Darwin's Doubt: The Explosive Origin of Animal Life and the Case for Intelligent Design

by Stephen C. Meyer

Whether in the analogies he drew to animal breeding or the computer simulations he used to demonstrate the supposed ability of mutation and selection to generate new genetic information, Dawkins repeatedly smuggled in the very thing he insisted the concept of natural selection expressly precluded: the guiding hand of an intelligent agent. He found Dawkins’s computer simulation particularly interesting. In The Blind Watchmaker, Dawkins described how he had programmed a computer to generate the Shakespearean phrase: “Me thinks it is like a weasel.”1 Dawkins did this in order to simulate how random mutations and natural selection could generate new functional information. Dawkins programmed the computer first to generate many separate strings (sequences) of English letters. He then programmed it to compare each string to the Shakespearean target phrase and select only the string that most closely resembled that target.2 The program then generated variant versions of that newly selected string and compared those sequences to the target, selecting, again, only the one that most closely resembled the desired target. This eventually generated—after many iterations—a string that matched the target perfectly. Axe recognized immediately the role that Dawkins’s own intelligence had played. Not only did Dawkins provide the program with the information that he wanted it to generate (“Me thinks it is like a weasel”), he imbued the computer with a kind of foresight by directing it to com...

If Sauer’s quantitative measures of rarity held up, then Axe thought it obvious that mutation and selection could not adequately search a space that large.

He was soon invited by Alan Fersht, a professor at the University of Cambridge and director of the Centre for Protein Engineering, part of the world-famous Medical Research Council (MRC) Centre at Cambridge, to join his research group.

Proteins comprise at least three distinct levels of structure:4 primary, secondary, and tertiary, the latter corresponding to a protein fold.

The specific sequence of amino acids in a protein or polypeptide chain make up its primary structure. The recurring structural motifs such as alpha helices and beta strands that arise from specific sequences of amino acids constitute its secondary structure. The larger folds or “domains” that form from these secondary structures are called tertiary structures (see Fig. 10.2).

But more often, proteins capable of performing new functions require new folds to perform these functions. That means that explosions of new life-forms must have involved bursts of new protein folds as well. FIGURE 10.2

The late geneticist and evolutionary biologist Susumu Ohno noted that Cambrian animals required complex new proteins such as, for example, lysyl oxidase in order to support their stout body structures. When these molecules originated in Cambrian animals, they also likely represented a completely novel folded structure unlike anything present in Precambrian forms of life such as sponges or one-celled organisms. Thus, Axe was convinced that explaining the kind of innovation that occurred during the Cambrian explosion and many other events in the history of life required a mechanism that could produce, at least, distinctly new protein folds.

the ability to produce novel protein folds provided a critical test for the creative power of the mutation and selection mechanism.

new protein folds could be viewed as the smallest unit of structural innovation in the history of life.

It follows that new protein folds represent the smallest unit of structural innovation that natural selection can select. Of course, natural selection can operate on smaller units of change—individual amino-acid changes that result in slight functional advantages or fitness gains, but not new folds, for example. But what if the functional or fitness gains that natural selection preserves and passes on never generate structural innovations? What if, instead, it only preserves slight differences in the sequence or function of proteins that confer an advantage without altering structure? Then, clearly, fundamental changes in the form of an organism will not occur. Building fundamentally new forms of life requires structural innovation. And new protein folds represent the smallest selectable unit of such innovation. Therefore, mutations must generate new protein folds for natural selection to have an opportunity to preserve and accumulate structural innovations. Thus, Axe realized that the ability to produce new protein folds represents a sine qua non of macroevolutionary innovation.

even though single mutations allow many proteins to retain some function, they still diminish or damage the function of the protein—often enough to ensure that they will be eliminated by the purifying effect of natural selection. Further,

mutations in preexisting genes will typically degrade functional genetic information.

when genes lose function, natural selection will eventually eliminate the organisms that possess these genes.

In work that he published in 2000, he showed that it is, indeed, exceedingly difficult to make extensive changes to functional amino-acid sequences without destabilizing a protein fold. Even best-case changes involving the most chemically similar amino acids in the exterior of proteins tended to destabilize protein folds.

