Uncommon Descent Serving The Intelligent Design Community

Open Mike: Cornell OBI Conference Chapter 7—Probability of Beneficial Mutation—Conclusion


Biological Information

To facilitate discussion, we are publishing the abstracts and conclusions/summaries of the 24 papers from the Cornell Conference on the Origin of Biological Information here at Uncommon Descent, with cumulative links to previous papers at the bottom of each page.

Note: The hard cover version is now shipping.

The conclusion for Multiple Overlapping Genetic Codes Profoundly Reduce the Probability of Beneficial Mutation by George Montañez, Robert J. Marks II, Jorge Fernandez, John C. Sanford:

Our analysis confirms mathematically what would seem intuitively obvious — multiple overlapping codes within the genome must radically change our expectations regarding the rate of beneficial mutations. As the number of overlapping codes increases, the rate of potential beneficial mutation decreases exponentially, quickly approaching zero. Therefore the new evidence for ubiquitous overlapping codes in higher genomes strongly indicates that beneficial mutations should be extremely rare. This evidence combined with increasing evidence that biological systems are highly optimized, and evidence that only relatively high-impact beneficial mutations can be effectively amplified by natural selection, lead us to conclude that mutations which are both selectable and unambiguously beneficial must be vanishingly rare. This conclusion raises serious questions. How might such vanishingly rare beneficial mutations ever be sufficient for genome building? How might genetic degeneration ever be averted, given the continuous accumulation of low impact deleterious mutations?

See also: Origin of Biological Information conference: Its goals

Open Mike: Origin of Biological Information conference: Origin of life studies flatlined

Open Mike: Cornell OBI Conference— Can you answer these conundrums about information?

Open Mike: Cornell OBI Conference—Is a new definition of information needed for biology? (Chapter 2)

Open Mike: Cornell OBI Conference—New definition of information proposed: Universal Information (Chapter 2)

Open Mike: Cornell OBI Conference—Chapter Three, Dembski, Ewert, and Marks on the true cost of a successful search

Open Mike: Cornell OBI Conference—Chapter Three on the true cost of a successful search—Conservation of information

Open Mike: Cornell OBI Conference—Chapter Four: Pragmatic Information

Open Mike: Cornell OBI Conference—Chapter Four, Pragmatic information: Conclusion

Open Mike: Cornell OBI Conference Chapter Five Abstract

Open Mike: Cornell OBI Conference Chapter Five – Basener on limits of chaos – Conclusion

Open Mike: Cornell OBI Conference Chapter Six – Ewert et all on the Tierra evolution program – Abstract

Open Mike: Cornell OBI Conference Chapter Six – Ewert et all on the Tierra evolution program – Conclusion

Open Mike: Cornell OBI Conference Chapter 7—Probability of Beneficial Mutation— Abstract

