Doug Axe and I share a number of similarities: we both profess faith in Jesus Christ; we both did our graduate training in cell/molecular biology in the late 1990s and early 2000s; we both rejected and opposed evolution during that time; and we both embraced the ideas of the Intelligent Design (ID) movement as a better explanation for the biological complexity that we saw.

Later, however, our paths diverged: I became convinced that the evidence supporting evolution is sound, and that ID arguments are not. Axe, as is clear from Undeniable, remains within the ID camp. We’ve both gone on to write books for lay audiences, but with very different conclusions.

Common Sense, Common Science

There is much that I could say about Undeniable. I found it a problematic book for many reasons, but I also found it interesting for the window it provides into Axe’s background and motivations. Like ID books in general, Undeniable is not aimed at experts in biology to convince them of the validity of ID. In fact, Undeniable is unapologetically aimed at non-specialists. Axe expends significant effort to convinceHumans are not well suited to have “common sense” intuitions about phenomena that are outside of our range of everyday experiences. his readers that their “common-sense” intuition that evolution just can’t work is in fact the right approach. This is a very odd tack for a scientific movement to take, to say the least—but it makes excellent sense for an apologetics movement.

Humans are not well suited to have “common sense” (or to use Axe’s term, “common science”) intuitions about phenomena that are outside our range of everyday experiences. We simply do not have the mental tools to intuit what might happen geologically over millions of years, what principles might be in play at speeds close to the speed of light, and so on. Quantum mechanics, to this non-specialist, looks like a violation of basic reason, but my colleagues in physics tell me that my intuitions are not valid. Similarly, what a process of evolution over billions of years might (or might not) be able to accomplish needs to be adjudicated based on evidence, not intuition. The church has enough of an anti-intellectualism streak within it already without Axe fanning that flame. How this sort of approach appears to outsiders should also be a concern if we value having a credible witness to the scientific community.

Protein Folding and Evolution’s Limits

Axe’s main contribution to the ID movement has been his work on protein folding and what he sees as the limits of what evolution can accomplish to produce new protein shapes (folds) and functions. In a nutshell, Axe argues that (a) proteins need to fold up into specific shapes in order to function, (b) the proportion of functional, folded proteins among the total number of possible protein sequences is vanishingly rare, on the order of 1 in 1074 , and therefore (c) evolution is not capable of producing new proteins with new functions, because the “search target” (functional, folded proteins) is too small for a chance-dependent process like evolution to find within the staggeringly vast “search space” of all the possible protein sequences out there.

This line of argumentation is very significant for ID. Here’s how Vincent Torley, a philosopher sympathetic to ID but nonetheless someone who critically examines its claims, sees Axe’s argument in his own (lengthy, but very interesting) review of Undeniable:

The importance of this particular argument to the case for Intelligent Design cannot be over-emphasized. Putting it succinctly: if it fails, then we’re back at square one, in terms of building a mathematical case for ID. Dr. Stephen Meyer’s two Intelligent Design best-sellers, Signature in the Cell and Darwin’s Doubt, are built on the bedrock foundation of this argument: their whole case would collapse without it. The same goes for The Design Inference, by Dr. William Dembski and Dr. Jonathan Wells. Speaking for myself, I can’t count how many times I’ve cited Dr. Axe’s argument… Whenever I’ve had doubts about Intelligent Design, this argument has always been my shining star.

Torley explains why he has come to have doubts about the validity of Axe’s argument, and it’s well worth the read.

Axe bases his estimate of the extreme rarity of functional, folded proteins on a paper he published in 2004, the backstory of which is part of the narrative of the book. The general contour of the argument—proteins need to be folded to have function, and stable folds are rare—was the prevailing view of protein biology when Axe and I were in graduate school. Recently, I had cause to re-read a chapter on protein biology from the standard textbook from that time—Molecular Biology of the Cell, 4th edition—and was struck by theExtrapolating from a small sample size to such a vast number is always something to be done with caution. parallels between it and Axe’s argument. Axe’s goal with his 2004 paper was to show just how rare folded proteins are, and his work led him to his estimate of about 1 in 1074, as he discusses in Undeniable.