Yet to turn one protein with a distinctive folded structure into another with a completely novel structure and function requires specified changes at many, many sites—far more than Axe altered in his experiments.10 The number of changes necessary to produce a new protein fold typically exceeds the number of changes that will result in functional loss. Given this, the probability of the evolutionary process successfully traversing a functional landscape from one functional peak to another—all the while escaping functional loss each step along the way—is extremely small, with the probability diminishing exponentially with each additional requisite change.

what most evolutionary biologists suspected—namely, that protein-to-protein (or functional gene–to–functional gene) evolution is a no-go where the mutation and selection mechanism must produce a new protein fold (see Fig. 10.3).

Based on the physical principles of protein function, the vast majority of protein functions simply cannot be performed by unfolded proteins. In other words, stability of protein structure is a precondition of protein function. Destabilized protein folds not only lose the three-dimensional structures they need to perform functional tasks, they are also vulnerable to attack by other proteins called proteases that devour unfolded proteins or polypeptides in the cell.12 As one structure is degraded as the result of multiple sequence changes, it will necessarily lose structural stability, resulting in a catastrophic loss of function.

Indeed, according to the equations of population genetics, the standard mathematical expression of neo-Darwinian theory, even slight losses in fitness will subject the disadvantageous traits that produce such losses to purifying selection, thus eliminating them. That means that even many protein sequences that retain a significant, though diminished, portion of their original function nevertheless will not survive the winnowing effects of the neo-Darwinian mechanism. Thus, the gradual transformation of one functional fold into another was a complete nonstarter.

This figure illustrates why many evolutionary biologists reject the idea that genes and proteins under selection pressure will evolve into new functional genes and proteins. Since genes, like English sentences, contain sequence-specific functional information, multiple changes in the genetic text will inevitably degrade function (or fitness) long before a new functional sequence will arise—just as random changes in a meaningful English sentence will typically destroy meaning long before such changes produce a significantly different meaningful sentence.

Thus, both experimental results and the physics of protein folding implied that random searches for novel proteins starting from preexisting protein-coding genes would result in functional loss long before a protein with a novel fold would emerge, as most evolutionary biologists already suspected.

randomly mutating the gene will soon destabilize a protein fold and/or generate nonfunctional intermediate sequences and structures long before a new gene (capable of generating a new fold) would arise. For this reason, this scenario involves not so much a climb up Mount Improbable, but a step out over Valley Impassable.

For all these reasons, like most evolutionary biologists, Axe thought the second neo-Darwinian scenario—in which new genes and proteins emerge from nonfunctional or neutral regions of the genome—provides a much more plausible means of producing the information necessary to construct novel protein folds. It was to this scenario that Axe turned his experimental energies. In this scenario, neo-Darwinists envision new genetic information arising from sections of the genetic text that can vary freely without consequence to the organism. According to this scenario, noncoding sections of the genome or duplicated sections of coding regions undergo a protracted period of “neutral evolution”15 in which alterations in nucleotide sequences have no discernible effect on the fitness of the organism. Functional genes and proteins gradually rise from a nonfunctional valley floor to a functional mountain peak—generating a new gene. Natural selection plays a role, but not until a new functional gene has arisen.

When this occurs, one of the two genes may begin to vary—to experience mutations—without adversely affecting the function of the organism, while the other performs the original function. In the jargon of evolutionary biology, mutational changes in gene duplicates are “selectively neutral”—they initially provide no advantage or disadvantage to an organism or population. These gene-duplication events allow nature room to experiment safely. Unhelpful but harmless genetic novelties can be passed on to future generations, where additional mutations one day may render the evolving genetic material useful. Eventually, as mutational changes accumulate, a new gene sequence may arise in a new organism that can code for a novel protein fold and function. At that point, natural selection can favor the new gene and its protein product, preserving and passing it along to future generations—or so the story goes. This scenario—which many evolutionary biologists now refer to as the “classical model” of gene evolution—has the advantage of allowing portions of the genome to vary freely through many generations, giving mutations many opportunities to “search” the space of possible base sequences without being punished for drifting into valleys of lost or diminished function. But this scenario faces an overriding problem: the extreme rarity of sequences capable of forming stable folds and performing biological functions. Since natural selection does nothing to help generate new folded, functio...

In other words, for sequences 150 amino acids long, only 1 in 1074 sequences will be capable of folding into a stable protein.

natural selection cannot select for the presence of a fold unless it also performs a function that confers a specific functional advantage on an organism.