For each code, we assume there is one optimal nucleotide of the four (for that code and that fitness function, whatever they may be),
That's exactly what I'm objecting to, not in general, but in a few select cases that fit into Behe's 1st rule of adaptive evolution, namely reproductive success is achieved via losing function. In the case of a loss of function scenario, there are several nucleotide configurations that can achieve improved reproductive success by disabling function, not creating it. Hence, the notion of "optimal" is reversed from your model, where breaking of the polyconstraints is possible in numerous ways and will result in disabling a functioning system in numerous ways. The proper way to model such a scenario is something of the reverse of your assumption. In a loss-of-function-results-in-reproductive-success scenario, one nucleotide sequence results in decrease of reproductive success, whereas 4^N-1 sequences results in reproductive increase, where N is the number of nucleotides under consideration. In such a scenario, the probability of a "beneficial" in the Darwinian sense is extremely high, even though it is clear this is damaging to the genome. What this does indicate is that selection, will not create more complexity or polyconstrained forms, it actually prevents them from forming or maintained in a population. If there were a two paragraph addendum to your paper to deal with the loss-of-function scenario (in agreement with Behe's 1st rule of adaptive evolution), I probably wouldn't have had anything to criticize in an otherwise stellar paper. Your paper demonstrates polyconstrained complexity cannot be reached via selection. My criticisms and recommendation for this addendum is to prevent Darwinists from citing things like anti-biotic resistance or pesticide resistance as some sort of counter example to the claims of the paper. In the case of antibiotic resistance, "optimal" may mean disabling of a pump or disabling proper expression of a protein. There might be many routes to disable function in those cases. In the case of pesticide resistance, in some cases it's through the malformed exoskeleton that almost suffocates the insect, but this near suffocating phenotype also prevents the insect from inhaling poisons at the same rate as its peers. Hence, this disability becomes a "beneficial" mutation! I believe there has been enormous damage to the genome in humans, but it hasn't necessarily resulted in reduced reproductive success, and my apologies for citing Octomom, but its the most poignant I could think of. My understanding is intelligence levels have declined, but humans have continued to be more reproductively successful than our smarter and more polyconstrained ancestors. There are a few cases of people with photographic memory, for example. It seems to me this is one example of loss of function in the general population. I would presume in ancient times, memory functioned far better than today. scordova
Hi Sal, Thank you for your comments, as they give me an opportunity to clarify several points of the paper. First, the model applies to any fitness function chosen (even those that confer large fitness benefits to octomom or blind cave fish). Choose your fitness function, and the model applies (as long as the independence assumptions among codes, and other assumptions outlined in the paper hold). Second, the paper models fitness on the nucleotide level, with each nucleotide having some tunable probability of being "optimal" (by whatever fixed definition or fitness function you want to choose). Thus your criticism about people having different notions of optimal, or it being hard to define, are irrelevant in the sense that the definition of (or probability of being) optimal is left open as a user defined parameter. We show results for a variety of parameter values for p(optimal). The model holds regardless of how you choose to define optimal, as long as you set that parameter to a probability value (between 0.0 and 1.0). Let's review the model used, for clarity. First, we have a stretch of DNA. For a given nucleotide in that stretch, it participates in between 1 to L independent codes. For each code, we assume there is one optimal nucleotide of the four (for that code and that fitness function, whatever they may be), and that if the base is not the optimum, there is one base that is better than it (the optimum for that code at that position), and two worse than it (harmful bases). Thus, for a non-optimal position, for a single code, there is a 1/3 probability of beneficial mutation, and 2/3 probability of non-beneficial (fitness decreasing) mutation. Now we apply this across L independent codes simultaneously, and you get the model presented in the paper. The probability p(optimal) is a tunable model parameter, which defines the probability of each nucleotide already being optimal for a given code at a given spot under a given fitness function. If it is optimal, the three other mutations will decrease fitness (assuming no truly neutral mutations, a simplifying assumption). You can think of the generative model in this way, given your fondness for gambling. You have a nucleotide in front of you, say the base G at position 55; generate a mutation by rolling an imaginary three-outcome die. Two of the faces will signal a harmful mutation, and one an improvement for that spot, for that code, if the nucleotide is non-optimal. Assume the outcome is T. To determine if G is already optimal or not for that code at that position, flip a weighted coin, with probability p(optimal) heads, 1-p(optimal) tails. If optimal (heads), the change to T will be harmful for that code (by definition of optimum, under the simplifying assumption). If non-optimal, see if the change to T was one of the two harmful mutations for that code (say A or C), or an improving mutation (say, T). Record the outcome (beneficial, yay!), then move onto the next code, flipping the weighted coin to see if G is already optimal for the next code, and if not, see if the new nucleotide T is beneficial or harmful for that code at that position. Count how many it was a beneficial change for versus a harmful change, for all your codes. Take the majority label (harmful/beneficial) across codes as the final outcome for that mutation (G->T) at that spot (position 55). You'll find a very low probability that your change was beneficial overall. (Reviewing the paper itself may prove instructive for clearing up any misunderstandings you have concerning the model or the results presented there.) Lastly, yes, the paper does give a toy example using inclusion in a fixed dictionary of Scrabble words as the criteria for fitness (a simple binary fitness function), but the model does not require a "grammatical" notion of fitness. You can use any fitness function you want. I used a "grammatical" one, since it is easily understandable, easily computable, and no one will can accuse me of stacking the deck with a favorable fitness function or against the process. It is non-arbitrary (I didn't make up the list) and meaningful. However, you can expand this dictionary to include all the slang words and teenage twitter vocabulary you like. The results will still be qualitatively the same. George George
I thought about this some more. A better way to frame the argument in the paper:
It is improbable that mutations randomly emerge that are grammatically correct according to the 12 codes
It does not mean a grammatically wrong mutation will necessarily be non beneficial in a reproductive sense. What do I mean. Take the case of human speech. Poor or expedient grammar in human relations has selective advantage for ideas. We say things like "ain't," or "nother" or "gonna". Such things might be grammatically suspect, but they are selectively advantaged constructs since they're so pervasive and they convey meaning expeidently. If I say, "it ain't gonna to work", in conveys a bit of derision that a can't be conveyed with the more grammatically and properly spelled phrase: "this will not work." Look at the grammar when teens text each other. Yikes! So "beneficial" should not be the criteria, grammatical constructs should be. With that in mind, the paper will have more force. Since, as we see in human relations, poor grammar can pervade a culture, so too, mutations that violate the gene grammars can be "beneficial" in the Darwinian sense. Octomom is an illustration of this (sorry to pick on Octomom, but she makes an illustrative example of the problem of "beneficial" in the Darwinian sense versus "beneficial" in the functional sense). scordova
Sal, just a comment on "chickens don't fly." In nature, they probably do, but only up to a height. I remember from childhood, the old-fashioned barnyard hen could flap up to the chicken shed roof, which is all she needed, to get away from dogs and coyotes. Actually, a chicken that couldn't fly at all would not be optimal. The question with any bird is, what does it need flight FOR? O'Leary for News News
Good work, but I offer a stiff criticism. Take the case of Octomom, who by any standards has a few screws loose, yet in the Darwinian world she is 14 times more reproductively successful than Richard Dawkins because she has 14 offspring, Dawkins only 1. We find such numerous examples in biology of "beneficial" traits. They are not that rare if we define beneficial in terms of reproductive success. The problem is the notion of beneficial in that paper is relative to function, and the notion of function is subjective. For example, many household appliances can be said to function as a paper weight if it is weighing down a piece of paper. How then do we characterize function without using subjective specifications? That said, a better way perhaps to frame the argument is the difficulty in finding functional forms. We then have to work on defining principles for identifying function in an engineering sort of way. The problem is when Darwinists will call a blind creature like a blind cave fish as the fittest phenotype, then they can argue "beneficial" is easy to achieve because even breaking of function creates reproductive advantage. How do we define function? I suggest medical notions of healthy are a start...it's arbitrary but it gets the point across. One challenge is that the paper refers to optimal. For example, what is the optimal resting heart rate for a human? Or resting respiration rate, etc.... In the engineering world, we have engineering specifications from the designer that tell us if the object is functioning optimally (i.e. a Car has certain operating parameters deemed optimal). The problem is how can we define optimal without the design specifications. We could fabricate them of course, but how can we justify them as valid? Perhaps there are certain consideration we can invoke from physics. For example a bird won't fly if it not operating optimally along several dimensions. A Darwinists might say, "chickens don't fly, but they are optimal." Suffice to say, I don't think we'll find resolution on this problem any time soon if ever... scordova

Leave a Reply