (despite your claim that it must), but this may be the way you want to set up the problem.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 04. February 2003 10:36
Rex and RBH,
Based on the feedback provided, I realize we need to go back and review some basic concepts.
To begin, we are in agreement that there is a descriptive causal relationship between ‘fertilized egg cells’ and ‘full grown organism’. We know from observation that there typically occurs a transformation from egg to mature organism. The purpose of the experimental paradigm proposed here is to evaluate the causal nature of the transformation. In more formal terms- is the observed transformation a simple ‘information neutral’ transformation as suggested by Darwinian concepts, or does the transformation require the generation of large quantities of information and ‘intelligent design processes’ as I am proposing.
To test between the two concepts, I am suggesting that the egg to organism transformation be viewed as an automated assembly or manufacturing process. Viewed from this perspective, we can consider the possibility of building an automated machine to simulate the transformation, manufacture or assembly from egg or full grown organism. By evaluating the types of instructions and information processing required for such a machine, I suggest, we will be able to answer the question about the type of transformation involved in biological development processes. I am in effect suggesting we use engineering techniques to reverse engineer the biological developmental process.
To understand what is being proposed here, you have to understand -1)the basic change of state or assembly instruction approach to modeling automated manufacture, 2) the current understanding/appreciation of the complexity of automated assembly, and 3)engineering ability to perform analysis based on incomplete knowledge.
ASSEMBLY INSTRUCTIONS- There are probably many different techniques for modeling assembly instructions. The approach used here is based on the ‘change of state’ concept. A manufacturing or assembly device is viewed as a closed device in a steady or stable state. For manufacture or assembly to occur, there must be an ongoing set of changes in the states or conditions of the system. It is not possible to model all the changes occurring in the assembly mechanism during the assembly process. It is, however, possible to model and simulate automated assembly processes using a discrete set of ‘change of state’ or assembly instructions. Each assembly instruction or change of state instruction involves 1)an input or trigger or stimulus and 2)an output or response which produces a change of state in the manufacture environment.
Consider as a simple example an automated device for baking or ‘assembling’ cookies from already assembled cookie dough. The simplest set of assembly instructions would consist of ‘turn on oven(reaction) at time t=0(stimulus)’ and ‘Turn off stove at time t=k’. In practice, we find such a set of instructions inadequate to produce properly baked cookies. Generally we will also need a set of instructions operating between time t=0 and t=k of the form "if temp greater than g and oven on, then turn oven off’ and ‘if temp less than h and stove off, then turn stove on’. In practice, it is found that except in a very controlled environments, even this set of instructions are not adequate. To consistently bake cookies properly, you need some type of instruction which can measure and react to ‘the cookies have reached the appropriate baked state’.
The techniques I defined for measuring the quantify of information, make it possible to evaluate both the complexity of the change of state instructions, and the volume of processing required to find the proper instruction.
KNOWLEDGE OF AUTOMATED ASSEMBLY- For those of you old enough to remember, when computers were first developed, lots of people thought it would be very easy to automate all the simple, mindless tasks performed by ‘low level employees’. People who worked with this type of problem quickly found that there was a big difference between what was simple and what appeared simple.
We know a great deal today about how to automate processes and the long complex line of instructions required to automate even simple assembly processes. It is my understanding that we have a rather extensive knowledge of some of the key processes associated with biological development. From our study of automating ‘simple’ processes, we have a reasonable understanding of the assembly instructions which would be required to automate the known developmental processes( or something logically similar to the known biological process). The experimental paradigm I am proposing would combine our knowledge of biological processes with our knowledge of automated assembly.
ENGINEERING AND PARTIAL KNOWLEDGE- It is useful to note that engineering analysis is based on partial or incomplete knowledge. Engineers are able to determine with a reasonable degree of accuracy that a certain design or approach will work or will not work by using mathematical or paper and pencil analysis applied to components of the overall design or model. With a reasonably high degree of accuracy, engineer techniques make it possible to predict whether a particular bridge design will or will not work. Engineers don’t actually build a bridge and watch it collapse to know that certain types of bridge design won’t work.
The goal of the proposed experimental paradigm is not to build a machine to assemble full grown organisms starting with the information contained in an egg cell. The purpose of the analysis being proposed is to determine first, what type and volume of information, as defined by assembly instructions would be required to transform or simulate the transformation from the egg to the organism. The second objective is to determine what type of information generating process would have the capacity to generate the information required for the ‘simple’ gene to organism transformation.
I am predicting that very early on in the proposed analysis, the ‘development based on information stored in genes’ or ‘development based on information neutral transformation’ designs or explanations will be shown to be unworkable. This conclusion, I predict, can be reached based on existing knowledge of biological development and automated assembly. [These predictions are not attempts to convince you of the conclusion I expect, but to give you a flavor of the direction this analysis is going.]
