I thought some of you might be interested to know that a paper of mine was recently published in BIO-Complexity – “Measuring Active Information in Biological Systems.” The goal of the paper is to provide a way of verifying whether (and how much) a mutational process is directed or undirected. I posted an overview of the paper at UD if you don’t have time to read it, but I thought this crowd might enjoy the details more.
Also, for background of why this is relevant, you might read my earlier paper, “Evolutionary Teleonomy as a Unifying Principle of the Extended Evolutionary Synthesis“.
I like the idea of cell–design directing mutations. yet it still means biology needs mutations. As a creationist i rather not see biology as the result of mutations unless mutations is the wrong word. instead just a diversity in the dna is being invoked and sometims its a random act which we call mutations. anyways thats just my thought and not directly about this paper.
The fact that you would or would not rather see anything counts neither as evidence nor logic, Robert.
Pretty much sums it up.
Just to help with feedback, Jon, I’d make a couple of points:
1. Biological evolution is not a search. Living organisms are just, well, living, surviving, breeding, or not.
2.There is not one solution and an obligation to look through a space of a certain size to find it. Mutations (in the broadest sense of novel arrangements in genotypes) turn up randomly and get selected for or against depending on whether they confer an advantage to individual organisms in a particular niche.
NB: https://www.blythinstitute.org/site/partnership/research-fellows/
Guess who are the two and only two Research Fellows? = P
At this rate, johnnyb will be accepting Robert Byers as the next Research Fellow of the Blythe Institute!
Whoever Trey Merkley is, allying with creationists & IDists, he did not choose wisely. : (
johnnyb – you’ve been duped by Expelled Syndrome and the DI to throw away your talents in defense of them as a trick; they make it appear that if you reject ID theory, then you must reject also the theological worldview that you embrace. Nothing could be further from the truth!
You can heal from this, but must depart as soon as possible from the DI’s tutelage & propaganda machine. It only makes you appear pseudo-sincere in your “research”, and easily seduced by conspiracy theories. Sad to see this work surfacing from you. ; ((
@Johnnyb
It’s not exactly clear what you mean by a “pure random search”(IΩ) in the context of biology. Let me try and come up with an example and you can perhaps clarify?
Supposing we just talk about random substitution constrained to a very small locus, we could imagine that we measure the fitness effects of all single-nucleotide substitutions in that locus. If there is a biochemical bias in the process of substitution (which of course we know there is), is that then going to count as part of the “search under analysis”(IS), or is that the “pure random search”(IΩ)?
If it’s part of the “search under analysis”(IS), then the absence of substitution bias must be what is meant by a “pure random search”(IΩ)? That should then imply, if I am to make sense of the definition, equiprobable substitutions. That all transitions and transversions are equally probable. Thus we could determine if any bias in the substitution process contributes active information to evolution of the locus, if such a bias increases the probability of beneficial mutation.
Do I have it correct that in this admittedly contrived example, the absence of a bias to the process of mutation is what would constitute the “pure random search”, aka (IΩ)?
@RobertByers –
The technical definition of mutation is *any* change in DNA sequence. The conflation of “mutation” with “random mutation” is due to the suppositions of the modern synthesis that got carried on far past their expiration date.
@AlanFox –
Thanks for taking the time to comment!
“Biological evolution is not a search. Living organisms are just, well, living, surviving, breeding, or not.”
As the paper mentions, it does not rely on evolution *ontologically* being a search to be valid. Only that there is a “problem” (i.e., selection pressure), a “goal” (relieving of selection pressure), and a multitude of potential paths.
“There is not one solution and an obligation to look through a space of a certain size to find it. Mutations (in the broadest sense of novel arrangements in genotypes) turn up randomly and get selected for or against depending on whether they confer an advantage to individual organisms in a particular niche.”
This is what is being tested against. I agree 100% that we should not assume that there is a solitary, isolated solution. The mechanism of Active Information is 100% in agreement with that. As long as we can judge that we hit the goal is all that is required. In other words, it only requires that we be able to measure fitness in some way.