Thus, Axe also estimated (a) the number of proteins of modest length (150 residues) that perform a specified function via any folded structure compared to (b) the whole set of possible amino-acid sequences of that size. Based on his experiments and data about the number of stable folded proteins that exist, Axe estimated that ratio to be about 1 to 1077. A telling conclusion follows from this experimental data: The probability of any given mutational trial generating (or “finding”) a specific functional protein among all the possible 150 residue amino-acid sequences is 1 chance in 1077—that is, one chance in one hundred thousand, trillion, trillion, trillion, trillion, trillion, trillion.

they refer to the number of opportunities an event has to occur as “the probabilistic resources.”18

How then should we assess the chance hypothesis for the origin of biological information—in particular, the hypothesis that random mutations generated the information necessary to produce a novel protein fold with a selectable function? What is the conditional probability that such a folded protein could arise as the result of random mutations in duplicated nonfunctional sections of a genome?

only those mutations in the genes (or DNA) in the reproductive cells of parent organisms can have any effect on the next generation.

the first appearance of bacterial life on earth (3.8 billion years ago), scientists have estimated that a total of about 1040 organisms have lived on earth since life first appeared.19

if every organism from the dawn of time had generated, by random mutation, one new base sequence in the sequence space of interest, that would amount to only one 10 trillion, trillion, trillionth of the sequences in that space—the space that needs to be searched.And, since conditional probability of a new gene arising in the manner envisioned by the classical model turns out to be almost unimaginably less than ½, the classical model turns out to be vastly more likely to be false than true.

The probabilistic resources available to the classical model of gene evolution are simply far too small to tame 1 chance in 1077 (see Fig. 10.4).

in the classical model of gene evolution, random mutations must thrash about aimlessly in immense combinatorial space, a space that could not be explored by this means in the entire history of life on earth, let alone in the few million years of the Cambrian explosion.

building new animal forms requires generating far more than just one protein of modest length. New Cambrian animals would have required proteins much longer than 150 amino acids to perform necessary, specialized functions.

In addition to a novel protein fold, these molecules (in living organisms) comprise over 400 precisely sequenced (nonrepeating) amino acids. Reasonable extrapolation from mutagenesis experiments done on shorter protein molecules suggests that the improbability of randomly producing functionally sequenced proteins

The Cambrian animals exhibit structures that would have required many new types of cells, each requiring many novel proteins to perform their specialized functions. But new cell types require not just one or two new proteins, but coordinated systems of proteins to perform their distinctive cellular functions. The unit of selection in such cases ascends to the system as a whole. Natural selection selects for functional advantage, but no advantage accrues from a new cell type until a system of servicing proteins is in place. But that means random mutations must, again, do the work of information generation without the help of natural selection—and, now, not simply for one protein, but for many proteins arising together. Yet the odds of this occurring by chance alone are, of course, far smaller than the odds of the chance origin of a single new gene or protein—so small as to render the chance origin of the information needed to build a new cell type fantastically improbable (and implausible) given even the most optimistic estimates for the length of the Cambrian explosion.

The sensitivity of proteins to functional loss, the rarity of proteins within combinatorial sequence space, the need for long proteins to build new cell types and animals, the need for whole new systems of proteins to service new cell types, and the brevity of the Cambrian explosion relative to rates of mutation—all conspire to underscore the immense implausibility of any scenario for the origin of Cambrian genetic information that relies upon random variation alone, unassisted by natural selection.

If only one out of every 1077 of the alternate sequences are functional, an evolving gene will inevitably wander down an evolutionary dead-end long before it can ever become a gene capable of producing a new protein fold.

Douglas Axe’s results highlight an acute dilemma for neo-Darwinism, a “catch-22.” On the one hand, if natural selection plays no role in generating new genes, as the idea of neutral evolution implies, then mutations alone must climb a Mount Improbable in a single leap—a situation that, given Axe’s results and Dawkins’s own logic, is probabilistically untenable. On the other hand, any model for the origin of genetic information that envisions a significant role for natural selection, by assuming a preexisting gene or protein under selective pressure, encounters other equally intractable difficulties. The evolving genes and proteins will range over a series of disadvantageous or nonfunctional intermediates that natural selection will not favor or preserve, but will, instead, eliminate. At that point, selection-driven evolution will cease, locking existing genes and proteins into place.

neo-Darwinism does not explain the Cambrian information explosion.

