Biology EP. 14 / 64 By SENSMAN

Why Life Never Drove Its Error Rate to Zero — Proofreading, Repair, and a Compromise at 10⁻⁹

Seventeen years behind a pharmacy counter trying to reach zero. Life never took that road. An error rate has a ceiling as well as a floor.

Why life never drove its error rate to zero — replication error and the stages of repair

At a funeral, a relative of the bereaved explained how the deceased had spent his last years, and used a name for the illness that nobody uses now. To him it had been the ordinary word his whole life. Several faces in the room tightened.

There was no malice in it. The word was simply where he had left it.

The same sentence, and the odds of being misread had gone up. The error rate of the language had changed.

Two renamings seen from behind a counter

I ran a pharmacy in Seoul from 2002 to 2019. In that stretch I watched two disease names change.

YearBeforeAfter
around 2010간질 (falling sickness)뇌전증 — epilepsy
2011정신분열병 (mind-split disease)조현병 — attunement disorder

The Korean Epilepsy Society renamed itself, and schizophrenia was given a new Korean name built on 調絃, the tuning of the strings of an instrument.

Medically nothing moved. Same condition, same drugs. What changed was the error rate of the word on its way to the patient.

Counselling someone on their medication, you see the difference immediately. The same explanation lands in a different place. A 2013 report in The Lancet examined this Korean renaming for exactly that reason.

Civilisation keeps refining its words to bring errors down.

And perfect transmission does not exist

There was one more place in that shop where I pushed for zero. Dispensing.

Drug names resemble each other more often than people expect. One letter apart, or the same sound, or a package that looks alike. The international term is LASA — look-alike, sound-alike. Handwriting on a prescription, an order taken by phone, and a busy hour together raise the odds.

For seventeen years my job was lowering that probability. Keep them on separate shelves, change the label colour, read it aloud twice, check once more at the end.

It never reaches zero.

Back then I took that as my own limit. Going back through the biology, it looked different.

Life did not choose zero either.

The numbers life works with

A human somatic cell holds about 6.1 billion base pairs — roughly 3.05 billion from each parent.

An adult body is built of about 30 trillion cells, and across a lifetime those cells rewrite that long sentence over and over.

Fidelity is not achieved in one step. It is stacked in three.

StageWhat it doesError rate remaining
① Base selectionthe polymerase picks the matching base10-4 – 10-5
② Proofreadinga 3′→5′ exonuclease removes the base just misinserted10-6 – 10-7
③ Mismatch repairleftover mismatches are excised and refilled after replication10-9 – 10-10

One in a billion. Rewrite 6.1 billion letters and the ones that come out wrong can be counted on your hands.

No copying device humans have built comes near this.

How many actually come out wrong

There is a measurement, not only an estimate.

In 2012 Kong’s group paired up parent and child genomes in Iceland and counted directly how many new mutations each child carried that neither parent had.

Around 60 per newborn.

Most of them came from the father. For every additional year of paternal age the count rose by about two. The cells that make sperm keep adding divisions as a man ages.

Sixty letters out of 6.1 billion. That is what one generation pays.

Why not zero ① physics blocks it from below

Why not copy perfectly. Zero percent error. Would that not be the ideal organism?

No. Two walls.

The lower one is thermodynamic.

Molecules live in a probabilistic world. Heat creates jitter; bonds lift and settle moment to moment. Even as adenine and thymine pull on each other, the atoms are vibrating.

In 1974 John Hopfield formalised the consequence. Kinetic proofreading.

There is a ceiling on the accuracy obtainable from equilibrium binding strength alone. Going above it requires inserting an extra step and burning ATP to make it irreversible. Every further step of fidelity carries an energy charge.

Accuracy is not free. It has a price tag.

As zero is approached, the price diverges. That is the first reason life did not buy it.

Why not zero ② there is a second wall above

Most accounts stop here, saying errors are needed for variation and variation for selection.

True, and half the picture. There is a wall on the other side.

In 1971 Manfred Eigen showed that above a certain error rate, repeated copying scatters the original sequence until the information is no longer held in the population at all. Error catastrophe.

That threshold sits roughly at the inverse of genome length. The longer the genome, the lower the per-base error rate it can tolerate.

Too lowToo high
Outcomeno variation → adaptation stopsinformation dissolves → the population collapses
Costenergy divergesreplication fails

Life lives in the narrow band between the two.

The rate is a setting, not a given

If this were simply a physical limit, that would be the end of it. There is evidence it is regulated.

