Evolution and Arithmetic

Chapter 7: Junk DNA and a Demolition Tool

← A Reality CheckChapter 8: Weasels →

A Second Chance: Gene Duplication

One of the oldest and most widely taught ideas in evolutionary biology proposes a way around the numbers this chapter has been raising. It is a textbook mechanism with more than fifty years of research behind it.

The idea, first laid out by geneticist Susumu Ohno in 1970, starts from something ordinary: genes get accidentally duplicated all the time, through routine copying errors. When that happens, an organism ends up with two copies of a gene that used to exist as only one. The original copy keeps doing its job, and stays under the same watchful pressure of natural selection it always has — any damaging change there still gets weeded out. But the second, spare copy is, in a sense, off the hook. The organism doesn't need it to survive. So the argument goes that this extra copy is free to drift, to pick up mutations that would normally be dangerous, without putting the organism at any real risk — because the original is still there, doing the necessary work. Every so often, so the theory holds, that freedom lets the spare copy stumble onto something new: a new protein, with a new job, that the organism didn't have before. If true, this would matter a great deal for everything this chapter has been asking, because it offers evolution something it otherwise doesn't have: a way to search for new function without betting the organism's survival on the outcome.

There's a fair amount of evidence in its favor. Duplicated genes are common — genomes accumulate them constantly, and entire genomes have been duplicated wholesale more than once over the course of vertebrate history, leaving whole families of related genes behind as a record. There are also specific, well-documented cases where researchers can trace a new protein directly back to an old one through exactly this route — an antifreeze protein in an Antarctic fish, the eelpout, evolved from a duplicated copy of an ordinary metabolic enzyme, is one of the more carefully studied examples. And a related, better-supported version of the idea, called subfunctionalization, has real support behind it too: rather than one copy inventing something new, the two copies simply split up the jobs the original gene used to do by itself, each specializing in part of what one gene used to handle alone. That's a smaller ask than inventing something from nothing, and the evidence for it is correspondingly stronger.

But look closely, and the picture that's supposed to rescue the Standard Model runs into several of the same problems this chapter keeps finding everywhere else.

The first problem is survival. The overwhelming majority of gene duplicates don't last long enough to do anything at all. Left without a job to protect them, most duplicate copies simply accumulate damaging mutations until they break entirely, becoming dead, non-functional stretches of DNA — what geneticists call pseudogenes. One influential estimate puts the average functional lifespan of a new duplicate at only a few million years before it's effectively gone. That's the demolition tool from earlier in this chapter, showing up again in a new setting: a spare copy buys some freedom from selection, but freedom cuts both ways, and decay wins the race far more often than invention does.

The second problem is more direct, and more recent. In 2024, researchers finally ran the actual experiment Ohno's idea calls for — evolving real populations of bacteria carrying either one copy or exactly two copies of a gene, under mutation and selection, and watching what happened. The two-copy populations did behave the way the theory predicts in one respect: they tolerated more mutations and built up more genetic variety, exactly as expected from having a spare copy to protect them. But that extra tolerance never translated into faster evolution of new function. In most cases, one of the two copies simply broke, worn down by accumulating damage, before anything useful had the chance to emerge. Having a spare copy, it turns out, is not the same thing as having a shortcut.

The third problem cuts closer to the heart of what this chapter has been arguing all along. In the specific cases researchers can actually point to and confirm — a duplicated gene that clearly went on to do something new — the number of changes involved tends to be small. Not the slow accumulation of many independent small mutations gradually building toward something unrecognizable, but a handful of specific changes, often concentrated in just the few positions that matter most for the new job.

Could larger transformations have happened and simply gone unnoticed? One might argue that researchers can only recognize a duplicate when the two copies still look alike, so a copy that changed dramatically would leave no trace. But this chapter's own findings point the other way. A protein's scaffolding tolerates a great deal of change while still holding the same overall shape — and the overall fold of a protein is known to survive long after the letter-by-letter resemblance has faded, which is why researchers now compare shapes as well as sequences. A new job, on the other hand, would mostly be done by changes in the working core. So a duplicate that had been turned to a genuinely new job should still be recognizable as kin: much the same frame, with a different working heart. Cases like that are exactly the kind researchers are equipped to find. What they actually find, again and again, are cases where only a handful of positions changed.

So gene duplication, taken as a whole, ends up looking like most of the other proposed escapes this book has examined: a real, well-documented mechanism, doing real work, and falling well short of what's actually being asked of it. It offers evolution more raw material to work with, and a genuine (if narrow) route to splitting up existing jobs between two genes instead of inventing new ones. What it doesn't offer is a demonstrated way of routinely producing large-scale, well-functioning novelty. The arithmetic problem this chapter keeps returning to doesn't go away because a spare copy is sitting around. It just gets a second seat at the table.

One way around all of this is to suppose that evolution does its real work not in genes that already function, but somewhere else entirely: out in junk DNA. That is the subject of the next chapter.

Last revised 2026-11-19

← A Reality CheckChapter 8: Weasels →