Kenneth Miller Resists Chloroquine Resistance

Part I: Introduction

Brown University biologist Kenneth R. Miller has posted a reply to my challenge to him to give a quantitative account for the extreme rarity of the origin of chloroquine resistance in malaria. I’m grateful to him for doing so. Although I strongly disagree with nearly everything he wrote, his essay gives the public a chance to see directly how one informed Darwinist reacts to a basic empirical challenge to the theory.

Last April a paper by Summers et al. (see my summary here) appeared in the Proceedings of the National Academy of Sciences confirming that at least two mutations to the protein PfCRT are required before chloroquine resistance appears in the malaria parasite, as I had surmised in The Edge of Evolution. I wrote a series of posts (here, here, and here) analyzing the result, and another one that challenged Kenneth Miller and PZ Myers (who had downplayed the unsettling implications of the result for Darwinism) to provide their own quantitative — not verbal — account for the extreme rarity of chloroquine resistance. Last month Miller posted an 11-page reply.

The first two and a half pages of the PDF version of Miller’s essay consist of stage-setting and throat-clearing. The last six pages are a reprise of his review of The Edge of Evolution and a defense of the evolutionary musings of University of Chicago biologist Joseph Thornton from my skepticism. I’ll deal with those later. Miller’s only response to my take on the importance of Summers et al. is in the section “Parasites and Drugs.” Although the section is less than three pages (including several large figures), as we shall see it includes a number of serious mistakes.

Unfortunately, Miller dodges my challenge to provide a quantitative account of the rarity of the origin of chloroquine resistance. I had asked him to “Please keep the rhetoric to a minimum.” Alas, to no avail. He cites no relevant numbers, makes no calculations — just words.

Miller begins the section by questioning whether the mutation K76T (which replaces a lysine, “K,” at position 76 with a threonine, “T”) in the protein PfCRT is important to its ability to transport chloroquine:

There is indeed one required mutation in the PfCRT protein, which is a change of an amino acid at position number 76 from lysine to threonine…. But Behe was dead wrong about it being “strongly deleterious.” In fact, it seems to have no effect on transport activity at all.

It’s nothing short of incomprehensible to claim that the K76T mutation has “no effect on transport activity.” Figure 2 of Summers et al. — the very paper to which Miller is referring — shows that one variant of PfCRT (dubbed “D39”) with a particular mutation (N75E) has no chloroquine transport activity. When the K76T mutation is added to it to make a double mutant (variant D32), it gains such activity. So it had no effect? Variant E1 has three mutations (none are K76T) but no activity. When K76T is added to those mutations to make the Ecu variant, activity appears. The only difference between the non-transporting and transporting variants is the addition of K76T, but Miller maintains it has no effect on transport activity?<\p>

Summers et al. did show in their Figure 4 that, on the background of the mature malaria resistant variant Dd2, which has a total of eight mutations, other amino acid residues could replace T at position 76 and retain activity. But that has nothing to do with the claim that a K76T mutation would have “no effect on transport activity” in a strain that is newly developing chloroquine resistance.

Miller’s claim that the paper shows K76T isn’t deleterious is equally unsupported. Summers et al. did not even try to test whether the K76T mutation is deleterious. The word doesn’t even appear in their paper. Rather, the workers were interested mostly in testing what mutations were required just for chloroquine transport activity.

To test if a certain mutation were itself “strongly deleterious” takes particular conditions. That mutation would at least have to be examined: 1) alone on the background of the wild-type sequence (that is, with no other mutations present; more about this later); and 2) in the relevant organism. Yet most of the work described in the Summers et al. paper used frog eggs (X. laevis) as a test system, not malaria (P. falciparum). Whether a mutation has “no effect on transport activity” in frog eggs says nothing at all about whether it would be deleterious to malaria.

Summers et al. did test a PfCRT variant (“D38”) that had only the K76T mutation. It did not transport chloroquine, showing that multiple mutations are needed — which is by far the most ominous result for Darwinism. But since the test system was frog eggs, that of course couldn’t determine whether K76T might have any deleterious effect in malaria. The authors did also test five PfCRT variants (encoded on plasmids) in malaria cells in the lab to see how they would affect the cells’ survival in the presence of chloroquine. But all had multiple mutations — not K76T alone — so that couldn’t determine possible deleterious effects of the single mutation. What’s worse, the malaria cells also retained their own, genomic, functional, wild-type PfCRT, which would likely mask any deleterious effects of a nonfunctional plasmid-encoded mutant protein.

