A sequencing method accurate enough to count mutations one molecule at a time

Two papers published in Nature on 8 October share a method, and the method is the story. Both use NanoSeq, a duplex sequencing approach with an error rate below five errors per billion base pairs [s1]. That precision is what lets researchers detect mutations present in a handful of cells inside a sample containing millions, which is the technical barrier that has kept most somatic mutation biology confined to tumours.

Why the error rate is the whole problem

As people age, tissues become colonised by microscopic clones carrying somatic driver mutations — some a first step toward cancer, others possibly contributing to ageing and other diseases [s1]. The difficulty has always been detection: a mutation present in 0.01 percent of the cells in a sample is indistinguishable from a sequencing error unless the error rate is lower than that.

Duplex sequencing solves this by reading both strands of the original DNA molecule and requiring agreement. The new version described in the first paper extends the approach to whole-exome and targeted capture, so deep sequencing of a polyclonal sample can profile large numbers of clones at once, yielding mutation rates, mutational signatures and driver frequencies in any tissue [s1].

What turned up in the lining of the mouth

The first paper applied targeted NanoSeq to 1,042 non-invasive samples of oral epithelium and 371 blood samples from a twin cohort [s1].

The result was a selection landscape the authors describe as extremely rich: 46 genes under positive selection in oral epithelium and more than 62,000 driver mutations, alongside evidence of negative selection in essential genes [s1].

Sixty-two thousand driver mutations in normal mouth lining. These are healthy people, and their cheek cells are already carrying tens of thousands of mutations that natural selection is actively favouring at the cellular level.

The resolution is fine enough that the researchers obtained high-resolution maps of selection across coding and non-coding sites for many genes, which they characterise as a form of in vivo saturation mutagenesis [s1] — reading, from natural variation in living tissue, what would otherwise require engineering every possible mutation in a laboratory.

Multivariate regression models then allowed what the authors call mutational epidemiology: studying how exposures and cancer risk factors such as age, tobacco and alcohol alter the acquisition or selection of somatic mutations [s1].

That is a genuinely new kind of measurement. Conventional epidemiology links an exposure to a disease outcome years later. This links an exposure to the molecular events that precede it, in tissue obtained without a biopsy.

What turned up in sperm

The second paper used NanoSeq on 81 bulk sperm samples from men aged 24 to 75 [s2].

Mutations accumulated linearly at 1.67 per year per haploid genome (95% CI 1.41–1.92), driven by two mutational signatures associated with human ageing [s2]. Deep targeted and exome sequencing identified more than 35,000 germline coding mutations [s2].

Forty genes showed significant positive selection in the male germline, 31 of them newly identified, with activating or loss-of-function mechanisms across diverse cellular pathways [s2]. Most of the positively selected genes are associated with developmental or cancer predisposition disorders in children [s2]. Four showed increased frequencies of protein-truncating variants in healthy populations [s2].

The consequence the authors draw is the one that will travel furthest: positive selection during spermatogenesis drives a two- to three-fold increased risk of known disease-causing mutations, resulting in 3–5 percent of sperm from middle-aged to older men carrying a pathogenic mutation somewhere across the exome [s2].

How to read the 3–5 percent

That figure is a property of sperm cells, not a prediction about children.

A single sperm carrying a pathogenic mutation somewhere in the exome is not the same as a child with a genetic disorder. Fertilisation selects one sperm from an enormous pool. Many pathogenic mutations are incompatible with development and never produce a live birth. Many others are recessive, or have low penetrance, or affect genes whose disruption produces no recognisable phenotype.

What the paper does establish is a direction and a mechanism. The authors state that the findings highlight a broader increased disease risk for children born to fathers of advanced age than previously appreciated [s2]. The word doing the work there is broader — paternal age effects were already known for specific conditions; this widens the set of genes involved.

Eighty-one men is also a small sample for a claim about population-level risk, and the paper is a measurement of germline mutation, not a study of offspring outcomes.

What connects the two

Both papers describe the same phenomenon in different tissue: cells in a normal, healthy body compete, and mutations that give a cell a growth advantage spread. In the mouth that competition is a step along the road to cancer. In the testis it changes what gets passed to the next generation.

Neither is new as a concept. What is new is being able to count it without a tumour, a biopsy, or a family history — from a cheek swab and a semen sample.

What to watch

The immediate applications the first paper points to are early carcinogenesis, cancer prevention, and the role of somatic mutations in ageing and disease [s1]. The near-term test of whether mutational epidemiology becomes a real discipline is whether the exposure–mutation associations found in this twin cohort replicate in independent populations, and whether the driver burden in normal tissue predicts anything about who later develops cancer. Neither has been shown yet.

Sources

• Somatic mutation and selection at population scale — Nature, 8 October 2025
• Sperm sequencing reveals extensive positive selection in the male germline — Nature, 8 October 2025