|
Author
|
Topic: Evaluation of neo-Darwinian Theory with Avida Simulations
|
Moderator
Administrator
Member # 1
|
posted 01. July 2004 14:23
Evaluation of neo-Darwinian Theory with Avida Simulations
*Note that an earlier version of this paper was posted at ISCID on January 23rd, 2004.
by Royal Truman
Abstract: Avida is a computer software platform designed to test neo-Darwinian scenarios including speciation, evolution of novel complex functions, and genetic drift. A large number of parameter settings are available to the researcher. We evaluate here published tests and intrinsic Avida characteristics in an attempt to extrapolate towards real biological organisms. Computer based models and simulations must capture enough realism to facilitate valid decisions. Neglect of key factors or unrealistic parameter settings permit conclusions to be claimed which merely reflect what the decision maker intended a priori.
The long-term effect on genetic systems of random insertions, deletions and point mutations over deep time may be degradation of function and specificity; or the generation of true biological novelty. Competing factors are involved and only detailed quantitative analysis with real world experiments can firm up our opinions in this matter. In Part 1 we examine some of these issues and in Part 2 explore some more quantitative details based on Avida runs.
It was found in this two part series that reports which demonstrate rapid increase in complex, genetically coded novel functions are an artefact intrinsic to how the Avida computer runs are planned and some neglected details constraints found in nature.
Using biologically realistic parameter settings and including neglected factors led to the opposite view, both theoretically and when using Avida experiments. Mutations plus natural selection in rapidly reproducing, asexual, small genomes fails to explain the original of ever increasing cellular functionality, including molecular machines.
Read part 1 of this essay.
Read part 2 of this essay.
[ 01. July 2004, 14:23: Message edited by: Moderator ]
IP: Logged
|
|
|
|
|
|
Salvador T. Cordova
Member
Member # 959
|
posted 09. July 2004 03:19
Greetings Dr. Truman,
I apologize for my delay on my promise to assist in evaluating Avida last January as I was tied up at ARN in some very difficult and technical threads for several months.
Recently, of all things I had gotten into a discussion at ARN with RBH (a reviewer of your earlier paper) at ARN. It is now the featured thread at ARN. I would not necessarily say the discussion was as scholarly at ARN as it would be here, and I said things in ways that I regret, however in terms of technical content some things of interest came out that may have relevance to your paper.
I independently confirm your statement by having done some experiments:
quote:
(3) Critically important functions for “life” are difficult to destroy by mutations • All functions necessary for an organism to replicate and survive are carried out initially by only 15 instructions. The physical machinery to transcribe, translate and perform metabolic processes are not coded genetically and thus not subject to mutational damage.
I made an Avida run with Cosmic Radiation mutation set to 50% per codon per update. That would be enough to obliterate DNA and everything else several times over in a real organism. To actually quatify what that would mean in physical terms would be hard, but figuratively speaking it would be like putting a human in a 10,000 watt microwave oven for a few days and then asking him to try to reproduce shortly thereafter. ![[Eek!]](eek.gif)
Statisically speaking with a generation time in Avida requiring 500 updates at a 50% per codon per update probability of Cosmic Ray point mutation, my figurative description is quite reasonable. Yet the creatures survived that level of Cosmic bombardment, never died, and the population actually expanded. There is an entropy containment strategy in Avida that guarantees critical information will never be lost. I
Avida circumvents critical considerations with respect to the propagation of destructive entropy.
Also, I would advise, and this is coming from a friendly perspective, that some minor spelling and grammar changes be made. I am not aware of how I can communicate my suggested grammar and spelling changes to your otherwise excellent paper. Let me know how I may communicate these changes, and I'd be glad to help do that.
Cordially,
Salvador
PS
After some foibles on the ARN featured thread that discusses Spiegelman and Avida, my more recent postings I feel have done a better job of representing your work and the scientific basis for Intelligent Design. I apologize for not doing so earlier in that discussion, but I feel I am not making amends for my earlier mistakes.