One of the variants (“E2”) had three mutations, including K76T. The protein could transport chloroquine modestly well in frog eggs, but in malaria cells in the lab it didn’t help increase survival rate much above background. That result may possibly indicate that more than those three mutations are needed for net beneficial activity in the wild. But it says nothing about whether a single K76T mutation would be deleterious to malaria cells.

Miller’s reading of Summers et al. is seriously mistaken. Sadly, a person who can’t accurately report the results of a paper makes for an unreliable guide. I urge everyone who has sufficient background to read at least the disputed parts of Summers et al. Determine for yourself which account is correct.

Part II: The Very Neutral Kenneth Miller

Brown University biologist Kenneth R. Miller has posted a reply to my challenge to him to give a quantitative account for the extreme rarity of the origin of chloroquine resistance in malaria. I’m grateful to him for doing so. Although I strongly disagree with nearly everything he wrote, his essay gives the public a chance to see directly how one informed Darwinist reacts to a basic empirical challenge to the theory. This is the second in a series of four responses to it. See yesterday’s post here.

Above, I showed Miller’s claim that the K76T mutation in the chloroquine-resistance protein factor PfCRT had “no effect on transport activity” was simply wrong. Also wrong was his claim that Summers et al. had shown the single mutation not to be deleterious. Here I rebut several other statements of Miller’s concerning whether the mutation is selectively neutral. He writes:

In fact, a 2003 study recommended against using the K76T mutation to test for chloroquine resistance since that same mutation was also found in 96% of patients who responded well to chloroquine. Clearly, K76T wouldn’t have become so widespread if it were indeed “strongly deleterious,” as Behe states it must be. This is a critical point, since Behe’s probability arguments depend on this incorrect claim.

Wrong again. A mutation could easily be both deleterious by itself yet widespread in a population if an organism has acquired other, compensating mutations that block its injurious effects. In the case of malaria this could happen in the following highly plausible scenario. First, as chloroquine is deployed in an afflicted country, the initial ultra-rare two required mutations eventually appear together in one cell (the deleterious-by-itself K76T plus a second mutation that confers transporting activity) and are selected to survive in the presence of chloroquine.

After those first two required mutations get their proverbial foot in the door and allow the altered gene to increase in the population, subsequent additional helpful mutations could do two things: 1) improve chloroquine-transporting, and 2) compensate for the destabilizing effect of K76T and/or other mutations on the normal, required function of PfCRT. (Malaria cells cannot survive without that protein, even in the absence of chloroquine.) The next milestone is that, once that multi-mutated protein becomes widespread, chloroquine is rendered medically useless and is prescribed much less frequently in the geographic region. With diminished pressure from chloroquine, the protein can then back-mutate to improve its necessary native function and spread in the population, without regard to maintaining the now-unnecessary transporting ability.

Consistent with this scenario are all of the following: The study Miller cites (Vinayak et al. 2003) did not check for other, possibly compensating mutations in PfCRT, only for K76T. Chloroquine had been the drug of choice in India, the locale of the study, but was discontinued as the first-line drug way back in 1973 — thirty years before the study Miller cites — when chloroquine-resistant malaria became endemic (Farooq & Mahajan 2004). Work in Malawi (Kublin et al. 2003) and China (Wang et al. 2005) has shown that chloroquine resistant malaria are replaced rapidly by sensitive malaria once the drug is discontinued in a region, likely reflecting selective pressure to alter the PfCRT protein in the absence of chloroquine. Although the Malawi work showed simple replacement of the resistant strain by the original unmutated one, other reversion pathways may well be possible. In doing so the PfCRT from the study Miller cites could easily have retained K76T if it also retained compensating mutations.

The result — widespread, chloroquine-sensitive K76T PfCRT in a population — is medically and epidemiologically interesting, but would have nothing to do with how resistance originally arose. It shouldn’t need pointing out to professional biologists that the major unsolved problem for Darwinian evolution is to explain how a new beneficial property first appears under a new selective pressure — in this case, how chloroquine resistance arises from the wild-type protein — not how a feature decays once selective pressure is removed.

