Of Whales and Timescales
UPDATE 05/27/19: The discussion of the paper “The Time Scale of Evolutionary Innovation” has been clarified and corrected. In addition, a footnote has been added with further discussion of Durrett and Schmidt’s paper.
Joshua Swamidass, assistant professor in the Department of Pathology and Immunology at Washington University, has responded to an Evolution News article about whale evolution. The original article concluded:
We don’t find the “pattern” that evolution predicts “should be found in the fossil record at certain times.” Rather, we find that truly aquatic whales appear abruptly. And even if we accept some of the fossils as “intermediates” between whale and land mammals, there is not enough time for the complex adaptations needed for whales’ fully aquatic lifestyle to evolve. Whatever the correct explanation is for the origin of whales, unguided evolutionary mechanisms are not the answer.
Looking at this progression [of skulls] we uncover an amazing fact. Surprisingly, whales have the same body plan as a terrestrial mammal! It’s the same body plan, with several intermediate forms. Looking at several features (e.g. ears, bone density, teach), we can see this transition beautifully. Look how we can see the nostrils slowly move back to the top of the head…
Yes, it is beautiful. One adapted for land, another for water, and one is intermediate. But take care; nothing is actually moving in those pictures. Any transition is in the interpretive imagination of the beholder.
Getting back to the claim that millions of years is “not enough time.” There is no genetic or mathematical analysis to back up this conjecture. What types of genetic changes are required for whale evolution? How unlikely or likely are they?
Consider a paper published in PLOS Computational Biology, “The Time Scale of Evolutionary Innovation.” The authors explore how long it should take for evolution to make a complex coordinated change to a sequence. They find that mutation alone would be little different from creating a completely fresh sequence each time using random letters, but that if natural selection is acting to “regenerate” the original sequence, and if the original sequence happens to be near the target, then evolution is much more likely to make the transition. This should be common sense, I think. Note a key result: a sequence of length L requiring only k specific coordinated changes will require at most Lk+1 trials. They describe this as “polynomial” because it is polynomial in L but it is exponential in k, and that means the time required grows very quickly with just a little bit of complexity. For example, analysis by Durrett and Schmidt calculated that the time required for just two specific coordinated mutations to co-occur in our human ancestors (assuming neither is harmful on its own) would have been over 200 million years. Mike Behe then found a factor of 30 they missed, implying it would take more like 6 billion years.*
Now consider the mutations that actually have occurred in humans in recent human history. Some have been interesting, including significant tweaks to melanism, and milk digestion, but none of them are spectacular (no X-Men) and certainly none have constructed new biochemical systems or new healthy morphology. The waiting times problem informs us that all the mutations in the history of the human species must have been similarly banal. Think about that for a moment. Are you surprised? You should be if you believe we evolved from something like an ape. Evolution has to work one step at a time. It cannot do the kind of complex coordinated magic that a human designer or engineer can. If you believe humans evolved, you have to believe it can happen without any complex coordinated changes at all (in this context complex means just 4 or more specific letters at the same time — not 4 new proteins). In fact, the exponential character of the waiting times problem tells us that all of evolution must have been similarly limited, right back to the Cambrian explosion. In turn, that raises the question of how the radical innovations of the Cambrian explosion could have occurred.
From a Batmobile to a Yellow Submarine
The Evolution News article argued that for a land mammal to become a whale, or a Batmobile to become a Yellow Submarine, it would require multiple coordinated changes. To many people this would be a trivial and common sense assumption, even without the detail given in the article. However, citing another paper, “Molecular evolution tracks macroevolutionary transitions in Cetacea,” Swamidass pushes back:
It is remarkable how many of the changes required for whale evolution are caused by loss of function mutations (which end causing “pseudogenes”), or small tweaks to proteins. This is one of the big surprises of mammalian evolution. Large changes can take place with tweaks to the genetic code. Eyes adapt to underwater vision by losing a rhodopsin gene. Hind Limbs are lost with the loss of a homeobox gene. Taste buds are lost when two genes are lost. Smell receptors are almost entirely lost in most species too. In all these cases, we see remnants of the broken genes, and in many cases the details of how these losses increase function are well understood.
This is all true. It is true that overall efficiency can be increased by losing unused functional components. On a design view, it makes sense to deactivate things that are not being used. Often this needs no more than a flip of a switch, and this no great challenge for evolution either. On the other hand, why would the random loss of information make a functional body plan? Researchers have created legless mice by knocking out a Hox gene, in an effort to understand snakes, but the resulting mice were simply paralyzed and could not mate. Also, note that at some point evolution has to explain the origin of all the proteins, Hox genes, as well as the rhodopsins and receptors that have been lost, and that is rather more difficult. An evolutionary process that creates nothing new is soon going to run out of other organisms’ proteins to borrow.
Remarkably, it does not appear any new enzymes or de novo genes are required in whale evolution. It appears that small tweaks to existing proteins, or loss or alteration of the function of existing genes, account for the changes we see at this point.
True, that is not where the challenge to whale evolution lies. But why is it remarkable to see no new genes? It turns out that a large number of genes are taxonomically restricted or ORFan genes. That means they seem to appear without evolutionary history in the twigs and leaves of the tree of life. Moreover, some even turn out to be essential, which would be very odd if they have been added last by evolution. The existence of these genes is a common problem elsewhere in the evolutionary story, even though it appears not to be relevant to whales. Protein coding genes are hard to explain when they appear de novo — see Doug Axe’s work as well as this recent EN article .