[ 09. July 2004, 13:53: Message edited by: Salvador T. Cordova ]
IP: Logged
|
|
Royal
Member
Member # 1060
|
posted 11. July 2004 15:25
[Salvador]: quote:
I made an Avida run with Cosmic Radiation mutation set to 50% per codon per update. That would be enough to obliterate DNA and everything else several times over in a real organism.
Statisically speaking with a generation time in Avida requiring 500 updates at a 50% per codon per update probability of Cosmic Ray point mutation, my figurative description is quite reasonable. Yet the creatures survived that level of Cosmic bombardment, never died, and the population actually expanded.
[Royal]: Excellent observation, Salvador. Extrapolating to the smallest of bacteria illustrates the problem. For example, C++ programs simply read and interpret instructions from Avida genomes. In the real biological world, translating mRNA alone requires on the order of 100 different kinds of macromolecules: proteins (aminoacyl tRNA synthetases, about 55 proteins for ribosomes, etc.) and RNA polymers (rRNA, tRNA, mRNA). These are all coded for by genes. Imagine using 50%, or 5% or even 0.05% mutational rates per codon per update on real, free-living organisms!
Avida logic functions are very easy to generate by chance. The closest analogy in biology to a logic function would be a novel metabolic network (a gene alone can’t do anything). These involve many enzymes with feedback inhibitation for control purposes. The network must deactivate the inhibitation itself when necessary.
Until about 10 proteins or so are properly synchronized natural selection would be negative: the cell would maintain unnecessary DNA, and waste energy and material producing worthless mRNA and polypeptides non-stop. The streamlined lineages would easily outreproduce these.
[Salvador]:
quote:
After some foibles on the ARN featured thread that discusses Spiegelman and Avida…
[Royal]: As I point out in Part 2, the leading Avida names also discusss Spiegelman, and have published the clear fact that massive genome truncation is to be expected. In fact, reproductive rate was estimated to be inversly proportional to genome length, ceteris paribus. It was precisely to permit any kind of evolution that this fact of nature has to be deactivated in Avida runs.
Does anyone know how the authors actually did these runs? How was fitness set to inverse genome length?
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 11. July 2004 16:00
Salvador wrote quote:
I made an Avida run with Cosmic Radiation mutation set to 50% per codon per update. That would be enough to obliterate DNA and everything else several times over in a real organism.
Statisically speaking with a generation time in Avida requiring 500 updates at a 50% per codon per update probability of Cosmic Ray point mutation, my figurative description is quite reasonable. Yet the creatures survived that level of Cosmic bombardment, never died, and the population actually expanded.
Again I ask, as I did on ARN: What specific values did Salvador assign to what specific variables?
Truman asked quote:
As I point out in Part 2, the leading Avida names also discusss Spiegelman, and have published the clear fact that massive genome truncation is to be expected. In fact, reproductive rate was estimated to be inversly proportional to genome length , ceteris paribus. It was precisely to permit any kind of evolution that this fact of nature has to be deactivated in Avida runs.
Does anyone know how the authors actually did these runs? How was fitness set to inverse genome length?
How about a reference. I don't want to fight my way through the pdf again hunting for what Truman means here.
RBH
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 12. July 2004 14:30
I have a couple of questions regarding fitness and simulation--but first I would note that it would be easier to read Part II if the figues and tables were actually in the main text instead of as twenty separate links.
The first question is: what fraction of an organism's metabolic output goes into replicating its DNA? Relatedly, what is the selective advantage as a function of relative genome length? (The answer obviously varies widely with the type of organism you're using, so this is largely a question about the type of organisms that one believes Avida models digitally.)
The second question is more technical: suppose you start out with the initial critters with, let's say, 15 replication instructions and 5 random instructions. What is the selective advantage of critters that have a NAND or NOT at random (using the original reward values of...2 and 4? I forget), over those who have no output function? I think that Truman has misinterpreted the fitness rewards, but that's easiest to determine if someone (e.g. Truman) actually spells out what they think happens.