@Gregory –
I’m not really sure what your point is. Did you agree or disagree with my paper? If I’m “throwing away my talents”, does that mean you agree with the paper? If you do, where do you think I should have published it? Usually the criticism is that I have no talent, so, I’m not quite sure what to make of this one 🙂
@Rumracket –
“It’s not exactly clear what you mean by a “pure random search”(IΩ) in the context of biology”
Remember, the “pure random search” is in the context of mathematics, not biology. The goal is not to say what biology will do, but to say what the expected value should be.
“If there is a biochemical bias in the process of substitution (which of course we know there is), is that then going to count as part of the “search under analysis”(IS) , or is that the “pure random search”(IΩ)?”
That’s definitely I_S.
“Do I have it correct that in this admittedly contrived example, the absence of a bias to the process of mutation is what would constitute the “pure random search”, aka (IΩ)?”
Yes, at least that’s the theoretical goal. It may only be approximatable though. However, I agree with Axe’s “coverage principle” – that an haphazardly chosen selection will generally have the same basic probabilistic features as a “pure random search”. Several fields (including machine learning) rely on this fact.
However, just to throw in a little complexity, as is mentioned in the section on “relative AI”, you can actually calculate AI relative to whatever background you like. My goal was to measure it in comparison to a pure random search. But, in theory, there is no reason why you *couldn’t* measure AI in relation to SNP biases. I’m not sure what use it would give you, unless you were specifically searching for non-SNP-oriented mechanisms or something. My goal was to be as agnostic as I can be about the nature of the information, and simply measure whether the cell is geared towards generating beneficial mutations or not. So, because of that, SNP biases are part of I_S.
Even if we assume targeted search, how does one calculate the probability the search will find the target without specifying the search space? A deterministic search algorithm will find the target or it will not. The only random factor is the starting point in the search space, and possibly the search space itself if you want to choose one randomly. We can add searches that allow randomness (genetic search, simulated annealing) which will create a region of search space where the probability of finding a target is strictly between 0.0 and 1.0, but this too will depend on the search space.
I don’t think this is a search problem at all, but rather an optimization problem. The algorithm climbs the hill to the top and stops. Algorithms with randomness might be able to avoid local maximums and find better solutions than deterministic ones. Probability of finding the target is simply the wrong question – It’s a matter of algorithm efficiency, and the best solution that can be found in the available time. The target solution isn’t even relevant, it only need to be “good enough”.
There is ongoing discussion of this paper at Peaceful Science too.
https://discourse.peacefulscience.org/t/bartlett-measuring-active-information-in-biological-systems/10096/
I’d also like to note that equation (3) is the very nearly a statistical likelihood ratio statistic, differing only by a constant. Everything that follows is essentially trying to interpret the magnitude of the p-value of an unspecified model.
Interpreted as search, the number of iterations is ignored. If the search algorithm is allowed to iterate, then random search much be allowed to iterate as well. This isn’t even a fair comparison to the straw man of random search.
Well i don’t like the technical use of the word mutation just to describe a change in dna. Thats not right. A mutation must mean a twist off something. Change in dna , as a option, must be allowed without it being said its mutation. Something wrong with definitions here.
Etymology of the word “mutation”.
Out of interest, Robert, what causes these ‘changes’?
My first reaction is that one has to be careful about biases in methodology. If your list of beneficial mutations are the product of a biased search then you can arrive at the false or unsupported conclusion that the biased search is better than a purely random search.
For example, naturally occurring substitution mutations favor transitions and CpG mutations in eukaryotes. Therefore, you will tend to see more beneficial mutations that are a result of those mutations. However, beneficial mutations may be evenly distributed between transitions, transversions, CpG mutations, and non-CpG mutations. There may be some beneficial mutations that the natural bias would be less likely to find.
All true, but you could map that out for a sufficiently small locus. It would definitely be unfeasible for the entire genome.
DNA is fantastic complex. There is no justification to say mankind has figured it out. So innate triggers within the system after thresholds are crossed is a option.
how genetics works is not fixed just because we see nothing going on today.
