Evolution and Arithmetic

Chapter 7: Junk DNA and a Demolition Tool

← A Lethal ToolWord Ladders →

Twenty Precise Positions

A protein can fail in two separate ways. One is about precision at a single point: does this exact position need this exact amino acid, or something very close to it? The other is about the protein as a whole: how much abuse, spread across several positions, can the whole structure absorb before it simply falls apart?

Take precision first. Studies that mutate a protein position by position, one at a time, consistently find that some fraction of positions are simply non-negotiable — swap in almost anything else, and the protein stops working. In one such study, of a small, common protein module called a PDZ domain, 83 positions long, 20 positions stood out as highly sensitive: on average, swapping in other amino acids there did serious damage, while most of the other positions barely mattered. In a much larger protein that acts on DNA, 360 positions long, the picture sharpens further once you separate the protein by region. In the small stretch that makes direct, specific contact with DNA, seven or eight out of every ten substitutions destroyed function outright — in one accounting, mutations at over ninety percent of the positions in that region changed the protein's behavior. In the much larger remainder of the same protein — the bulkier body of the molecule, away from the part that reads the DNA — only three substitutions in ten caused the same kind of failure. Overall, across the full 360-position protein, researchers counted forty-two positions that tolerated no substitute at all. That's not a small difference.

It's logical to suppose this is exactly what you'd expect if precision tracks with function: the part of the protein doing the specific, information-carrying job — recognizing one exact sequence of DNA out of everything else in the genome — turns out to be the part that can least afford to change, while the bulkier body around it has much more room to vary. The data lines up with that expectation directly.

It would be a mistake, though, to conclude that the bulkier body of a protein never needs precision. It's logical to suppose that even the scaffolding will have its own non-negotiable points — the exact spot where a fold turns a sharp corner, or where two distant parts of the chain need to lock onto each other to hold the whole structure together, may well need one specific amino acid, or a very narrow set of them, even while the positions around that spot are free to be almost anything. This fits the same underlying pattern found throughout these studies: positions buried in the working architecture of a protein consistently prove less forgiving than positions simply sitting on the surface, filling space, and not otherwise pulling their weight in the fold.

Now the second question: even granting that most individual positions have some flexibility, how much can a protein take before the whole structure gives out? Here the picture is, if anything, less forgiving than it first appears. Most proteins are only marginally stable to begin with — the energy difference between a protein correctly folded and a protein falling apart is often startlingly small. A large, direct study measuring this effect across tens of thousands of individual mutations found that roughly seven in ten measurably pushed the protein toward instability, with only a small fraction being neutral or helpful. Individually, many of these losses are too small to matter. But they add up, spending down a thin and finite margin — and once that margin runs out, the protein doesn't fail gracefully. Function collapses all at once, right at the point where the margin disappears.

So precision and stability are two different obstacles, not one. A random mutation in a working gene can destroy it outright, by swapping the amino acid at a critical position for one that won't do. Or, separately, several mutations at positions where each one alone would do no harm can together run up a stability bill the structure can't pay. Both are real constraints, and both narrow the room in which a series of mutations can build instead of destroy.

Notice, too, where the changes do the most harm: in the part of the protein that does the real work. If a protein is slowly being converted into a different one, that's exactly where the important changes would have to happen — in the working part, not the scaffolding. But if the working part breaks at almost any change, how can it be assumed that one-by-one changes there could slowly, surely, and continually make the protein fitter, step after step, as evolution requires?

Put all of this together, and the picture is hard to avoid: when it comes to an existing, functional gene, random point mutation behaves far more like a demolition tool than a construction tool. It is considerably better at breaking something that already works than at building something that doesn't yet exist. Changes to working genes tend to break them fairly quickly. Once a gene stops working, it no longer provides any benefit to the organism, and further slow changes to it no longer carry the selective advantage evolution needs to spread and keep them. That makes point mutation in functional genes an unlikely engine for evolutionary progress.

Last revised 2026-11-19

← A Lethal ToolWord Ladders →