IP: Logged
|
|
Salvador T. Cordova
Member
Member # 959
|
posted 12. July 2004 14:47
In response to RBH's request.
Take the baseline Avida download for 1.3.
Make the following modification in the Genesis file:
POINT_MUT_PROB 0.5
The authors elaborated in 2.0 (presumably the same meaning in 1.3) at: Explanation of 2.0 Genesis File
quote:
Point mutations (sometimes referred to as "cosmic ray" mutations ) occur every update; the rate set here is a probability for each site that it will be mutated each update
I highlight the fact that I deliberately deviated from the authors recommendation:
quote:
In other words, this should be a very low value if it is turned on at all.
The reason for my experiment, deviating from the author's recommended setting was to highlight by way of extreme demonstration the immunity Avida grants to critical life functions. In no way was I trying to model evolution, I was trying to highlight a flaw in the way Avida models physical reality.
Avida allows no propagation of point mutations to critical life functions, neither does it grant propagation to copy, divide, insert, delete mutations to critical life functions. The critical life functions are absolutely immune to modification while the "Avida genome" mutates away.
Salvador
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 12. July 2004 15:19
Salvador blew it. See here.
RBH
IP: Logged
|
|
Salvador T. Cordova
Member
Member # 959
|
posted 12. July 2004 16:29
RBH linked to something he wrote at ARN, I encourage reading of my responses there.
There was the question of the meaning of the setting for POINT_MUT_PROB. At question is, should a multiplier of 10^-6 be presumed when setting POINT_MUT_PROB. The Avida documentation for version 1.3 is very sketchy and misleading.
Thus RBH goes on the presumption of mulitplying POINT_MUT_PROB by 10^-6 to get a probability of point mutation. That however leads to interesting results, and if true, Avida 1.3 has bugs.
I report the following (and it was also at the ARN thread):
Avida with setting to 50000.00 for point mutation enabled poplulation growth. Settings of 5,50,500,5000, and 500000 locked the system.
In contrast, 0.5 and 50000 seem to work and allow population growth. There is an apparent inconsistency in the software itself as numbers like 5,500,500000 locked the system but not 50000 (fifty thousand).
There is another aspect why these result don't make any sense: 50,000 is 100,000 larger than 0.5. The poplulation growth was faster with POINT_MUT_PROB set to 50000.00 than 0.5. This does not make any sense. In addition to the considerations in Royal's paper, now there is the distinct possibitly Avida may have a software bug.
For pure discussion in the run with the .5 setting. I got some of the following results:
Consider the two genomes, the ancestor named 020-aaaaa (presumably 020-aaaaa, where 20 means how many instructions) and some descendent named 026-aaaaa (or maybe the same creature, but just morphed) after 396 updates (presumably 026-aaaaa, where 26 means how many instructions).
It is not clear if "aaaaa" refers to the same creature, if so, then "aaaaa" morphed by 95% . Otherwise the descendent evolved 95% in only 396 updates. The 95% value is justified by the table below.
Either interpretation conveys that POINT_MUT_PROB set to 0.5 is a very high mutation probability.
code:
020-aaaaa 026-aaaaa comparison outcome
search-f copy divergence
nop-A allocate divergence
nop-A jump-f divergence
add dec divergence
inc search-f divergence
allocate allocate same
push swap divergence
nop-B jump-f divergence
pop put divergence
nop-C nop-A divergence
sub allocate divergence
nop-B call divergence
copy nop-A divergence
inc pop divergence
if-n-equ get divergence
jump-b add divergence
nop-A sub divergence
divide if-bit-1 divergence
nop-B inc divergence
nop-B sub divergence
dec
put
swap
divide
nop-B
get
We see 95% divergence in only 396 updates. If 0.5 means a mutation rate of 0.00005% rather than 50%, then the above sequence divergence is not consistent with that interpretation. Frankly, the Avida authors should have been clearer about what they meant.