Now to some of the specific issues raised. Avida like most GA systems ASSUMES the existence of a simple genotype to phenotype map. For biological systems, we know that no such simple map exists. We, in fact, have no evidence that any functional map simple or complex exists. While it is probably theoretically possible to modify Avida to deal with realistic assumptions, it is certainly not practical. As should be apparent, Avida and GA systems in general would not be useful for the analysis being performed here without very, very, major enhancements.
Terminology involving teleology and intelligent design processes may be not be very helpful in discussing Darwinian concepts. The terminology is, however, both useful and appropriate in addressing automated assembly. Purposeful or teleological assembly instructions of the type ‘generate action X if goal or sub goal (stimulus) G is (or is not) achieved’ are very useful, and in fact essential, in most complex automated assembly processes. As may or may not be obvious, the ‘evolution’ of such clearly teleological or purposeful instructions poses some interesting logical/mathematical challenges.
If an automated assembly process is designed to produce a full grown organism, a protein, or properly baked cookies, it is relatively easy to precisely define, measure and/or determine if the automated assembly process in fact succeeded in producing the full grown organism, the protein or the properly baked cookie.
Rex- You appear to be arguing that
1) There must be a causal relationship between genetic codes and the full grown organism because under normal conditions one normally follows the other. And
2) If there is a causal relationship then that causal relationship must have the form of a one to one correspondence or mapping between genotypes and phenotypes and phenotypes to genotypes.
This logic is fundamentally flawed. There are causal relationships other than ‘functional mappings’ which can explain the observed pattern. The experimental design proposed here is designed to demonstrate that 1)the one to one mapping type of explanation does not fit the data and 2)there are other types of complex causal relationships which will fit the observed data.
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 04. February 2003 11:12
Warren states:
quote:
This logic is fundamentally flawed. There are causal relationships other than ‘functional mappings’ which can explain the observed pattern. The experimental design proposed here is designed to demonstrate that 1)the one to one mapping type of explanation does not fit the data and 2)there are other types of complex causal relationships which will fit the observed data.
You seem to be jumping to conclusions before the experimental data has even been collected. I would be interested to see when you have performed your experiments, if the data still support your conclusions.
I doubt btw that Rex is arguing a one-to-one mapping of genotype to phenotype and vice versa. Rex is merely pointing out that RM&NS maps quite nicely to your definition of intelligent design process.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 04. February 2003 13:18
In the OP warren wrote quote:
The purpose of this thread is to discuss the experimentally verifiable hypothesis that "biological development is a biological design process which can not occur without the presence of information generating 'intelligent design processes'". The hypothesized existence of scientifically measurable and verifiable intelligent design processes as part of biological development processes would appear to contradict most of the widely held beliefs in both biology and ID. The fact that the hypothesis is unconventional should not make it an inappropriate topic for discussion here. (emphasis added)
My mention of Avida, with its distinction between genotypes and phenotypes, and with the ability to closely analyze the developmental mapping between them, was intended to move toward the "experimentally verifiable hypothesis" warren writes of in his first sentence. Rather than merely discuss it in an empirical vacuum, as we have so many such ideas, I would strongly prefer to see some systematically gathered data to be able to evaluate the degree and force of the experimental verification.
In the OP warren specifies "four types of experimental design," all requiring a simulation model to execute. The simulation model has to come from somewhere, and until it exists and is running, discussion of the proposed hypothesis is unlikely to be fruitful. Then one might experimentally test the hypothesis; whether it would be verified is an empirical question, not an a priori given.
RBH
Edited to note that I added the emphasis to the quotation from warren's OP.
[ 04. February 2003, 15:11: Message edited by: RBH ]
IP: Logged
|
|
Evan
Member
Member # 164
|
posted 04. February 2003 14:05
RBH has found Avida, and listed its many features. Warren says that Avida would need "many enhancements" to model his view of the design process.
Warren also, however, makes it clear that simulating the design process is what he is interested in.
So my questions to Warren are:
1) what specific enhancements would need to be made to Avida (both additions and subtractions) to make an accurate model of your concepts? - and
2) what parts of Avida would be common to both the existing program and a program that simulates your version of the design process?
Please be specific. I don't think we need any further review of the concepts.
[ 04. February 2003, 14:06: Message edited by: Evan ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 05. February 2003 09:45
Francis- Common interpretations of neo-Darwin concepts suggest the existence of both a ‘one to many genotype to phenotype map’ and a ‘1 to many phenotype to genotype map’. This would typically be interpreted to suggest a one to one genotype to phenotype map if you take neo-Darwinian concepts literally. If neo-Darwinian concepts are simply metaphysical or descriptive or non-causal, then of course mathematical principles need not apply.