In fact there are sea creatures ho change bodyplans simply upon moving ten feet that way and poof they adapt to the colour schemes there to be camouflaged.
Robert Byers,
What triggers, what threshold?
No part of this sentence is true. You have been shown multiple times what genetic change can be seen going on today.
T_aquaticus,
Yes, I see the Texas Sharpshooter Fallacy in Bartlett’s analysis (left column of page 10 — why no section numbers?) of a particular mutation that occurred in the Long-Term Evolution Experiment.
Consider replacing an n-base gene g with (1) a sequence of n bases drawn uniformly at random or (2) a sequence of n bases obtained by randomly shuffling the exons of gene g. The latter is much more likely to improve fitness than the former. But it doesn’t follow that a process sampling a space of DNA sequences by shuffling the exons of an existing gene is itself informed of the fitnesses associated with novel DNA sequences.
As I have said many times — I’m not interested in going through all of it yet again — sampling processes are biased, not informed. When an event is more likely to occur in one sampling process than in another, that merely indicates a difference in bias of the processes.
The game that ID proponents are playing, when they refer to sampling as search, is to make “obvious” the false notion that the outcome of the process depends on information intrinsic to the process.
A sampling process is never, in and of itself, a search. You are categorically wrong in suggesting otherwise. A sampling process has no information whatsoever of properties of as-yet unsampled elements of the sample space. That is the fundamental reason that there is “no free lunch” for someone who selects a sampling algorithm for use in a search for a solution to a problem. The algorithm-selecting agent possibly exploits information, but in no case is the sampling process described by selected algorithm informed. I have addressed this matter with great rigor in “Sampling Bias Is Not Information.”
Consider spinning a U.S. penny on its edge, say, on a kitchen countertop. Tails is about 4 times as likely to come up as heads. I can use prior information of the bias in favor of tails to improve the expected value of a bet on the outcome. But it would be ridiculous to say that the process of spinning a penny itself has information relative to the nominally unbiased process of flipping a penny.
Tom English,
What does this mean Tom, a penny spinning on its edge is more likely to land on tails? It is something about the weight of a penny? A US penny? A new one, a clean one?
Tom English,
I saw this, but the author seems to contradict himself. Perhaps he has a poor editor.
https://www.tested.com/science/math/451830-spin-penny-land-tails/
But then, if a penny is weighted unevenly surely that also effects it when flipped too.
Well, why not do as Tom suggests and try it? You’d be conducting scientific research.
A slight offset of the center of gravity has a fairly big effect on the outcome of a spin, and a negligible effect on the outcome of a flip. Years ago, I read about the physics. But I don’t recall enough to give you a good explanation of what’s going on.
For a detailed account of coin-flipping, see Dynamical Bias in the Coin Toss (there is a slight difference in the probabilities of heads and tails). I don’t know, offhand, a good reference on coin spinning.
Well, I have just read up more on it, and I think it is more legend than fact. If fact what I read said to try it yourself, but that you might not notice it because the penny gets dirty. Furthermore, if the weight difference was very much at all, you could never balance a penny on its edge. And on top of that if the edge is at all uneven, which it almost certainly would be that will also throw it off. So I bet you won’t see the so called 80 percent results either way. And I don’t have a penny. I am not American, remember.
Years ago, I experimented with the metal lid of a big pickle jar. The center of gravity is offset more for a lid than for any coin I’ve ever seen. The results of flipping were close to 50-50. The intuition that the heavier side should land down most of the time, when the lid is spun, is wrong. It’s the other way around. In fact, the lid was beveled at the top edge. That would seem to bias the outcome in favor of top-down. But the outcome of spinning the lid on its edge was top-up for more than 90 percent of the trials. (I don’t recall the exact number.)
I didn’t know that. I thought you were USAmerican currently rice farming in China. For a non-American, you seem quite invested in US politics.
Tom English,
Interesting.
In other words, we don’t and can’t ever know. How’s that working out for you?
That’s for someone else to do. IDists don’t do work.