Given the organisms aren't reproducing very much, the other mutation sources are mostly ruled out, although we can try zeroing those mutation sources out to see what happens.
Salvador
[ 13. July 2004, 17:28: Message edited by: Salvador T. Cordova ]
IP: Logged
|
|
Salvador T. Cordova
Member
Member # 959
|
posted 12. July 2004 18:02
Royal,
This is your thread and about your paper. I don't necessarily want all of the ARN debate over Avida to spill over here. Thus, with respect to the point mutation issue, I only offer it as an independent support for your claims.
If there are any reasons for me to offer material ammendments or reversals to my independent confirmation of your suggestion that Avida immunizes critical life functions from harmful mutation, I will appraise you of it here in this thread.
I am preparing my recommendations for changes in grammar and spelling to your paper and will be sending them to you via the private channel you offered.
I will continue reviewing your paper as it is a focal point in my ARN debate and of interest generally with regard to modelling of evolutionary biology. The ARN debate has looser rules than ISCID, it would not be completely appropriate for ISCID, but any developments of a purely technical nature in that debate relevant to your paper I will post here.
I expect the majority of my technical contributions to this thread will be independent confirmations of your paper's findings.
I have made identification of tautological flaws in Avida, particularly the way avida rewards emergence of novel complexity, and that is in independent agreement with your paper's findings. I will post those as well here.
Cordially,
Salvador
PS
I sent you an e-mail today, 7/12/2004, with my new e-mail address.
[ 12. July 2004, 21:07: Message edited by: Salvador T. Cordova ]
IP: Logged
|
|
Salvador T. Cordova
Member
Member # 959
|
posted 13. July 2004 15:19
For control purposes of duplicating any experiments, please note the version number of Avida being run. At this time, Avida 2.0 is unstable and buggy (RBH concurs) and therefore many of my runs are done with Avida 1.3
When I began fiddling with Avida 2.0 Beta 6, it bombed when I tried to set cosmic ray mutation to 0.5.
Given my previous posts, it is possible the authors of Avida have potentially published findings with faulty software. The bugs don't necessarily invalidate their numbers (as some of those parameters are not even used by most people), but such problems do not instill confidence in the Avida software.
Salvador
IP: Logged
|
|
Salvador T. Cordova
Member
Member # 959
|
posted 13. July 2004 18:23
Royal,
With regards to page 5 you wrote:
quote:
For realistic Avida simulations, generation times must be inversely proportional to amount of superfluous genetic material, ceteris paribus, at least until an advantageous logic function evolved. The authors permitted deletions, but fail to provide any reward when they occur: generation times would be shortened; and less (biological) energy and material (nucleic and amino acids to produce proteins) would be needed. Some Avida researchers seem to be aware of this fact, as further discussed in Part 2, but a parameter is not provided in the genesis file to include this effect during simulations.
The bolded is not in my opinion entirely correct. I could be wrong, but I urge caution.
I would plead your review of this.
In a cross verified experiment which I participated in, RBH provided a genome with mostly nop-instructions. It looked like:
quote:
search-f # calc Size
nop-A
nop-A
add
inc
allocate
push # Move size into CX and clear BX
nop-B
pop
nop-C
sub
nop-B # Copy Loop
copy
inc
if-n-equ
jump-b
nop-A
divide # And divide...
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
nop-B
The creature de-volved in about 10 generations, losing its superfluous genetic material, and then evolved upward with novel complexity. Excess metabolic load was indeed penalized and improved metabolic fitness was rewarded. This was even under the condition of SIZE_MERIT_METHOD = 4, where shorter genomes still have a natural advantage over longer genomes, but not quite as severe as would be the case if SIZE_MERIT_METHOD = 0.
RBH elucidated an opposing view point which I bring to your attention as I think it warrants your consideration.
Please see : Devolution then upward evolution example
quote:
Figure 3 illustrates how rapidly the population would shift from 1% to >99.9% members with 10 fewer codons: in just 70 generations.