Here’s a homey analogy to help explain. Suppose you wanted to replace the support columns of an old porch. If you simply pulled one away, the porch roof might collapse. But if you first braced the roof with strong poles, the columns could safely be replaced. Ken Miller is in the position of someone who points to a braced porch under repair, and claims that shows removing a column from a normal, unbraced structure would not be harmful.

He writes:

A neutral mutation like this can easily propagate through a population, and field studies of the parasite confirm that is exactly what has happened.

In my last post I showed Miller’s claim that Summers et al. found K76T to be selectively neutral was incorrect. Above I showed that the study he cites, showing the mutation is found in chloroquine-sensitive malaria, is easily compatible with its being deleterious on its own. But is there any direct, positive, experimental evidence indicating whether a single, uncompensated K76T mutation is deleterious or neutral?

Yes, there is. As I wrote earlier, to see if a mutation is harmful by itself, at the very least you have to test it without any other mutations present in the relevant organism. Lakshmanan et al. did this carefully in the lab in 2005:

To test whether K76T might itself be sufficient to confer VP-reversible [chloroquine resistance] in vitro … we employed allelic exchange to introduce solely this mutation into wild-type pfcrt (in GC03). From multiple episomally transfected lines, one showed evidence of K76T substitution in the recombinant, full-length pfcrt locus (data not shown). However, these mutant parasites failed to expand in the bulk culture and could not be cloned, despite numerous attempts. These results suggest reduced parasite viability resulting from K76T in the absence of other pfcrt mutations. This situation is not reciprocal however, in that parasites harboring all the other mutations except for K76T (illustrated by our back-mutants) show no signs of reduced viability in culture. [Emphasis added.]

Frankly, it was never a good bet that the K76T mutation — a nonconservative change in a required protein that’s likely to be near a binding site — would be selectively neutral. The best relevant experimental evidence indicates that K76T is indeed strongly deleterious by itself in the wild-type protein, but not if compensatory mutations are present. Miller’s claim that the individual mutation is neutral is wrong.

Part III: The Many Paths of Kenneth Miller

Brown University biologist Kenneth R. Miller has posted a reply to my challenge to him to give a quantitative account for the extreme rarity of the origin of chloroquine resistance in malaria. I’m grateful to him for doing so. Although I strongly disagree with nearly everything he wrote, his essay gives the public a chance to see directly how one informed Darwinist reacts to a basic empirical challenge to the theory. This is the third in a series of four posts responding to it.

In the I showed Miller’s claim that the K76T mutation in the chloroquine-resistance protein factor PfCRT had “no effect on transport activity” was simply incorrect, and that his argument that the mutation is selectively neutral strongly conflicts with the best relevant experimental evidence. Here I deal with his discussion that the existence of several pathways leading from partial to full chloroquine resistance somehow mitigates the improbability of its origin from zero resistance, and the baleful implications for Darwinism.

Miller writes:

Directly contradicting Behe’s central thesis, the PNAS study also showed that once the K76T mutation appears, there are multiple mutational pathways to drug resistance.

And once you have jumped over the Grand Canyon, there are multiple pretty trails you can explore.

Here Miller’s mistake helps us to see the bad effects of the lack of quantitative rigor. He exclaims that there is not just one pathway from partial to full chloroquine resistance, there are “several mutational routes.” His italics indicate that he thinks this is very important. But a moment’s quantitative thought shows that the number of pathways makes precious little practical difference. Suppose you were given a choice of a billion trillion roads to travel, but were told that only one of them led to safety; the others all led to certain death. You would likely feel pretty pessimistic about your chances. But suppose someone came along to tell you that there are actually several paths — in fact, five of them! So your odds of finding a safe path home have jumped from one in a billion trillion to five in a billion trillion. Feeling better? I didn’t think so.

If the odds of finding something by one path are an astronomically unlikely 1 in 1020, then the odds of finding it by either of two equally probable ones are 2 in 1020 — still astronomically unlikely. Even with a hundred paths the odds would be a profoundly prohibitive 1 in 1018. So it turns out that Miller’s strongly emphasized point has little practical importance. To avoid being misled by our imprecise intuitions, it is necessary to be as quantitative as the data allow.

