Evolving complex features

The Lenski et al 2003 paper, The evolutionary origin of complex features, is really worth reading.  Here’s the abstract:

A long-standing challenge to evolutionary theory has been whether it can explain the origin of complex organismal features. We examined this issue using digital organisms—computer programs that self-replicate, mutate, compete and evolve. Populations of digital organisms often evolved the ability to perform complex logic functions requiring the coordinated execution of many genomic instructions. Complex functions evolved by building on simpler functions that had evolved earlier, provided that these were also selectively favoured. However, no particular intermediate stage was essential for evolving complex functions. The first genotypes able to perform complex functions differed from their non-performing parents by only one or two mutations, but differed from the ancestor by many mutations that were also crucial to the new functions. In some cases, mutations that were deleterious when they appeared served as stepping-stones in the evolution of complex features. These findings show how complex functions can originate by random mutation and natural selection.

The thing about a computer instantiation of evolution like AVIDA is that you can check back every lineage and examine the fitness of all precursors.  Not only that, but you can choose your own environment, and how much selecting it does.  There are some really key findings:

  • Irreducibly complex functions evolve, where IC is defined as “if you take something away, it breaks”
  • Functions evolve via Irreducibly Complex pathways, i.e. pathways in which there are several neutral, or even quite steeply deleterious steps.
  • The more complex functions do not evolve without selection.
  • All the functions evolve via at least some neutral steps
  • There are many pathways to each function
  • Not all the functions are achieved in the same way.

The AVIDA findings are very good evidence that there is no reason, in principle, why irreducibly complex things can’t evolve, that some degree of natural selection aka heritable variance in reproductive success in the current environment is necessary for for certain features to evolve, but that not all steps need be advantageous – because of the role of drift, even quite sharply deleterious steps can prove key to the emergence of a complex function.

Then there’s the Lenski lab e-coli experiments.  But let’s start with AVIDA, because it really does rebut the idea that Irreducible Complexity, whether defined as the property of a function or of a path to a function, is an in-principle bar to evolution by natural selection.

242 thoughts on “Evolving complex features

  1. No, they have meteor storms. Did you see Allan’s video on gait evolution? There was one scenario where it was raining boulders.

    I love writing these things. I always make my hazards stochastic – it works way better if you let some no-hopers through, even if it means you lose some potential goodies sometimes too.

  2. Elizabeth: No, they have meteor storms.

    Heh. That actual meteor crack was a callback to this amusing post of yore that became something legendary, eventually:

    All computational evolutionary algorithms artificially isolate the effects of random mutation on the underlying machinery: the CPU instruction set, operating system, and algorithmic processes responsible for the replication process.

    If the blind-watchmaker thesis is correct for biological evolution, all of these artificial constraints must be eliminated. Every aspect of the simulation, both hardware and software, must be subject to random errors.

    Of course, this would result in immediate disaster and the extinction of the CPU, OS, simulation program, and the programmer, who would never get funding for further realistic simulation experiments.

    http://www.uncommondescent.com/intelligent-design/a-realistic-computational-simulation-of-random-mutation-filtered-by-natural-selection-in-biology/

  3. William,

    Finite resources are part of natural selection. So, you agree with me.

    I agree that finite resources impose a selective constraint, and that RM would eventually produce every possible adaption in the total absence of selection; but I disagree with what you are claiming in these threads: that selection has no role in building or producing complex adaptations.

    Finite resources are a given in nature. Populations can only support a limited number of individuals. If selection were not awarding the limited “slots” to fitter organisms, on average, then complex adaptations would not arise on evolutionary timescales.

    Again, this is exactly why Weasel works. Without selection, Weasel would not converge in your lifetime. With selection, it converges in seconds. In fact, I remember calculating somewhere that even under generous assumptions, Weasel without selection would take quintillions of human lifetimes to hit the jackpot.

  4. I suspect it will eventually be possible to make a computer model of evolution in which the hardware and OS replicate with variation. It will require massive parallel processing, but it’s within sight. I would guess someone will attempt it within ten or twenty years.

  5. petrushka,

    I suspect it will eventually be possible to make a computer model of evolution in which the hardware and OS replicate with variation.

