EVOLUTION’S MECHANISM PROBLEM

Ask people how evolution works and most will answer something like this: mutations happen by accident, natural selection keeps the useful ones, and given enough time we can get from a bacterium to a blue whale. Sounds sensible. But there’s a question hiding inside the answer, a question the whole debate turns on.

Where do brand new genes come from?

Not slightly different versions of genes that already exist. Not small tweaks to old instructions. Genuinely new genes, carrying new information, doing jobs that nothing in the parent organism could do.

This post looks at the machinery of evolution—the actual, proposed, physical mechanism—and asks a simple question: has anyone ever watched it build something new? The answer, we will argue, is far more interesting than either side usually admits. Nature is full of change. But almost all of the change we can measure moves in one direction: things get broken, blunted, borrowed or rearranged. Very little gets built.

If that is true, then similarity between creatures needs a different explanation than inheritance from a shared ancestor. It may point instead to shared design—a single Architect reusing good parts.

That last point is worth pausing on, because it does the heavy lifting in everything that follows. Selection cannot invent. It can only sort. So the entire creative burden falls on mutation—on accidents.

The real question isn’t “how much change?” but “what kind?”

Christians are often accused of accepting “small” evolution but rejecting “big” evolution, as though we simply cannot imagine large numbers. That isn’t the argument. Nobody doubts that dogs vary, that finches’ beaks change, or that bacteria become resistant to drugs. We know all of that happens.

The real question is about the kind of change, not the amount. Michael Behe, a biochemist at Lehigh University, sorted every adaptive change ever recorded in the laboratory into three simple categories. Once you see these three categories, the whole landscape becomes clear.

Table 1: Michael Behe’s three categories of adaptive change. Only the first category can build a body plan. It is also, by a very wide margin, the rarest.

Behe’s survey of four decades of laboratory evolution led him to what he called the First Rule of Adaptive Evolution: if an advantage can be gained by breaking or blunting something, that’s the route evolution will take. Why? Because there are vastly more ways to break a machine than to improve it, so damaging mutations arrive first. And arrive constantly. Evolution takes the cheap route because the cheap route is the only one it can reliably find.

Douglas Axe: why finding a new protein is like finding one house on a billion planets

Imagine a protein 150 amino acids long—quite a modest size. At each of those 150 positions, life uses one of 20 possible amino acids. The number of possible chains is 20 multiplied by itself 150 times. That’s a number with almost 200 digits.

Scientists call this vast space of possibilities sequence space. Almost all of it is rubbish. A random chain of amino acids doesn’t fold into anything useful; it usually collapses into a sticky, functionless clump. Working proteins sit on tiny islands of function in an enormous, empty ocean.

Douglas Axe, a molecular biologist trained at Caltech who spent years at the University of Cambridge, set out to measure how tiny those islands are. Working with a section of a bacterial enzyme, he published an estimate in the Journal of Molecular Biology in 2004: roughly one workable sequence in 10 to the power of 77.

Axe also has an answer to those who dismiss ordinary people’s instinct that living things look designed. He calls it the universal design intuition, and he argues it’s not ignorance but sound reasoning: we all know, from a lifetime of experience, that clever coordinated arrangements come from minds.

Michael Behe: the edge of what mutations can reach

Douglas Axe’s argument is mathematical. Michael Behe’s is observational, and in some ways harder to escape, because it uses real creatures fighting for their lives.

Consider the malaria parasite. When the drug chloroquine was introduced, resistance to most drugs appeared within a few years. Chloroquine resistance took decades—and when researchers worked out why, they found that it needs at least two changes to the same protein, and that one of those changes is harmful on its own. So the parasite has to pass through a valley before reaching the hill. Behe cites malaria specialist Nicholas White’s estimate that this happens roughly once in every 10 to the power of 20 parasites.

That number is enormous, yet malaria clears it, because there are astronomically many parasites in the world. This is Behe’s point. He’s not saying two coordinated mutations are impossible. He’s saying we can now measure the price, and then ask which creatures can afford it.

Table 2: Behe’s “edge of evolution”. The creatures with enough numbers to buy multi-step changes are microbes—and even they have never been observed to build a new molecular machine. The creatures said to have evolved lungs, eyes and wings are precisely the ones that cannot afford the price.

HIV is the sharpest test we have. It has mutated inside millions of human bodies, with a mutation rate far higher than ours, for decades. It has changed. It has escaped drugs. It has not invented a single new protein machine.

John Sanford: the genome is running downhill, not uphill

John Sanford spent his career as a plant geneticist at Cornell University and co-invented the “gene gun”, the technology used to insert genes into crops. He holds dozens of patents. He is not a man unfamiliar with how genomes work.

His argument, which he calls genetic entropy, runs like this.

• Every human baby is born with roughly 60 to 100 brand new mutations that neither parent had.
• The overwhelming majority are either neutral or very slightly harmful. Outright disasters are rare; tiny scratches are constant.
• Natural selection has limited eyesight. If a mutation’s effect on survival is smaller than the ordinary noise of life—accidents, weather, who happens to find a mate—selection simply cannot see it. Geneticists call this the nearly-neutral zone.
• Therefore slight damage accumulates faster than selection can remove it. Over many generations, the genome degrades.