The problem is actually somewhat worse than the above discussion indicates. Here’s how. Suppose there were an enormous number of routes a traveler could take. Almost all lead to oblivion, but one or more (it’s unknown how many) lead to a particular safe destination. After counting many, many travelers embarking and arriving by whatever route they happened upon, we reliably determined that approximately one in a billion trillion arrived at the safe port. That means the odds of finding any safe route to the destination is one in a billion trillion. It does not matter if in reality there are a thousand routes or just one, the odds of finding one of them remain one in a billion trillion. That’s because our numbers are derived from statistics, not from some pre-conceived, theoretical way of arriving at the destination.

The same goes for chloroquine resistance. The number of 1 in 1020 against developing chloroquine resistance comes from estimating the number of malaria cells without resistance that it takes to produce and select one with resistance, no matter what genetic route is taken. So the number of routes that Miller emphasizes turns out to have no effect at all on the statistical likelihood of developing chloroquine resistance. Each route itself is actually less likely than the cumulative probability. All of the routes together add up to only 1 in 1020.

Miller writes:

In most of these [pathways], each additional mutation is either neutral or beneficial to the parasite, allowing cumulative natural selection to gradually refine and improve the parasite’s ability to tolerate chloroquine. One of those routes involves a total of seven mutations, three neutral and four beneficial, to produce a high level of resistance to the drug. Figure 4, taken from the Summers et al. paper, makes this point in graphic fashion, showing the multiple mutational routes to high levels of transport, which confer resistance to chloroquine.

A chain is only as strong as its weakest link. And a Darwinian pathway is only as likely as its most improbable step. It matters not a whit whether later helpful mutations are easily acquired. It matters only that some steps are exceedingly unlikely. All of the pathways in Miller’s Figure 4 (which is Figure 3 from Summers et al.) require K76T — the most difficult, daunting change — as the first or second step. The pleasantly colored crisscrossing arrows of the figure might distract a person’s gaze, but they are not intended to quantitatively represent the improbabilities of the transitions.

Here’s an analogy. Suppose to win a prize you had to match seven numbers. The last five can be any number between one and three. The first two numbers, however, might be anything between one and a billion. Of course it doesn’t matter that the last five are pretty easy to guess, or that you might be permitted to guess them in any order. The steps that overwhelmingly control your odds of winning are the first two, the most improbable ones.

Here’s another one. Suppose there were a nice, pretty, level meadow where a person could easily walk around, smelling the flowers — but the meadow was situated on the top of a sheer-cliffed butte. If a travel agent told you how easy it was to walk around the pretty meadow without mentioning the brutal climb it would take to get there you would rightly conclude that, whatever other admirable qualities he may have, he was an unreliable guide.

Let me also emphasize, if any of the pathways or other factors Miller discusses made much difference, then the odds of malaria developing chloroquine resistance would be better than they are known to be. Miller’s argument has both a quantitative and a conceptual problem. He agrees that the development of chloroquine resistance is an extremely rare 1 in 1020, but he doesn’t know why. He seems to really want one beneficial mutation to be available at a time, but the mutation rate in malaria is about a trillion times greater than the origin of chloroquine resistance. So why is the origin of resistance so rare? Resistance to another anti-malarial drug (atovaquone) arises de novo in nearly every infected person it’s given to. So why is de novo chloroquine resistance much, much, much less frequent?

That question is the bane of Miller’s perspective. Figure 3 of Summers et al. (Miller’s “Figure 4”) shows that it takes a minimum of two mutations for chloroquine transport function to appear, that before both of them appear there is zero activity. That is the big problem for the evolution of resistance. That is the reason why de novo chloroquine resistance appears so much less frequently than resistance to other anti-malarial drugs that require only one mutation.


There are indeed several mutational routes to drug resistance, and they are indeed the result of sequential, not simultaneous mutations.

“Sequential or simultaneous” is the wrong distinction. The only question relevant to Darwinian evolution is whether the helpful, selectable activity appears incrementally, with each additional mutation. Summers et al. shows that it doesn’t. There is zero chloroquine-transport activity until two mutations have occurred to the wild-type sequence. The relevant activity appears discontinuously, not incrementally.

Part IV: Kenneth Miller Steps on Darwin’s Achilles Heel

Brown University biologist Kenneth R. Miller has posted a reply to my challenge to him to give a quantitative account for the extreme rarity of the origin of chloroquine resistance in malaria. I’m grateful to him for doing so. Although I strongly disagree with nearly everything he wrote, his essay gives the public a chance to see directly how one informed Darwinist reacts to a basic empirical challenge to the theory. This is the last in a series of four posts responding to it.