I find Tom English’s response very intriguing. Here’s what he says:
What I understand English to be saying is that he doesn’t think that organism have information about how they should mutate towards likely beneficial results. And his evidence for this is that he can point to specific processes that exist in organisms and mutation that make it more likely to mutate towards beneficial results (exon shuffling, in this specific case).
I think most people would classify such processes as “information”.
So, in other words, the reason why Tom English thinks there isn’t any information biasing mutations is that it has already been identified, and he is describing where it is and what it looks like.
I mean, I really don’t know what to say about that.
This is a good point. Wasn’t obvious to me at all before you said it, so thank you for dropping by to point it out.
To be clear then, you believe that a process sampling a space of DNA sequences by shuffling the exons of an existing gene is itself informed of the fitnesses associated with novel DNA sequences?
Y/N?
But you want to apply your ideas to biology. Since “active information” is measured relative to a “pure random search”, if you can’t define what the “pure random search” is in a biological context, you can never measure “active information” in a biological context and your application of “active information” to biology is a non-starter.
OR if we allow that genetic drift is a random sampling, this will contribute active information within some portion of the biological population, and the value of “random search” has been misstated. Useless or wrong – take your pick.
IIRC toward the end of the paper it mentions that any two “search” strategies can be compared. This might be true: if active information can be viewed as a likelihood ratio statistic then there is a mature body of theory and methods on how to do this. This just to show that a few month in the laboratory (or at the chalkboard) can save you a few hours in the library.
Roy –
I think you misunderstood what I was saying. My point was that the question is to see what *biology* does compared to *expectation*. If you are just comparing biology-to-biology, then you don’t get an answer about whether or not biology follows expectation.
Take Tom English’s coin-flipping example. If someone told us that a coin is perfectly equivalent on both sides, then we could use Active Information to show that this is not the case. If someone told us that tails would be likely to come up, again, we can use Active Information to show that this isn’t the case. If we tried to show instead that it “doesn’t come up heads more often than coins normally do”, we haven’t learned anything.
The claim in biology is that “mutations are random with respect to fitness outcomes”. This is either true or false. Active Information gives us an expected value for the claim and a methodology for examining its truth.
Here’s my challenge. If you do not like the method I have proposed for determining whether or not mutations are directed towards function, please do one of the following:
1) Give your own methodology for determining whether or not a mutation is random with respect to the biological needs of the organism. Be specific and quantitative. How can we know that a mutation is random with respect to the needs of the organism.
or
2) In the question of whether or not a mutation is random with respect to the needs of the organism, take the position that we don’t have the ability to know this at all, and claims for either side lack justification. This would mean explicitly denying what most evolutionary biologists involved in this debate (Swamidass, Moran, Coyne, etc.) believe.
I’m happy to have a discussion about #1. That’s perfectly reasonable. The reason I wrote the paper is because there was simply a lot of assertion back-and-forth about “directedness”, but nobody seemed to have a method for determining it. I believe I have come up with a method from basic principles that will help answer the question. If you think it’s bad, great! Let’s have a discussion about what would be better. If you think that there is no way to do it, then you should be in equal disagreement with Swamidass, Moran, Coyne, and others as you are with me.
In answer to your (1), I can provide you with a way of testing whether the process by which mutations arise is random wrt fitness. This is what biologists mean when they state that “mutations are random wrt fitness outcomes”.
Asking, as you did, whether a mutation (singular) is random wrt fitness appears nonsensical. We can ask how it arose, but its effect on the organism is uncontroversially not random. Could you clarify?
I think this is where the problem lies. I’m not sure one can tell whether the process is random without knowing the results. I’d be interested to hear what you have to say, but I’m skeptical that I would conceive of any ability to understand whether a process is random without knowing anything about the results.
I agree that we need to have more than a single data point (if that’s what you were saying about “a mutation (singular)”), and my proposal relies on multiple data points (in fact, it relies heavily on the law of large numbers).
Of course we look at the results. I never suggested otherwise.
If I understand you correctly re the law of large numbers, you meant to write
“Random” is an invitation to equivocation. We can show that the proportions of different mutations that arise are unbiased with respect to the “needs of the organism”. Would that suffice?