RBH-
Quote: My mention of Avida, with its distinction between genotypes and phenotypes, and with the ability to closely analyze the developmental mapping between them, was intended to move toward the "experimentally verifiable hypothesis"
Avida ‘assumes’ a simple one to one mapping from simple code to simple phenotypes. Hardly an ability to closely analyze developmental mapping.
Quote: Rather than merely discuss it in an empirical vacuum, as we have so many such ideas, I would strongly prefer to see some systematically gathered data to be able to evaluate the degree and force of the experimental verification.
While your sentiments may be admirable, two point need to be kept in mind. First, it must be recognized that ‘existing genetics and evolutionary biology’ can not generate testable, predictive models of evolutionary processes. It is possible to generate models which appear to simulate some small components of evolutionary change under limited and highly dubious assumptions. Proposing techniques to produce predictive models and testable simulations is something new in evolutionary biology.
Second, the models and simulations being proposed are based on engineering standards and concepts. Not the standards and concepts typically associated with theoretical science. A biological system is viewed here as a complex machine. The experimental paradigm being proposed is designed 1)to test the adequacy or inadequacy of certain types or classes of models or simulations and 2)to develop by successive approximation models and simulations of how the biological machines actually function. Using engineering standards, a model of evolution is not a one step pass fail process. On a gradual basis we develop knowledge of 1)what doesn’t work and 2)what actually does or could work.
As we start this gradual engineering process, we do not need to gather new information so much as analyze and interpret the existing body of knowledge.
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 05. February 2003 12:41
Dear Warren
I fail to see how you conclude that from a one-to-many and many-to-one mappings one would 'typically interpret' to suggest a one genotype to one phenotype map. Perhaps you could explain to me the mathematical foundations for such a conclusion?
Are you also saying that one-to-many or many-to-one mappings cannot be mathematical?
Please explain these 'common interpretations' since they seem to be all but.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 05. February 2003 14:05
To Francis warren wrote quote:
Common interpretations of neo-Darwin concepts suggest the existence of both a 'one to many genotype to phenotype map' and a '1 to many phenotype to genotype map'. This would typically be interpreted to suggest a one to one genotype to phenotype map if you take neo-Darwinian concepts literally.
That is simply incoherent. Do you really imagine that biological scientists are so blind as to interpret some relationship that is many-to-one in both directions as being "typically" (your word) interpreted to suggest a one to one map? I think you need to provide some real strong support for that assertion. If it is typical, as you claim, it shouldn't be hard for you to supply references to the professional literature that illustrate your claim. Where can we read that claim in the biological literature? Specifically which "neo-Darwinian concepts" imply that interpretation, what is your source for those "concepts," and how is the interpretation drawn from those "concepts"?
warren wrote quote:
Avida 'assumes' a simple one to one mapping from simple code to simple phenotypes.
Perhaps you'd be good enough to show me where in the source code (available on the site referenced above) that mapping is assumed. This morning I've been doing an analysis of an Avida run in which there is a many to one mapping from gene strings (creature code) to species of artificial creatures (groups of morphologically - in Avida's terms - indistinguishable phenotypes).
warren wrote quote:
While your sentiments may be admirable, two point need to be kept in mind. First, it must be recognized that 'existing genetics and evolutionary biology' can not generate testable, predictive models of evolutionary processes. It is possible to generate models which appear to simulate some small components of evolutionary change under limited and highly dubious assumptions. Proposing techniques to produce predictive models and testable simulations is something new in evolutionary biology.
Your first point is irrelevant, since you yourself claimed in the OP that quote:
The purpose of this thread is to discuss the experimentally verifiable hypothesis that "biological development is a biological design process which can not occur without the presence of information generating 'intelligent design processes'". (emphasis added)
The request is for you to support your claim of experimental "verifiability" for your allegedly superior formal mathematical simulation model. You say it's experimentally verifiable. Where is the verification?
Your claim that "techniques to produce predictive models and testable simulations" are "something new in biology" is flatly false. Predictive models and testable simulations are not new in biology, and to assert that they are requires that you ignore a large and growing body of literature.
Finally, warren wrote quote:
As we start this gradual engineering process, we do not need to gather new information so much as analyze and interpret the existing body of knowledge.
To analyze and interpret the "existing body of knowledge" one must know that body of knowledge and provide an accurate description of it. The new model must both account for existing knowledge and data and provide novel predictions of new observations. The former requires an informed description of the already known biological phenomena that your model purportedly accounts for and how it does so.