    What would be the point?

  6. Keiths said:

    I agree that finite resources impose a selective constraint, and that RM would eventually produce every possible adaption in the total absence of selection;

    Yes, we agree on this.

    but I disagree with what you are claiming in these threads: that selection has no role in building or producing complex adaptions.

    I never claimed anything of the sort. In fact, I’ve agreed that it plays an essential role in producing adaptations of any sort. Please note: a feature is not an adaptation, and “playing an essential role” doesn’t necessarily mean “actually builds the thing in question”.

    If selection were not awarding the limited “slots” to fitter organisms, on average, then complex adaptations would not be arise on evolutionary timescales.

    If there are no resource constraints and no death, everything is awarded a slot because “slots” are not “limited”. It is Natural Selection that limits slots in the first place, thereby severely constraining the true capacity of RM to build only what NS allows. NS doesn’t “help” to build any feature whatsoever, although it is essential (by definition) for the establishment of the prevalence of those features in a population (by weeding out lesser competitors in a system organized by NS).

  7. William J. Murray: Please note: a feature is not an adaptation, and “playing an essential role” doesn’t necessarily mean “actually builds the thing in question”.

    Ah, finally you are providing footnotes and annotations as you go.

  8. Elizabeth: There can be dying without NS.So what “happened” in the NS scenario is that they died.Shit happens sometimes.

    Okay, let’s call meteor showers and those kinds of deaths (not related directly to fitness) non-selective. I’ll grant you that arguendo.

    It makes no significant difference to my “world without NS” scenario. A volcano here and a meteor there is only going to slightly, randomly interfere with the progression of RM produced features unhindered by limited resources and fitness-related death. It would still produce any feature RM+NS could produce, and lkely faster.

    Are you now going to claim that “limited resources” is not an aspect of NS?

  9. William J. Murray: It would still produce any feature RM+NS could produce, and lkely faster.

    Why? Support this please.

    EDIT: And how much faster? If you don’t know or can even give a guess then how do you know it’s faster at all? I mean, on what are you basing that “faster”?

  10. OMagain: Why? Support this please.

    EDIT: And how much faster? If you don’t know or can even give a guess then how do you know it’s faster at all? I mean, on what are you basing that “faster”?

    Logic. The only difference between an RM&NS world and a RM world is that NS kills off variations and lines of variation in its world due to selective pressures. No variations get killed off in the non-NS world due to selective pressures. All individual configurations and variations, at all times, continue to reproduce exponentially. Since no trait lines are eliminated an no obstacles exist on the way to any feature in the RM world, and they exist in the RM&NS world, logically the RM-only world would likely get there faster – because there are no NS obstacles.

  11. William J. Murray: world, and they exist in the RM&NS world, logically the RM-only world would likely get there faster – because there are no NS obstacles.

    But you use the word “world” twice but in two very different ways. No world I know of can support the latter never mind get there faster.

    A world that accommodates RM+NS and “gets to places” can exist. For example, the one you are on.

    A “world” that accommodates only RM and “gets to places” cannot exist, for it would no doubt turn into a black hole immediately due to the required size of it. Do you see what I mean? Every point from every point is a somewhat different proposition. It requires more real estate. More then a mere “world” can provide.

    So you see why your “thought experement” is incomplete? And hence why the conclusions you may draw from it would likewise be limited in their value?

  12. keiths:
    petrushka,
    What would be the point?

    What is the point of Avida? I’m not sure.

    But single CPU AI seems pretty much dead to me. I suspect if we look ahead 50 years we will have chips with billions of very simple CPUs having very simple OSs, that will mutate their own code, possibly die, and be “eaten” by other CPUs (reprogrammed in a kind of replication).

    I suspect the future of AI is hive intelligence.

  13. This is thread is not about more infinite monkeys resources.. It is about the AVIDA program.

    In the AVIDA program there is competition for resources.

    If you turn off RM you will get nothing, because at the beginning all the creatures are identical.

    If you turn it on you will get very few logic functions, and you never, in the life of any run, get EQU.

    If you turn on NS for all logic functions (i.e. critters that can do logic functions get better access to finite resources) then EQU evolves.