You need results. You also need to specify the form of randomness (a null hypothesis) such as IID uniform, or perhaps multinomial proportional of the original base gene population frequencies as Tom suggested. IID uniform doesn’t seems biologically plausible, but it is simpler to think about. Then a test of heterogeneity to the null.
BUT any selection for fitness, including censoring of deadly mutations, will tend to reject the null and appear non-random.
“Needs” is hard to pin down. An organism needs to be effective at gathering and dissipating energy with respect to reproduction. “Effective” should equate to population survival by any means. If you could specify the fitness function, the fitness landscape, and the available survival strategies, THEN you might be able to test for non-randomness independent of fitness.
Outside of a simplified model, this doesn’t seem practical.
Non-randomness independent of fitness might suggest you should be tenting for fitness that is lower than expectations, rather than higher. I am implying that a Designer could prevent a particular mutation, not just promote one. In all seriousness, consider this carefully.
Just wondering if JohnnyB thinks the Lederberg replica-plating experiments are relevant.
I also wonder if it’s relevant that we see the same mutation spectrum in junk DNA as we do in DNA under selection.
Not to mention the Luria-Delbrueck experiments.
I think you’re misunderstanding my point. Unless you can define “pure random search” in a way that is relevant to biology, you have no expectation to compare biology to.
Unless you define what pure randomness is for mutations, which you have not done (and I suspect cannot do since it would require not just examining how mutations occur and are corrected for in existing life forms, but also how they occurred in both long-extinct life forms and also in life forms that could have evolved with different genetic mechanisms) then you cannot determine that expected value nor use your methodology.
By challenging me to produce an alternative means of resolving this question, rather than addressing the flaws of your own approach, you are effectively admitting that your work is unsustainable, and that I should try something else.
I am interested in whether mutations are random w.r.t. fitness – but there doesn’t seem to be much point in discussing it here with you.
The expected value alone is not sufficient to describe a population, or calculate probabilities. There will be variability in a biological population, and the expected variance is also needed if you want to calculate probabilities (with some additional assumptions about multivariate normality and covariance).
Johnnyb: You should have another, much longer chat with your statistician.
DNA_Jock –
Just to note, I find each person is different regarding which “wording” they find appropriate. However, I am fine with “unbiased with respect to the needs of the organism”, and I believe that my measurement measures the amount of bias there actually is or isn’t.
John Harshman and Joe Felsenstein-
Just to note, my first full paper was actually on these two experiments. See Statistical and Philosophical Notions of Randomness in Creation Biology. In short, what the Lederberg and Luria-Delbruck experiments actually show is whether or not a mutation happened prior to selection or because of selection. It does not show whether or not these mutations are random with respect to the needs of the organism.
Remember, both of these methods were developed prior to the discovery of DNA, so some of the things that people thought they showed were invalidated (not necessarily shown false, just shown that the reasoning was not correct) by the discovery of how DNA and genetics works. Luria-Delbruck was 1943, Lederberg was 1952, and Watson-Crick was 1953. People still hang on to the idea that this shows that mutations are random, but it just isn’t so. As an example, let’s say that organisms mutate in a way that is more likely to be biologically meaningful than not. This would not be random (or unbiased) with respect to the needs of the organism. Now, it does not directly correlate to its needs now, but, under this idea, if you compared the option that occurred compared to all the options available, it is certainly biased in favor of the needs of the organism, as it would not mutate in many possible ways that are definitively counter to the needs of the organism.
In any case, for my methodology, the Lederberg replica plating experiment would be useful as a means to verify sure that the mutation didn’t already exist in the population prior to perform my suggested experiment. Luria-Delbruck might be usable to measure the active information of non-selective changes, though I haven’t really put a lot of deep thought into that. Though that’s possibly an interesting direction to take.
Tomato –
First of all, the assumption going in to the experiment are that the organisms are basically identical (it’s a simplifying assumption, but all biological mathematics makes simplifying assumptions such as this). Therefore, the successes in the experiment are binomially distributed. Population count and expected value are precisely the only two things you need to calculate probabilities (technically, the parameters are population count and probability, but it is easy enough to work backwards as from expected value to probability). I’m surprised you are unaware of this.