A prerequisite to analyzing and interpreting existing knowledge and data is accurately describing it, and that requires describing the knowledge and data as it has been published by the proponents of the old model. That means citations to their published literature, not general statements about casual inferences from a superficial description of the old model. As I've said before, in order to criticize a model one must know it well. I'm not real interested in a discussion in an empirical vacuum.
A few months ago, in response to your remark that you didn't have access to it, I provided you with the URL of PubMed, the publicly available search facility for the most extensive database of professional published work in biology and allied disciplines out there. I suggest you use it to attempt to provide systematic support for your claims about current biology.
RBH
[ 05. February 2003, 14:10: Message edited by: RBH ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 05. February 2003 15:49
Remember that Warren claims that a single genetic code does not map to a single phenotype. I think this is the basis of the issue as Warren brings up. (Of course supporting his claim is a separate issue -- but Warren's comments seem to be based on this).
Without looking at an evolutionary algorithm in detail (Avita) I can't say for sure, but I would assume that the algorithm does not also simulate develoopmental processes that (Warren claims) could change the "phenotype" after the establishment of the genetic code. Does it do so?
(I'm just trying to establish the basis of what Warren is saying -- not commenting on its validity.)
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 05. February 2003 16:53
Gedanken wrote quote:
Remember that Warren claims that a single genetic code does not map to a single phenotype. I think this is the basis of the issue as Warren brings up. (Of course supporting his claim is a separate issue -- but Warren's comments seem to be based on this).
Without looking at an evolutionary algorithm in detail (Avita) I can't say for sure, but I would assume that the algorithm does not also simulate develoopmental processes that (Warren claims) could change the "phenotype" after the establishment of the genetic code. Does it do so?
Ahhh. Well, if that's what warren is saying I'm just going to watch a while, until it's clearer how he plans to implement his model in order to experimentally verify it. As I understand Avida so far, it does not implement developmental variability in the mapping from gene to phene. I thought one question was whether there was a many to one mapping from gene strings to expressed phenotypes, which there is in Avida. In other words, several different genotypes can produce the same phenotype (to which the selective pressure applies), where "phenotype" is defined as the expressed behavior of the code. "Species" in Avida are typically (rather than atypically) composed of bunches of individual critters with several different genotypes. Avida's critters can (among other things) carry "junk" code along with the executing code. (Added in edit: I should add that a "gene" in Avida is not a single assembly language instruction; it is a sequence of instructions that execute some function, such as copy self. Individual instructions are more analogous to bases or perhaps DNA triplets.)
By the way, on another thread you were trying to convince warren that populations can adapt to average trends underlying high-frequency variability. This paper that I ran onto in the course of a search for something else speaks directly to that issue. quote:
Dynamic Fitness Landscapes in Molecular Evolution
Authors: Claus O. Wilke (1), Christopher Ronnewinkel (2), Thomas Martinetz (2)
((1) Caltech (2) Medizinische Universitaet zu Luebeck)
Comments: LaTeX, 60 pages, 14 eps figures, expanded introduction, minor corrections, submitted to Phys. Rep
Subj-class: Biological Physics; Adaptation and Self-Organizing Systems; Soft Condensed Matter; Statistical Mechanics
Journal-ref: Phys. Rep. 349:395-446 (2001)
We study self-replicating molecules under externally varying conditions. Changing conditions such as temperature variations and/or alterations in the environment's resource composition lead to both non-constant replication and decay rates of the molecules. In general, therefore, molecular evolution takes place in a dynamic rather than a static fitness landscape. We incorporate dynamic replication and decay rates into the standard quasispecies theory of molecular evolution, and show that for periodic time-dependencies, a system of evolving molecules enters a limit cycle for $t\to\infty$. For fast periodic changes, we show that molecules adapt to the time-averaged fitness landscape, whereas for slow changes they track the variations in the landscape arbitrarily closely. We derive a general approximation method that allows us to calculate the attractor of time-periodic landscapes, and demonstrate using several examples that the results of the approximation and the limiting cases of very slow and very fast changes are in perfect agreement. We also discuss landscapes with arbitrary time dependencies, and show that very fast changes again lead to a system that adapts to the time-averaged landscape. Finally, we analyze the dynamics of a finite population of molecules in a dynamic landscape, and discuss its relation to the infinite population limit. (emphasis added)
I commend it to warren's attention both because it shows adaptation to time-averaged high frequency environmental changes, and because it is an example of formal mathematical modeling in evolutionary research.
RBH
[ 05. February 2003, 17:16: Message edited by: RBH ]
IP: Logged
|
|
|
Topic Closed
Printer-friendly view of this topic