In the previous three sections, I showed that Miller’s claims concerning the evolution of the chloroquine-transporting protein PfCRT of malaria were at best contrary to strong experimental evidence and at worst simply wrong. That section occupied less than a third of his essay. Now I’ll move on to the remaining two-thirds. Mercifully, that can be dealt with more briefly here, because in previous writings I have already answered almost all the objections he raises. For most of that final portion Miller simply calls on the prominent biologists Sean Carroll and Joseph Thornton for help. My answers to Carroll’s old arguments are here and here.

And to Thornton’s writings here, here, here, here, here, here, here, here, here, and here.

And while I’m at it, my responses to Miller’s and Richard Dawkins’s attack reviews of The Edge of Evolutionare here, here, and here.

Just one issue remains. In a section titled “Rigging the Odds” (the link to that section in HTML is more bluntly called “Fabricating the Odds”), Miller objects to my saying that if two new protein binding sites were required to evolve for some new, useful, selectable function, the likelihood would be about the square of the odds of one new selectable protein binding site evolving. I put the latter odds at about 1 in 1020, about the same as the odds of malaria developing chloroquine resistance (which I dubbed a “Chloroquine Complexity Cluster,” or CCC). The odds of two required sites evolving in my model would then be around 1 in 1040 — a very large number indeed, and what I argued was the “edge,” the limit, of Darwinian evolution. That would be a big problem for the theory since most proteins occur in cells as complexes of six or more.

Miller grants the value of 1 in 1020 for purposes of discussion (“Let’s accept Behe’s number of 1 in 1020 for the evolution of a complex mutation like his CCC”). But he balks at multiplying the odds for the development of two required sites to get the ultra-huge value of 1040, calling it a “breathtaking abuse of statistical genetics.” It’s more likely he lost his breath from the speed at which he switched models.

He points to what might happen with chloroquine and a second anti-malarial drug of the same efficacy:

Chloroquine resistance arose in just a decade and a half, and is now common in the gene pool of this widespread parasite. Introduce a new drug for which the odds of evolving resistance are also 1 in 1020, and we can expect that it will take just about as long, 15 years, to evolve resistance to the second drug. Once you get that first CCC established in a population, the odds of developing a second one are not CCC squared. Rather, they are still 1 in 1020. Behe gets his super-long odds by pretending that both CCCs have to arise at once, in the same cell, purely by chance. They don’t…

Miller shows here that he has simply misunderstood the central argument of The Edge of Evolution. In my book I stipulated that the two sites were interdependent, linked: “Now suppose that, in order to acquire some new, useful property, not just one but two new protein-binding sites had to develop.” The sites are defined as belonging to the subset of beneficial traits named “Only Selectable When Partner Site Has Also Evolved.” In my thought experiment, there is no selectable property in the presence of only one such site. I postulated that only when the two have developed do we get such an effect. (For example, suppose in Miller’s inapt illustration above the two drugs were chemically tied together — perhaps they were simply different regions of the same molecule. In that case resistance would have to be developed against both at once to do any good. And the likelihood of that would indeed be the multiple of the odds of developing resistance against each one separately.)

Miller’s incongruous response essentially is to say: “No, I have decided to change your own model. I will switch the premise to one in which each protein binding site will necessarily be beneficial by itself.” Much worse, he doesn’t tell his readers that’s what he’s doing. His writing leads them to believe he is describing the same situation as I did. Let me be clear, if Miller had simply said that he thought there would be no actual situations in nature like I modeled — that the subset was empty; that never in reality would two new protein binding sites be needed before any new selectable property resulted — then that would have been fine. We could have argued amicably about whether that was true. But he didn’t. Instead he conjured an entirely separate scenario, and then claimed it was I — not he — who was “fabricating,” trying to deceive readers with a “statistical trick”!

Miller’s efforts to divert readers’ attention from features that require multiple mutations follows inexorably from Darwinism’s profound Achilles’ heel. Let’s play off Miller’s two-antimalaria-drugs example to help see what it is. He wrote that, sure, the odds of malaria developing resistance to chloroquine are about 1 in 1020. But if a second drug came along the odds would still be 1 in 1020. They wouldn’t be multiplied, he said.