There was some question as to whether or not failures were under-counted for lethal mutations. I don’t think so, since, as is usually the case in such experiments, you count successes, not failures. Population size is counted in generalities. We don’t go and count each individual cell, but look at population size and density. The number of successes are sufficiently small that they can be counted individually.
I agree wholeheartedly. There is nothing in my experiment which would suggest otherwise. My goal for my experiment was to minimize any assumption of what the mechanism would be. A mechanism preventing bad mutations and pointing to good mutations would both equally work to generate positive active information for these calculations. Additionally, note that if the mutational mechanisms were pointing the other way – away from fitness – these calculations would reflect that by giving a negative quantity for active information.
johnnyb:
According to your article:
Do you believe that the probability
of generating immunoglobulin 
in response to a novel antigen depends on the particular antigen?
More concretely, suppose that one group of lab mice is infected with virus A, and another group is infected with virus B. Do you expect the frequency distributions of mutations in the CDR to be different for the two groups? (ETA: Suppose that binding to virus A requires that different sites be mutated than does binding to virus B.)
I don’t think there’s any known evidence for it. One slight exception – If I recall correctly the immune system seems to know when to turn off and when to keep going, so that will increase the probability of finding a solution eventually (and not breaking known solutions). But, to what I think your point is, as far as we can tell, there is no correlation between the antigen and *which* specific base pairs in the targeted region of interest get mutated.
Additionally, your question made me think of an application. We could use the relative form of active information to actually measure whether or not mutations are biased based on the antigen. I imagine the result you will get is that active information will be zero, but active information would give us an experimental result. That would tell us whether looking deeper for a matching mechanism would be worthwhile (again, my guess is that Active Information will be zero and there is not a deeper mechanism).
johnnyb,
I believe you’re wrong. If a mutation happens before it’s needed, that means it’s not responding to the needs of the organism. Unless you assume that organisms are trying (successfully) to predict future environments different from what’s currently experienced, they argue against your position.
You are conflating two different ideas – (1) whether a mutation is *responding* to the needs of the organism, and (2) whether a mutation is *biased* considering the needs of the organism. Those are two related, but different questions. I agree that you are correct that if it happens before it is needed, then, by definition, it’s not responding to needs. I disagree that this means that it is unbiased with respect to the needs of the organism. For instance, if a mutation is more likely to confer a configuration that is biologically consistent than one that is not, it is biased to the needs of the organism, plain and simple. And that bias requires explanation.
I do think that this happens, too. They needn’t be successful 100% of the time, they just need to be more successful at it than if they tried at random.
Another metric that would be worth looking at is mutation rates compared to the Kelly Criterion for probability of success. i.e., does the mutation spectrum make a rough match to the expected future possibilities as gauged by the Kelly Criterion? That would be tough to measure (and would take equal parts good philosophy, good math, and good experimentation) but may be interesting. Microbial “Bet hedging” was an active area of research in the 80s, if I recall correctly, but petered out because no one knew how to measure it. I think using the Kelly Criterion would be helpful, but, as I suggested, might need to be modified.
johnnyb,
I agree with this. It is how I often think about design. The design doesn’t have to be tweaked by an invisible hand at every turn and enable an organism infinite possibilities of survival adaptation to be designed. It just has to have mechanisms in place that give it abetter then random chance at survival. And I’m fact this is what we see all around us. We see body systems that have so many complex adaptation features that can be called upon to assist the survival chances of the organism. But evolutionist like to claim that these systems somehow evolved as a fluke first, and then the systems became semi intelligent. But it was all by accident. Like like they see the genetic code or the system of epigentic switches. They want us to believe those were random events that caused them but it turned out useful. That is credulity stretching to them extreme if you ask me.
Do you mean like an all-powerful magical being speaking things into existence? Like that?
I don’t know about what other may or may not want you to believe. I don’t want you to believe anything. I’d be content if you understood, even if you didn’t believe it, and thus learned that you have no reason to be so angry and offended that we don’t believe in magical beings in the sky as you do.