Here’s the part that should make everyone stop and think: this isn’t only a creationist worry. Michael Lynch, a leading evolutionary geneticist, published a warning in the Proceedings of the National Academy of Sciences in 2010 about measurable decline in human genetic health under modern relaxed selection. Alexey Kondrashov, another mainstream geneticist, published a paper whose title asks why, given the mutation load, we have not died out many times over. The disagreement is about the size and seriousness of the problem, not its existence.

The wait time problem

There’s a second and even more awkward difficulty. Suppose a creature needs a handful of specific mutations, in specific places, to produce something new. How long must it wait?

John Sanford and colleagues modelled this in 2015 in the journal Theoretical Biology and Medical Modelling, using a human-like population of 10,000 individuals with realistic mutation and reproduction rates.

Table 3: Sanford and colleagues’ waiting-time estimates for a population of 10,000, compared with the roughly 6 million years usually allowed for human evolution. Note how quickly the waiting time explodes as the requirement grows: this is why the problem is not solved by adding more time.

An older version of this problem was raised, remarkably, by one of the founders of modern evolutionary theory. JBS Haldane calculated in 1957 that a population can only fix roughly one beneficial mutation every 300 generations. Applied to the human line, that allows perhaps 1,700 beneficial changes since the supposed split from other apes—against tens of millions of genetic differences that need explaining. This is still known as Haldane’s dilemma. It hasn’t been convincingly answered.

Orphan genes: the genes with no family tree

Now to the newest and, perhaps, most striking evidence.

Common descent makes a clear prediction. If all life is one branching family, genes should have relatives. Compare any gene in a mouse and we should find a recognisable cousin in a rat, a fainter cousin in a fish, a fainter one still in a fly. Genes should nest inside one another like Russian dolls.

As genome after genome has been sequenced, biologists have kept meeting genes that refuse to do this. They’re called orphan genes (or taxonomically restricted genes). They have no detectable relatives outside a narrow group of creatures. In many species, somewhere between 10 and 30% of genes fall into this category. Some are essential. Many produce the very features that make a species distinctive.

Some apparent orphans aren’t really orphans: a 2020 study in PLoS Biology by Caroline Weisman, Andrew Murray and Sean Eddy showed that our search tools lose the trail when sequences are short and change quickly, so relatives can exist but stay hidden. And a small number of genuinely new genes have been documented—the antifreeze protein in Antarctic fish, for example, which was reshaped from a digestive enzyme.

But look closely at what these documented cases actually deliver. They’re typically short, floppy, weakly used and modestly helpful. Not one is a new folded machine. Not one is a new organ, a new tissue type or a new body plan. The distance between “a short new peptide appeared” and “an eye appeared” is the entire question, and it isn’t bridged by these examples. It’s illustrated by them.

“But genes get duplicated”—the standard answer, examined

The usual reply is gene duplication. A gene is accidentally copied twice; the spare copy is free to change; eventually it becomes something new. It is a clever idea, and it certainly happens. But notice the difficulty it creates rather than solves.

• A spare copy is invisible to selection. While it drifts, it is doing nothing useful, so nothing protects it. Selection cannot guard a work in progress.
• Unguarded DNA decays. Mutations accumulate in the spare copy at random. The usual observed fate of a duplicate gene is not a new invention. It’s a pseudogene—a broken, unread remnant.
• The gap still has to be crossed blind. Duplication gives us a second copy of an island we’re already standing on. It doesn’t tell us how to cross the ocean to a different island. Douglas Axe’s problem returns unchanged.
• Copying is not composing. Photocopying a page of a novel twice doesn’t write a new chapter. The information content hasn’t increased; it has been repeated.

A further difficulty: the genome does several jobs at once

John Sanford makes one more point that deserves to be better known. DNA is not a simple list where each letter does one job. The same stretch of DNA can, at the same time, code for a protein, contain a switch controlling another gene, carry a signal for how the DNA should be folded and packed, and hold instructions for how the message should be cut and spliced. Sanford calls this a poly-functional genome.

The consequence is severe. If one letter serves four different codes simultaneously, a mutation that improves one code will almost certainly damage the other three. The proportion of mutations that are beneficial overall collapses towards nothing. The more we learn about how much the genome is doing at once—and the collapse of the old idea that most of it is useless “junk”—the tighter this constraint becomes.

The longest experiment ever run: what actually happened

Since 1988, Richard Lenski at Michigan State University has grown E. coli bacteria continuously, generation after generation, for more than 75,000 generations. On a human timescale that is well over a million years of reproduction. This is the cleanest test of the mechanism that anyone has ever performed. What did the bacteria do?