When I first heard that Douglas Axe had succeeded in making a rigorous estimate of the rarity of proteins in sequence space, I wondered what neo-Darwinists would say in response. Given the experimental rigor and mathematical precision of the work he reported in the Journal of Molecular Biology in 2004, and the long odds against mutation and selection ever finding a novel gene or functional protein, what could they say?

his paper had passed through the careful scrutiny of peer review,

The same year, I published a peer-reviewed scientific article about the Cambrian explosion and the problem of the origin of the biological information needed to explain it.1 In the paper, I cited Axe’s results and explained why the rarity of functional proteins in sequence space posed such a severe challenge to the adequacy of the neo-Darwinian mechanism. The article appeared in a biology journal, Proceedings of the Biological Society of Washington, published out of the Smithsonian Institution by scientists working for the Smithsonian’s National Museum of Natural History (NMNH). Because the article also argued that the theory of intelligent design could help explain the origin of biological information (see Chapter 18), its publication created a firestorm of controversy. Museum scientists and evolutionary biologists from around the country were furious with the journal and its editor, Richard Sternberg, for allowing the article to be peer-reviewed and then published. Recriminations followed. Museum officials took away Sternberg’s keys, his office, and his access to scientific samples. He was transferred from a friendly to a hostile supervisor. A congressional subcommittee staff later investigated and found that museum officials initiated an intentional disinformation campaign against Sternberg in an attempt to get him to resign. His detractors circulated false rumors: “Sternberg has no degrees in biology” (actually he has two Ph.D.’s, one in evolutionary biology and one in...

Nearly all of the scenarios developed in the papers that Long cites start with an inferred common ancestral gene from which two or more modern genes diverged and developed. These scenarios treat the similarity of sequence (the information) in two or more genes as unequivocal evidence for a common ancestral gene (see Fig. 11.2). As I noted in Chapters 5 and 6, standard methods of phylogenetic reconstruction presuppose, rather than demonstrate, that biological similarity results from shared ancestry. Yet, as we saw in Chapter 6, similarity of sequence by itself is not always an unequivocal indicator of common ancestry. Sometimes similarity appears between species where it cannot be explained by inheritance from a common ancestor (e.g., the similar forelimbs on moles and mole crickets) and, at the very least, there are other possible explanations for sequence similarity.

For example, molecular biologists have discovered that both whales and bats use similar systems—involving similar genes and proteins—for echolocation. The striking similarity of these systems used in two otherwise disparate mammalian species has led biologists to posit the parallel evolution of echolocation, including the gene sequences and proteins that make it possible, from a common ancestor that did not possess this system.14

it is possible that similar genes might have been separately designed to meet similar functional needs in different organismal contexts. Viewed this way, similarity of sequence does not necessarily reflect descent with modification from a common ancestor, but could reflect design in accord with common functional considerations, constraints, or goals. I recognize, of course, that to this point I have not given any independent reasons for considering the design hypothesis, and that, as a hypothesis for sequence similarity by itself, intelligent design may not seem compelling.

All of these scenarios attempt to explain the origin of two similar genes by reference to descent with modification (via mutation) from common ancestral genes. Yet genomic studies are now turning up hundreds of thousands of genes in many diverse organisms that exhibit no significant similarity in sequence to any other known gene.15 These “taxonomically restricted genes” or “ORFans” (for “open reading frames of unknown origin”) now dot the phylogenetic landscape. ORFans have turned up in every major group of organisms, including plants and animals as well as both eukaryotic and prokaryotic one-celled organisms. In some organisms, as much as one-half of the entire genome comprises ORFan genes.16

Thus, even if it could be assumed that similar gene sequences always point to a common ancestor gene, these ORFan genes cannot be explained using the kind of scenarios that Long’s article cites. Since ORFans lack sequence similarity to any known gene—that is, they have no known homologs in even distantly related species—it is impossible to posit a common ancestral gene from which a particular ORFan and its homolog might have evolved. Remember: ORFans, by definition, have no homologs. These genes are unique—one of a kind—a fact tacitly acknowledged by the increasing number of evolutionary biologists who attempt to “explain” the origin of such genes through de novo (“out of nowhere”) origination.

As scientists have explored and sequenced more genomes, they have discovered more and more ORFans without finding anything like a corresponding number of homologs. Instead, the number of “unpaired” ORFan genes continues to grow with no sign of the trend reversing itself.17

Duplication mutations and various other modes of random mutational change along with natural selection clearly constitute “causes now in operation.” No one disputes that. But have these processes demonstrated the capacity to produce the effect in question, namely, the genetic information necessary to structural innovation in the history of life? There are several good reasons to think that they have not.