    Therefore, adding NS makes EQU happen when it wouldn’t without.

    William, do you dispute this?

  14. William,

    If there are no resource constraints and no death, everything is awarded a slot because “slots” are not “limited”. It is Natural Selection that limits slots in the first place, thereby severely constraining the true capacity of RM to build only what NS allows.

    It’s the environment that limits the number of slots, not natural selection. If the slots are awarded randomly, with no preference for one variant over another, then there is no selection.

    (What Lizzie doesn’t realize is that this also applies to my unlimited resource, 100% survival and reproduction scenario. There is no preference for one variant over another in that scenario, either, so there is no selection.)

    If the environment discriminates between variants, so that some variants have a better chance of grabbing a slot than others, then you have selection.

    Thus the AVIDA and Weasel no-selection cases really are no-selection cases, despite the fact that slots are limited.

  15. Keiths if you want to discuss unlimited resources scenarios and why they are not NS, start your own thread.

    This one is about AVIDA. And biology.

  16. Which is why my comment is about AVIDA, and how the finitude of the slots does not automatically mean that selection is happening, contra William.

    Lighten up, Lizzie.

  17. Well, don’t leave derail bait. I’ve yet to see an ID proponent actually grapple with the implications of AVIDA for the Irreducible Complexity argument.

  18. I’ve yet to see an ID proponent actually grapple with the implications of AVIDA for the Irreducible Complexity argument.

    IDers ( such as Marks and Dembski) claim that the program itself smuggles in information so any appearance of IC is an illusion. It has no bearing on the natural world

  19. Remind me again why the program rewards logic functions and not illogic functions?

  20. Mung,

    Remind me again why the program rewards logic functions and not illogic functions?

    To confuse you. It seems to be working.

  21. Mung,

    Remind me again why the program rewards logic functions and not illogic functions?

    Because it’s modeling evolution, not ID.

  22. If Elizabeth really wants to know what IDers think about Avida she could probably get an OP hosted over at UD. Maybe Salvador would host it for her. You know, what with him being “The Heretical IDist” and all.

    From the paper:

    Darwin realized that “organs of extreme perfection and complication”, such as the eye, posed a difficulty for his theory. Such features are much too complex to appear de novo, and he reasoned that they must evolve by incremental transitions through many intermediate states, some-times undergoing changes in function.

    I don’t understand why “organs of extreme perfection, such as the eye” posed any difficulty for his theory. iirc, Darwin didn’t think they posed any particular difficulty for his theory.

    Why would anyone think these “features” could not just appear de novo? After all, as we all know, highly improbable things happen every single day and twice on Thursdays.

    As for Avida, which logic functions underwent a “change in function”?

    And in answer to my question as to why the program rewards logic functions and not illogic functions, the answers all seem to invoke teleology (aka design).

  23. From the paper:

    To convey what digital organisms are, it is helpful to compare them with computer viruses. Digital organisms are self-replicating computer programs, as are viruses. However, computer viruses require direct intervention to mutate and evolve…

    What is the point here? Computer viruses require direct intervention to evolve but digital organisms do not? And from this we can conclude that digital organisms are not designed but that computer viruses must be designed?

    If a computer virus that was designed to evolve was created, what would be the impact, if any, on the conclusions drawn from Avida?

  24. Elizabeth:
    Well, don’t leave derail bait. I’ve yet to see an ID proponent actually grapple with the implications of AVIDA for the Irreducible Complexity argument.

    Oh the dilemma! A computer program which is written by humans, who programs all of the constraints and parameters of the program, which produces two dimensional images on a flat electric monitor. How does ID overcome these implications!

    You earlier made the absurd assumption that all parts of irreducibly complex systems have aspects which are of small reproductive advantages. Any thinking person should be able to take apart a system like our lymph glands, or respiratory system, and look at each individual part, and ask, what small advantage could alveolar ducts have without a complete respiratory system? And on and on. What good is a partial gland? What good is skin without pores, or pores without skin.

    Suggesting IR is so easy to overcome because of a dumb program is a complete fools card trick.