Table 4: A summary of Richard Lenski’s long-term experiment. Note that even the celebrated citrate gain was a rearrangement of parts the bacterium already owned—a new wiring diagram, not a new component. Meanwhile the losses were broad, repeated and predictable. Richard Lenski himself reads these results as evidence of open-ended innovation; we simply ask readers to weigh what is in each column.

This is Behe’s First Rule under a microscope. Given the freedom to adapt for a million equivalent years, the bacteria adapted mostly by throwing equipment overboard.

Antibiotic resistance: the example that proves the opposite of what is claimed

Antibiotic resistance is presented as evolution in action, and in a limited sense it is. But when you look at the molecular detail, resistance nearly always arrives by one of three routes—and none of them creates anything.

Table 5: The three main routes to antibiotic resistance. Notice that the first two carry a real cost, which is why resistant strains often lose ground once the drug is withdrawn. Resistance is survival by sacrifice, not by invention.

The same pattern appears everywhere once we look for it. Sticklebacks lose their protective pelvic spines because a control switch beside the Pitx1 gene is deleted. Our dog breeds—from chihuahuas to great danes—come overwhelmingly from losses and reshuffling within an existing wolf genome, which is precisely why so many pedigree breeds carry inherited disease. Cave fish lose their eyes. Flightless birds lose their flight.

Life is remarkably good at adapting by breaking things. That’s a real and fascinating capability. It’s simply not the capability the story of common descent requires.

The positive case: what design actually explains

Critics say design is only a gap-filler—that we point at what science cannot yet explain and insert God. That charge would stick if our argument were merely negative. It is not.

• We know of one cause of large volumes of functional information, and it is mind. Every book, every program, every blueprint in human experience traces back to intelligence. When we find a four-letter coded language that is read, copied, proofread, edited and translated, inferring an author is not a failure of imagination. It is the ordinary logic of every other field, including forensics and archaeology.
• Design predicts what we keep finding. Reused parts across unrelated creatures. Function turning up in DNA once dismissed as junk. Layers of code stacked on top of one another. Firm limits on how far breeding can push a species.
• Design explains similarity without requiring descent. A good engineer reuses a good component. Shared parts are evidence of a shared source; whether that source is an ancestor or an Architect is exactly the question, and cannot be assumed at the start.

Stephen Meyer, in Signature in the Cell, presses the point that DNA carries not merely order but specified information—the kind of arrangement we recognise instantly as linguistic. And R. C. Sproul, in Not a Chance, made a philosophical observation that cuts deeper than any laboratory result: chance is not a force. It is a word describing our ignorance of causes. Chance has no being, and what has no being can do no work. To say a gene arose “by chance” is not an explanation. It is a decision to stop explaining.

Table 6: Both models make testable predictions. This is why the debate is a scientific one and not merely a matter of belief—and it is why we are content to let future research judge between them.

What Scripture says about a world running down

For the Christian, none of this should be surprising. Scripture describes a creation that was made good, made in distinct kinds, and then damaged.

Notice how well this fits the data. The direction of measurable genetic change is downhill. Genomes accumulate damage. Species lose abilities. Breeds bred for one striking feature carry hidden weaknesses. A world under the curse of Genesis 3 is a world where the arrow of change points towards decay—which is precisely what geneticists measure, whatever framework they use to interpret it.

Where this leaves us

Let us gather the threads.

• Natural selection cannot create. It can only choose between options that already exist, so every creative act must come from random mutation.
• Working proteins are astonishingly rare in the space of possible sequences, and selection cannot help you reach one until you have already arrived.
• We can now measure the cost of even two coordinated mutations, and the creatures said to have evolved most cannot afford it.
• Mutations accumulate as damage faster than selection can clear it, and the waiting time for specific new sequences vastly exceeds the time available.
• Thousands of genes have no family tree, appearing suddenly and in bulk, exactly where nested inheritance says they should not.
• In the longest evolution experiment ever run, and in every famous case of observed adaptation, the mechanism worked mainly by breaking things.

Put together, these are not scattered doubts. They are a single, consistent finding about the mechanism: it’s a superb editor and a poor author. It tunes, trims and dismantles with great skill. It has never been caught writing.

And if the mechanism cannot write, then the writing in every living cell needs an author. The similarities running through all of life are real, and they demand an explanation. We suggest the better explanation is not a long unguided accident but a common Designer, reusing excellent designs, and writing His signature into every cell of every creature that has ever lived.

“I praise you, for I am fearfully and wonderfully made. Wonderful are your works; my soul knows it very well.” (Psalm 139:14, ESV)

Related Reads

• Common Descent or Common Design: Which Explains Life’s Unity?
• Can Random Mutations Create Complex Life? The Math Says No
• Failing the Math Test: Darwinism’s Wait Time Problem
• Haldane’s Dilemma: Why Can’t Evolution’s Math Add Up?
• Dependency Graph of Life: Why Genes Don’t Fit Darwin’s Tree
• The Cambrian Explosion: Why the Fossil Record Runs Top-Down
• Fruit Fly Mutations: What Decades of Research Actually Tell Us
• The Big Hoax: Turns Out ‘Junk’ DNA Isn’t Junk After All