Your theoretical calculation was actually in agreement with Avida behavior in that case, in my opinion, so I would plead reconsideration.
Salvador
[ 15. July 2004, 00:56: Message edited by: Salvador T. Cordova ]
IP: Logged
|
|
Royal
Member
Member # 1060
|
posted 16. July 2004 11:59
[Salvador]:
quote:
In a cross verified experiment which I participated in, RBH provided a genome with mostly nop-instructions… The creature de-volved in about 10 generations, losing its superfluous genetic material, and then evolved upward with novel complexity. Excess metabolic load was indeed penalized and improved metabolic fitness was rewarded. This was even under the condition of SIZE_MERIT_METHOD = 4
[Royal]:
In principle, Avida setting SIZE_MERIT_METHOD = 4 in the genesis file should lead to a slow, average streamlining in genome length (unless compensated for by novel rewarded logic functions).
Unnecessary instructions after the last one required to define a replication don’t count in the case of option 4 when calculating the fitness (SIPs) of that lineage. Theoretically the shorter genomes should reproduce slightly faster due to this effect.
I hope to comment later of professor RBH’s run, in which about 35 nop-B’s were removed by natural selection in about 10 generations. I have performed hundreds of Avida runs, and am now systematically examining this issue using different RANDOM_SEED values.
Here is what I am finding: in the absence of any logic function rewards, contra expectations, the genome size starts to increase initially, plateaus, then decreases to a value smaller than the expected lowest number of instructions needed to define replication.
Dr. RBH and I agree that for very small genomes which reproduce very rapidly, natural selection (not “evolution”!) will favor genome truncation while at the same time attempt to retain larger genomes if these encode richly rewarded logic functions. In my essays I show that the proportion of logic functions among random sequences is huge. In virtually all Avida work I am familiar with, long before any significant loss of junk has been removed, some logic functions have already been stumbled on and are building up throughout the population. Truncation is halted. This is an artefact of logic functions and is biologically implausible as discussed in detail in the two essays.
I chose to dedicate the last several months performing Avida runs to substantiate my claims in the two essays instead of arguing in circles with Dr. RBH. He does understand the Avida platform well, but has not yet worked out in sufficient detail how various parameters need to be set to permit extrapolation to real, self-contained organisms such as the smallest of bacteria.
Professor Charlie d. has also not given careful enough thought as to the very strong tendency towards genome compaction for the kinds of organisms suitable for neo-Darwinian Theory evaluations using Avida. Generation of huge amounts of worthless mRNA and non-functional polypeptide would be strongly disfavored by natural selection (this is a prediction which has absolutely nothing to do with evolutionary theory).
Perhaps a dialog can be re-started. I will not participate in the kind of tone going on at the ARN site, though! The highly educated PhD evolutionists I deal with every day have good reasons for what they believe and I think we all have to review our own reasoning more humbly. I am quite shocked at the kinds of exchanges I read on the US sites, this is very foreign to us. (I know from the last interaction with Dr. RBH that I mucked up badly in my reasoning once, and he did so too. I hope we are both better informed from the exchanges!).
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 16. July 2004 19:08
quote:
Professor Charlie d. has also not given careful enough thought as to the very strong tendency towards genome compaction for the kinds of organisms suitable for neo-Darwinian Theory evaluations using Avida. Generation of huge amounts of worthless mRNA and non-functional polypeptide would be strongly disfavored by natural selection (this is a prediction which has absolutely nothing to do with evolutionary theory).
I am not sure I understand what you mean here. Are you saying that natural selection strongly disfavors organisms with large amounts of duplicated, redundant DNA? Or just organims producing a vast eccess of mRNA and proteins from a specific gene?
Also, I am not sure that metabolic costs of mRNA and protein synthesis really apply to AVIDA, in which all the metabolic costs are associated with genome replication.
IP: Logged
|
|
|
Printer-friendly view of this topic