  25. Mung: What is the point here? Computer viruses require direct intervention to evolve but digital organisms do not? And from this we can conclude that digital organisms are not designed but that computer viruses must be designed?

    Yes. Digital organisms evolve by random mutation and natural selection within the environment provided.

    Mung: If a computer virus that was designed to evolve was created, what would be the impact, if any, on the conclusions drawn from Avida?

    None, but it would be a hell of a virus. I’m not sure that one hasn’t been done.

    Oh, look, yes, maybe.

    Do note, though, that to designa virus that evolves what you design is something that self-replicates with variance in its environment. In other words, what you provide is the OoL part and the environment part of the story, and then let Darwin’s mechanism do the rest.

    The initial population in AVIDA is designed. The can do one thing, which is that they can replicate. None of them can perform a logic function. The final population can do lots of logic functions, including the most complex. They were not designed to do that – they evolved to do it.

  26. phoodoo: Oh the dilemma! A computer program which is written by humans, who programs all of the constraints and parameters of the program, which produces two dimensional images on a flat electric monitor. How does ID overcome these implications!

    I asked you to consider the implications for Irreducible Complexity.

    phoodoo: You earlier made the absurd assumption that all parts of irreducibly complex systems have aspects which are of small reproductive advantages.

    I did no such thing. I said nothing like it. I don’t even know what your sentence means.

    phoodoo: Suggesting IR is so easy to overcome because of a dumb program is a complete fools card trick.

    What?

  27. The point, Mung, is very simple.

    Behe argued that a thing that was Irreducibly Complex could not evolve.

    His initial definition of an Irreducibly Complex thing was one which, if you took away any part, would no longer function.

    His argument was that if, say, you need five components to form a structure, and none of those components by themselves, or in sub-combinations, would give thing any advantage, then it would be very unlikely that those features would ever occur together (outside William’s infinitely reproducing world). That is why UD has an artist’s idea of a bacterial flagellum in its masthead.

    Other people argued, well, the precursors could be MORE complex, but not Irreducibly so, and the IC thing could evolve by subtraction, which is true (the “Arch” mechanism. Others argued that the precursors could have parts that served some other function (e.g. the Pallen and Matzke hypothesis for the bacterial flagellum). This is also true.

    However, what Lenski in AVIDA showed is that an Irreducibly Complex function (all the logic functions in AVIDA are IC), can evolve even without the precursors having useful functions, and without going via more complex but not-IC functions.

    At one point (sorry don’t have a reference) Behe changed his definition of IC to refer to the pathway not the object itself. He defined an “IC pathway” as one with many steps that were not advantageous. The more steps the greater the “degree” of IC. And argued that if the only evolutionary pathway to a function were via a high IC pathway, that function couldn’t evolve.

    Again, AVIDA shows that this is not the case. Because the lineages of each organism can be traced, the actual pathway can be examined (as it cannot in life, unless you do something like Lenski’s e-coli experiment of course, which Lenskii of course also did, and is still doing AFAIK).

    And not only is EQU IC by Behe’s first definition, the pathways by which it evolves are High Degree IC by his second. In fact even more stringently so – it turns out that a necessary step (well it seems to be necessary) on the way to EQU is quite sharply deleterious. In otherwords the immediate precursors to EQU are selected against.

    And yet EQU evolves.

    Darwin didn’t know this. He thought IC would finish his theory (hence his famous quotation about the eye). That’s because he didn’t know about drift (not surprisingly as he knew nothing about genetics, let alone population genetics).

    But he was wrong. As long as SOME of the steps in AVIDA are advantageous (the simpler functions are rewarded), many many neutral steps and deleterious steps can occur between one advantageous step and the next even more advantageous steps.

    And some functions, like EQU, appear only to be reachable via a sharply deleteriou step.

    And lest people object: “Ah! Told you so! EQU can evolve, but only at the cost of some other function! It’s devolution not evolution!” – no, that is not the case. Having evolved, the organisms that can perform EQU then go on to re-evolve the functions they lost to get there.

    So Behe’s principle (shared by Darwin) is falsified. The discovery that a function cannot have evolved by incremental consistently advantageous steps does not imply that it cannot have evolved.

    Turns out it can, and reliably does.