Now, there are good reasons to doubt that this is an accurate estimate of all possible functional, folded proteins. Axe used only one protein with one function as the “test bed” for his analysis, and extrapolated from that result. Extrapolating from a small sample size to such a vast number is always something to be done with caution. As such, biologists took Axe’s work as one estimate among several, but not the definitive estimate that Axe seems to suggest that it should be seen as. (In Undeniable, he laments that evolution, as a field, does not simply fold up shop in light of his results). Other estimates in the field, using other test proteins and techniques, do not show the extreme rarity of functional proteins that Axe estimated.

Recent Developments in Protein Folding

In the decade plus since Axe published his paper, other work in protein folding and function has been done. In light of this work, the consensus that Axe and I were taught as graduate students has changed in two important ways: we now know that proteins do not need to be stably folded in order to function, and we also know that functional proteins are not rare within sequence space. Both of these advances in our knowledge fatally undermine Axe’s thesis, and with it a key argument upon which much of ID is built. Let’s examine both of these developments in turn.

HarperOne, 2017

The test protein that Axe used for his estimate does require a stable, folded structure in order to do its function. Axe estimated the rarity of functional protein sequences by swapping in varied protein sequences into his test protein and noting how many could provide function, indicating that they were stably folded. One of the issues with this analysis is that protein sequences that are functional but do not have a stable, folded structure would be completely missed by this analysis. This was not thought to be a big issue in 2004, but it is known to be significant now: we now know that there are many proteins that are functional but not stably folded. Not surprisingly, these proteins—called “intrinsically disordered proteins”—were highly controversial when they were first described. Eventually, biologists came to accept them as real, and widespread, because of the evidence. So, the first claim in Axe’s argument—that proteins need to be stably folded in order to function—simply is not the case. Axe’s 2004 paper is thus a significant underestimate of the prevalence of functional proteins even beyond the above noted shortcomings, since it misses this class of proteins altogether.

The second development in the last decade is the widespread sequencing and comparison of genomes of all manner of organisms—from humans to flies to yeast. (As an aside, one strong impression I get when I read ID literature is that I’m in a parallel reality where comparative genomics doesn’t exist—the evidence just isn’t discussed, despite its absolute relevance to the topic at hand. Michael Behe is the rare exception in ID circles, since he (briefly) acknowledges and accepts the evidence that humans and chimpanzees share a common ancestor. (Axe does not interact with this evidence in Undeniable. But I digress). What we see in these data is strong evidence not only that species share common ancestors, but that new genes that code for novel, functional proteins can pop into existence from sequences that did not previously encode a protein. OneWe now know that proteins do not need to be stably folded in order to function, and we also know that functional proteins are not rare within sequence space. Both of these advances in our knowledge fatally undermine Axe’s thesis, and with it a key argument upon which much of ID is built. such example is a protein found in baker’s yeast called BSC4. This protein arose in the relatively recent past from a sequence that did not code for a protein at all – something that we can test by comparing this sequence in its genomic context in closely related species of yeast. Despite being a young gene, there is good evidence that it is functional. Recent work on its protein structure indicates that it is “sort of” folded: it has some regions that are more stable than others, and some regions that seem to be disordered. As such, it provides a picture of what a young gene might look like—and it very well might become more stably folded over time. The idea that a protein might be selected for its function in the absence of a stable fold and then become more stable through subsequent mutation just isn’t on Axe’s radar.

A second recent paper in protein evolution is also problematic for ID, because it addresses the question of functional rarity directly. Once the evidence for new protein genes being “born” from non-protein sequences became convincing because of comparative genomics, researchers have sought to do more direct estimates of how often a random sequence might be converted into a functional gene—either as a protein gene, or merely as DNA transcribed into RNA. Ideally, the best way to estimate the proportion of functional sequences within a random pool would be to make a large number of random sequences and test each of them individually for function. The technological capability to do this sort of analysis directly is now available, and a recent paper has done just that: made a large pool of random sequences and directly tested how many of them are functional in a model organism (a bacterium). The results were surprising: about 25% of random DNA sequences had a beneficial function that was subject to natural selection. In at least one case that the researchers investigated closely, the function came from a new protein sequence (rather than just from the DNA sequence being made into RNA). These results show that new proteins with selectable functions are easily accessible to evolution.

Put more plainly: if Axe’s thesis is correct, we would never observe even a single new functional protein arising. Since we see strong evidence for not just one but several—and this with the first few studies to directly address the question—we can conclude that the central scientific claim of Undeniable, and with it a major tenet of ID, is incorrect.