Well, we can note that the odds of developing resistance to the malaria drug atovaquone are only about 1 in 1012. We can then ask, why is the probability so much better for atovaquone than for chloroquine? And following Miller’s lead we can ask, if malaria developed resistance to atovaquone at a frequency of 1 in 1012, shouldn’t it subsequently develop resistance to chloroquine at 1 in 1012? Why not just another round of 1 in 1012? Why the jump to 1 in 1020?

Enter Achilles and his heel. It turns out that the odds are much better for atovaquone resistance because only one particular malaria mutation is required for resistance. The odds are astronomical for chloroquine because a minimum of two particular malaria mutations are required for resistance. Just one mutation won’t do it. For Darwinism, that is the troublesome significance of Summers et al.: “The findings presented here reveal that the minimum requirement for (low) CQ transport activity … is two mutations.”

Darwinism is hounded relentlessly by an unshakeable limitation: if it has to skip even a single tiny step — that is, if an evolutionary pathway includes a deleterious or even neutral mutation — then the probability of finding the pathway by random mutation decreases exponentially. If even a few more unselected mutations are needed, the likelihood rapidly fades away.

Without telling his readers, Miller switches from my model to one with the tendentious assumption that new protein binding sites would necessarily always be helpful on their own. (New protein binding — that sounds so nice. What could possibly go wrong?) Yet one mutation in the chloroquine-resistance protein isn’t helpful at all. In fact the best evidence indicates it is harmful on its own. Two mutations are needed before it’s helpful. So why should we think that just one binding site must always be helpful? Who made up that rule? The answer is that we have no particular reason to think it, and good reason to disbelieve it.

So what should we conclude from all this? Miller grants for purposes of discussion that the likelihood of developing a new protein binding site is 1 in 1020. Now, suppose that, in order to acquire some new, useful property, not just one but two new protein-binding sites had to develop. In that case the odds would be the multiple of the two separate events — about 1 in 1040, which is somewhat more than the number of cells that have existed on earth in the history of life. That seems like a reasonable place to set the likely limit to Darwinism, to draw the edge of evolution.

References for the series:

Farooq, U. and Mahajan, R. C. 2004. Drug resistance in malaria. J. Vector. Borne. Dis. 41:45-53.

Jucker, M. and Walker, L. C. 2013. Self-propagation of pathogenic protein aggregates in neurodegenerative diseases. Nature 501: 45-51.

Kublin, J. G. et al. 2003. Reemergence of chloroquine-sensitive Plasmodium falciparum malaria after cessation of chloroquine use in Malawi. J. Infect. Dis. 187:1870-1875.

Lakshmanan, V., et al. 2005. A critical role for PfCRT K76T in Plasmodium falciparum verapamil-reversible chloroquine resistance. EMBO J. 24:2294-2305.

Paget-McNicol, S. and Saul, A. 2001. Mutation rates in the dihydrofolate reductase gene of Plasmodium falciparum. Parasitology 122:497-505.

Summers, R. L. et al. 2014. Diverse mutational pathways converge on saturable chloroquine transport via the malaria parasite’s chloroquine resistance transporter. Proc. Natl. Acad. Sci. U. S. A 111:E1759-E1767.

Wang, X. et al. 2005. Decreased prevalence of the Plasmodium falciparum chloroquine resistance transporter 76T marker associated with cessation of chloroquine use against P. falciparum malaria in Hainan, People’s Republic of China. Am. J. Trop. Med. Hyg. 72:410-414.

White, N. 1999. Antimalarial drug resistance and combination chemotherapy. Philos. Trans. R. Soc. Lond B Biol. Sci. 354:739-749.

White, N. J. 2004. Antimalarial drug resistance. J. Clin. Invest 113:1084-1092.

Michael J. Behe

Senior Fellow, Center for Science and Culture
Michael J. Behe is Professor of Biological Sciences at Lehigh University in Pennsylvania and a Senior Fellow at Discovery Institute’s Center for Science and Culture. He received his Ph.D. in Biochemistry from the University of Pennsylvania in 1978. Behe's current research involves delineation of design and natural selection in protein structures. In his career he has authored over 40 technical papers and three books, Darwin Devolves: The New Science About DNA that Challenges Evolution, Darwin’s Black Box: The Biochemical Challenge to Evolution, and The Edge of Evolution: The Search for the Limits of Darwinism, which argue that living system at the molecular level are best explained as being the result of deliberate intelligent design.