  28. Elizabeth: I asked you to consider the implications for Irreducible Complexity.

    I did no such thing.I said nothing like it.I don’t even know what your sentence means.

    What?

    Lizzie did you write this:

    “What I am saying, and what Darwin said, was that if those mutations result in slight reproductive advantage, they will aggregate.

    You might not believe it – but that’s what the theory says. I understand the theory.”

    This doesn’t mean exactly what I just said you claimed?? That all these little parts of the irreducible system came from mutations that each gave a small advantage then aggregated? How can you act like you said no such thing, and you have no idea what I am talking about?

    You don’t even understand the things you yourself write? Is it a definition problem? For crying out loud…

  29. phoodoo: This doesn’t mean exactly what I just said you claimed?? That all these little parts of the irreducible system came from mutations that each gave a small advantage then aggregated? How can you act like you said no such thing, and you have no idea what I am talking about?

    Because you didn’t say it very clearly. Yes, that’s what I said.

    However, population genetics tells us that to become common in a population, a mutation doesn’t have to confer an advantage in the current environment, or alone. It can become common simply by drift.

    And what AVIDA shows that that means that not every mutation necessary for some future function has to be advantageous at the time it first appears, or alone.

    So if, for instance, three separate mutations are necessary for an advantageous function, and none are advantageous, and some are even disadvantageous, when not combined with the other, the advantageous function still evolves with high probability, because the three necessary mutations can become quite prevalent simply through drift.

    So Darwin was correct that adaptive features evolve through the accumulation of incrementally advantageous variation over many generations. However, population genetics and AVIDA shows that those variations do all have to be advantageous to accumulate.

    The lineages of the AVIDA organisms that could perform EQU show that some advantageous steps were required, but that many intervening steps could be neutral or even deleterious.

  30. Elizabeth,

    Oh please, stop equating dots on a computer screen to what is necessary for advanced life systems. That is such a pathetic metaphor.

    If you want to argue that all the parts necessary for a highly specific, finely tuned system like an eyeball, or the central nervous system, can come about because the parts arising along the way are not really useful during the construction phase, but just so happen to come together to be useful one day, fine. Just take a bunch of spare parts lying around in a junk draw, and build a satellite, without even using blueprints. A little dried up superglue, a couple of broken paperclips, the top from a coke can, some dixie cups with holes in them…voila!, the space station!

    Some theory.

  31. phoodoo: Oh please, stop equating dots on a computer screen to what is necessary for advanced life systems. That is such a pathetic metaphor.

    Do you feel the same way about Behe’s mousetrap metaphor?

  32. phoodoo: Oh the dilemma! A computer program which is written by humans, who programs all of the constraints and parameters of the program, which produces two dimensional images on a flat electric monitor. How does ID overcome these implications!

    If you think that “written by humans” is particularly significant here, then you could jump to the conclusion that evolution was intelligently designed by God.

    If ID proponents were making the religious claim that God created evolution, instead of a false scientific claim that biological creatures were intelligently designed and not a product of evolution, then you would see far less push-back.

  33. I have a problem understanding how this refutes Behe’s IC.

    Behe does not say that IC systems cannot evolve, he states that they cannot evolve by a direct Darwinian route. I must admit that I don’t really know the ins and outs of the working of Avida, so I’m sure my misunderstandings will be explained to me.

    Behe states that it is possible for IC systems to be built up using co-option. Isn’t that what is going on in Avida?

    Also, in Avida, the separate systems and sub systems may produce different logical operations, but these systems are all of like kind and so can be easily combined to make a more complex system. They may produce different logical operations but they all have the same “function” of producing a binary string. The components of Behe’s IC systems can comprise parts with vastly different functions. For example in the flagellum the function of the secretory system is to transport the protein parts through the cell membrane while the function of the flagellar tail is to provide thrust. These components do not naturally fit together unless they are designed to do so. You cannot take a lawn mower engine, fit it to a truck and expect it to work as a means of transporting a payload.

  34. CharlieM: You cannot take a lawn mower engine, fit it to a truck and expect it to work as a means of transporting a payload.

    Therefore ID.