In 1991 John Drake assembled DNA-based microbes whose genomes differ in length by a factor of thousands.

Per-base error rates varied widely between species. And yet mutations per genome per generation held at about 0.003 across all of them.

As the genome grew longer, the per-base rate dropped to match, holding the total steady.

An error rate is not a defect. It is a setting. The strength of the repair machinery is the dial.

A drug proves it

Back to the pharmacy side.

If the error rate is a device, then breaking the device is also a drug.

Molnupiravir works this way. It does not kill the virus directly. It inserts itself into the RNA, makes base pairing ambiguous, and lets errors accumulate with each round of copying.

In 2021 Kabinger’s group worked out the molecular mechanism: at the site where the drug sits, two different pairings become possible, and the sequence degrades as replication continues.

It induces, as therapy, the error catastrophe Eigen calculated in 1971.

The name for it is lethal mutagenesis. Pushing a virus over the wall rather than shooting at it.

The most practical proof that the error rate is a regulated quantity sits on a pharmacy shelf.

This article’s definition

Fixed here.

An error rate is not a defect but a device. Civilisation tries to eliminate error; life regulates it.

Miss that difference and vaccines, new drugs and gene therapy all get read wrongly, because demanding perfection is the same as forbidding change.

The hexagram here is 大有, and its Image reads 遏惡揚善 — check what harms, raise what is good. Not erase everything, but block what must be blocked and keep what must be kept. That sentence is the job description of the repair systems.

Pushing dispensing errors towards zero for seventeen years was the right call. Dispensing has no upper wall, because one person is not a population.

Life runs a different calculation. The accuracy of a single individual and the persistence of a whole lineage do not point the same way.

Perfection is a stop, and a stop is extinction. Life continues only while it is twisted by 10-9.

We are the outcomes that survived that twist.

In short

  • Base selection alone misreads about one letter in a hundred thousand
  • Proofreading improves it ~100-fold and mismatch repair another 100–1,000-fold, reaching one in a billion to ten billion
  • Measured in people, a newborn carries about 60 mutations neither parent had
  • That count rises by roughly two per additional year of paternal age
  • Kinetic proofreading: each further step of accuracy must be paid for in burned energy
  • Error catastrophe: past a threshold the information is not retained. There is a ceiling as well as a floor
  • Drake’s rule: mutations per genome hold near 0.003 across genomes of very different length
  • Molnupiravir raises the error rate deliberately, inducing error catastrophe as therapy
  • The Image of 大有 reads 遏惡揚善 — not erase everything, but block and keep

This piece is rebuilt for search readers from “Error Rate” of the serial essay The Great Corona Swindle. The experience from the years of running a pharmacy is the author’s own; the mechanisms and numbers have been reset against primary sources.

Frequently asked

What is the actual error rate of DNA replication?
It depends on the stage. The polymerase choosing a base gets it wrong about once in a hundred thousand. The 3′→5′ exonuclease attached to the polymerase removes a wrongly inserted base and improves this roughly a hundredfold. After replication, mismatch repair removes another 100 to 1,000-fold. The final rate lands around one in a billion to ten billion.
Has it been measured in people?
Yes. In 2012 Kong's group compared parent-child trios in Iceland and counted directly how many mutations a child carried that neither parent had. The average was around 60 per newborn. Most came from the father's side, and the count rose by about two for each additional year of paternal age — the cells producing sperm accumulate more divisions as a man gets older.
Why not drive the error rate to zero?
Two walls. From below, physics. Hopfield's 1974 kinetic proofreading shows that every further step of accuracy requires additional energy to be burned, so the cost diverges as zero is approached. From above, evolution: no errors means no variation, and no variation means selection has nothing to act on. But there is also a wall on the other side. Eigen's error catastrophe shows that past a certain rate the information is not retained at all.
Is there evidence that the error rate is regulated?
In 1991 Drake compared DNA-based microbes whose genomes differ in length by a factor of thousands and found that mutations per genome per generation stayed almost constant at about 0.003. A longer genome runs a correspondingly lower per-base error rate. That is direct evidence that the rate is a setting rather than a physical given.

References

#error rate#replication fidelity#proofreading#mismatch repair#kinetic proofreading#error catastrophe#Drake's rule#de novo mutation#molnupiravir#look-alike sound-alike

#오류율#DNA복제충실도#교정기능#불일치수선#키네틱프루프리딩#오류재앙#드레이크법칙#신생돌연변이#몰누피라비르#유사약품명

Published October 4, 2026