Also, there does not appear to be any reason that a large number of these changes must happen at the same time. They appear gradually in the tree, and it’s not clear at all why they would need to be “coordinated”. They do not appear to need to occur at the same place and time to be useful. So this does not make these transitions unlikely.
This is where we have to disagree. Some of the changes listed might be independent, small, and easy, but there are also some pretty massive, complex ones that would seem to need coordination. For one example, losing tooth enamel does not make baleen plates. For another, whale testes are inside the body. In itself this appears to be a trivial change, and it makes good design sense in terms of streamlining. That is, until you try to implement it, and find that mammalian testes become infertile if kept too warm, so now you need a cooling system, or else a redesign of the reproductive system. It is not so trivial any more. And the changes do need to be coordinated. What selective advantage is there in a cooling system? None unless you have testes there. What selective advantage is there in internal testes? None, unless a cooling system is there. It turns out that dolphins and whales have mysteriously acquired an elaborate counter-current cooling system that keeps the testes the same cool temperature as its fins! That system is not trivial, and it is not going to evolve with just one or two mutations.
A Design Perspective
Considering this from a design perspective, and speaking as one experienced in doing design, it’s going to be tough to convince me that one could “evolve” a program with a series of single-letter changes, deletions, and random copy-pasting, all while it continues to compile and function. The notion is in conflict with our experience of how complex functional systems actually work.
We have also argued that homoplasies constitute evidence of common design. Swamidass argues these can be explained by convergent evolution.
Also, we also see convergent mutations between whales, bats (echolocation), and beavers (diving adaptations to blood). These “homoplasies” are the rare exceptions to the nested clade pattern of common descent, and are exactly what we expect in evolutionary process, just like we see recurrent mutations in cancer, and convergent evolution in human HLA variation. Everyone agrees that human variation arises by natural processes, and that cancer arises by natural processes, yet we see homeoplasies here too; this is what we expect from common descent.
But another way of looking at it is that evolutionists have adapted their expectations in response to the evidence that homoplasies exist. They have long known about character traits that don’t fit the canonical Darwinian explanation of a branching tree, and convergent evolution has long been proposed as an explanation, but the truth is the authors of the original paper on bats and whales found this particular result “surprising” and “remarkable.” From my own experience, I remember trying to persuade an evolution-evangelist that molecular homoplasies exist between extremely distant species, and he wouldn’t believe me! I will try to explain why.
Now, convergent evolution can happen, but it really depends on the particular circumstances. Convergence in cancer and HLA are different from convergence between bats and dolphins because they both involve very high mutation rates, coupled with strong selection effects acting on very small changes. In HLA the changes are concentrated in a tiny region of the genome. In cancer, strong positive selection acts on mutations that each help the cancer, but destroy some normal function.
It is one thing to find weak points where cars tend to break independently in the same place, or even how one breakage leads to another (e.g., brakes first then everything else as it careens of the road). It would be quite strange if cars independently acquired sonar capabilities, and even more so if the software upgrades were identical.
Imagine you assign a coding exercise to computer-science students. It is quite possible that two students would come to roughly the same solution, since there are likely only a few good solutions. That is how convergent evolution is supposed to work. However, what if they had not only the same general solution, but identical code too? Or imagine two history students write an essay about the causes of World War I. It is possible they come to the same conclusion. But if you see identical prose, you have to suspect plagiarism: it strongly suggests that the text or code was designed once and then used multiple times.
And that is why my friend did not believe me; because he did not believe the coding sequences would converge. It is easy to imagine, if evolution could find complex solutions at all, that it could find something similar again, but it is much harder to imagine that evolution would converge on the same code, since the mutations that write it are supposed to be random, especially if the code has been diverging for some time. However, now that we have found that there are molecular homoplasies at great taxonomic distance, the committed evolution-believers are surprisingly unfazed: all it means, they argue, is that there must be only one solution that works and natural selection finds it every time (while listening to Wagner). That’s an interesting theory, but can you prove it? If the evolution of complex traits is so predictable and reliable, it seems we should be able to set it up and see it happening.
Meanwhile, can you hear the students accused of plagiarism? But sir, it’s the only solution! And we are both geniuses! Hmm. If you are both geniuses, I look forward to your next assignment.
There is a much more parsimonious solution: common design.
Either way, the larger problem is that the changes involved in adapting a generic mammalian template into a whale are certainly not all simple, independent, single-letter changes. It seems obvious that multiple coordinated changes would be needed, and it turns out that would require a lot longer than mere millions of years.
*By the way, there is much more to say about this topic. Durrett and Schmidt’s calculation involves a very interesting second-order process called ‘stochastic tunnelling’ in which a second mutation occurs before the first one is fixed. However, they also claimed Mike Behe had performed an incorrect calculation. In the end it turned out to be the other way around. The key difference between Mike Behe’s real-life chloroquine resistance (CQR) case, and Durrett and Schmidt’s theoretical double mutation scenario, is that CQR is significantly more difficult to evolve: it appears that the intermediate mutations are not neutral or harmless. One of the mutations, K76T, has been shown to be very harmful on its own, and it seems likely the other one would be as well. Even with their own model, the timescales Durrett and Schmidt found were far too long to have occurred during human evolution. This lead them to seek an escape route by arguing that changes need not be so specific, but could have occurred in a larger region, perhaps anywhere in the genome. Mike Behe pointed out this move is not biologically justified but is ‘gratuitous multiplication of probabilistic resources.’ You can follow the full exchange here, here, here and here.
Photo credit: t4berlin, via Pixabay.