  35. CharlieM: Behe does not say that IC systems cannot evolve, he states that they cannot evolve by a direct Darwinian route.

    Again, I have to ask if you have a point.

  36. petrushka,

    Yes. Elizabeth stated, “Behe’s claim was that Irreducibly Complex systems couldn’t evolve”. This is inaccurate

  37. CharlieM:
    petrushka,

    Yes. Elizabeth stated, “Behe’s claim was that Irreducibly Complex systems couldn’t evolve”. This is inaccurate

    Well, he hedged his bets and said “unlikely”. Same difference really.

    “An irreducibly complex system cannot be produced directly (that is, by continuously improving the initial function, which continues to work by the same mechanism) by slight, successive modifications of a precursor system, because any precursor to an irreducibly complex system that is missing a part is by definition nonfunctional. An irreducibly complex biological system, if there is such a thing, would be a powerful challenge to Darwinian evolution.”
    Behe 1996 p. 39.

    “Irreducibly complex systems appear very unlikely to be produced by numerous, successive, slight modifications of prior systems, because any precursor that was missing a crucial part could not function. Natural selection can only choose among systems that are already working, so the existence in nature of irreducibly complex biological systems poses a powerful challenge to Darwinian theory.”
    Behe 2002.

    http://www.talkorigins.org/faqs/comdesc/ICsilly.html#footnote3

    It can hardly pose a powerful challenge to Darwinian theory if it could have evolved, can it?

  38. Actually, CharlieM, Top Gear is an entertainment show.
    Check out some serious cars or an example of lawnmower-payload
    Actually, my point was that you, along with every ID-proponent I know of, fail to appreciate the usefulness of “better than the current competition”, and rather insist that an adaptation must be optimized before it can offer any benefit at all.
    This is wrong.

  39. DNA_Jock,

    Your hybrid car is making my point for me. It is nowhere near a simple plug in conversion. Just look at all the modifications that must be carried out before the lawnmower engine can become part of the power generation system. And a small car is hardly a truck.

    As for the flagellum:
    Ishiguro K., “Revealing the mystery of the bacterial flagellum: A self-assembling nanomachine with fine switching capability,” Interview of Keiichi Namba, Japan Nanonet Bulletin, No. 11, February 5, 2004:

    Nature created a rotary motor with a diameter of 30 nm. Motility of bacteria, such as Salmonella and E. coli with a body size of 1 ~ 2 micron, is driven by rapid rotation of a helical propeller by such a tiny little motor at its base. This organelle is called the flagellum, made of a rotary motor and a thin helical filament that grows up to about 15 micron. It rotates at around 20,000 rpm, at energy consumption of only around 10-16 W and with energy conversion efficiency close to 100%. Prof. Namba’s research group is going to reveal the mechanism of this highly efficient flagellar motor that is far beyond the capabilities of artificial motors.

  40. CharlieM: Yes. Elizabeth stated, “Behe’s claim was that Irreducibly Complex systems couldn’t evolve”. This is inaccurate

    But I’m not asking about what Lizzie said of what Lizzie thinks.

    I’m asking you if you have a point. Are you agreeing that evolution can result in complex adaptions, or are you arguing against this?

  41. CharlieM: Your hybrid car is making my point for me.

    Well, I guess we`re even then
    🙂
    To wit:
    That`s a really impressive flagellar motor. But why bring its efficiency up, unless you believe that
    either

    a less efficient rotor would not confer a selective advantage

    or

    evolution is incapable of optimizing the rotor

    ?
    You appear to be suffering from the “What You See Is All There Is” fallacy [Kahneman].
    As I wrote:

    Actually, my point was that you, along with every ID-proponent I know of, fail to appreciate the usefulness of “better than the current competition”, and rather insist that an adaptation must be optimized before it can offer any benefit at all.
    This is wrong.

    Thanks for the confirmation.
    🙂

  42. Elizabeth: The initial population in AVIDA is designed. The can do one thing, which is that they can replicate. None of them can perform a logic function. The final population can do lots of logic functions, including the most complex. They were not designed to do that – they evolved to do it.

    They were designed to do that. I have no idea why you think design and evolution are mutually exclusive.

Leave a Reply