A tube of blood can flag a chromosome problem in a fetus or a tumour nobody has found. It works by counting fragments of dead cells

Every day, a large number of the cells in your body die on schedule and are cleared away. As they break up, short pieces of their DNA end up in the bloodstream. Draw a tube of blood, spin off the cells, and the clear plasma left behind contains a soup of these scraps — cell-free DNA, shed by tissues all over the body within the last few hours.

That soup turns out to be readable. In pregnancy it contains DNA from the placenta. In someone with a tumour it contains DNA from the tumour. Two of the most commercially successful genomic tests of the past decade are built on this observation — and both are widely misunderstood in the same way, because neither is doing what people assume.

What is actually floating there

The fragments are remarkably consistent: around 170 bases long. That number is not arbitrary. DNA in a cell is wound around spool proteins called nucleosomes, and when a cell dies in an orderly way its DNA is cut in the exposed stretches between spools. What survives is the wrapped portion plus a little linker — so the fragment length is a fingerprint of how the DNA was packaged, and researchers now read patterns in fragment length and cut position as a signal in their own right.

The other important property is that cell-free DNA is cleared quickly, with a half-life measured in minutes to a couple of hours. The sample is not an archive. It is a snapshot of what was dying recently, which is precisely what makes it useful for monitoring something that changes.

In pregnancy, a fraction of the plasma DNA comes from the placenta — usually around a tenth of the total after ten weeks, rising as the pregnancy progresses. This is called the fetal fraction, and the name is slightly wrong in a way that matters later: the DNA is placental, not fetal.

Prenatal screening is a counting exercise

Here is the part that surprises people. A non-invasive prenatal test does not read the fetus's genome. It cannot — the fetal material is a minority component of a mixture dominated by the mother's own DNA, and there is no way to separate the two.

What it does instead is count. Millions of fragments are sequenced just enough to identify which chromosome each came from, and then tallied. In a pregnancy where the fetus has the usual two copies of chromosome 21, the proportion of fragments mapping to chromosome 21 sits at a predictable value. If the fetus has three copies, that chromosome is slightly over-represented in the mixture — by an amount roughly equal to half the fetal fraction. A tenth of the DNA being placental, with one extra chromosome out of two, shifts the count by a few per cent.

Detecting a few per cent shift requires counting a great many fragments, and it explains the test's central dependency: if the fetal fraction is too low, the expected shift disappears into the noise and the test simply cannot report. Low fetal fraction is more common early in pregnancy and at higher maternal weight, and a "no result" is a genuine outcome rather than a failure of the laboratory.

It is not reading the baby's genome. It is noticing that one chromosome is slightly over-represented in a soup. Every limitation of the test follows from that sentence.

Why it is a screen and not an answer

Two separate things make this a screening test.

The first is arithmetic that applies to any test for a rare condition: when the condition is uncommon, a substantial share of positive results are false, even with excellent sensitivity and specificity. This is why the same test performs very differently for a common trisomy in an older mother than for a rare microdeletion in a young one — the test did not change, the underlying frequency did.

The second reason is biological and more interesting, because these are cases where the test correctly reported what was in the blood and the blood was misleading.

The DNA is placental. In confined placental mosaicism, a chromosomal abnormality is present in the placenta and absent from the fetus. The test is right about the sample and wrong about the baby.

A vanishing twin. DNA from a twin that stopped developing early can persist and be counted.

Maternal biology. The mother's own DNA is the bulk of the sample. Maternal chromosomal variation can produce an abnormal result — and occasionally an unexpected pattern across several chromosomes turns out to reflect an undiagnosed maternal cancer, which is a genuinely serious incidental finding that the test was not designed to look for.

None of this makes the test bad. It makes it a screen, whose positive results require confirmatory diagnostic testing before anyone acts. In India these tests are marketed heavily and often directly to patients, frequently without access to genetic counselling, and that gap — not the assay — is where the harm happens.

The cancer version is the same idea, much harder

Circulating tumour DNA works on identical principles and faces a far worse signal problem. Where placental DNA might be ten per cent of the sample, tumour DNA in early disease can be well under one per cent, sometimes under a tenth of that. You are looking for a handful of mutated fragments among millions of normal ones.

At that level, the sequencer's own error rate becomes the limiting factor: if the machine miscalls one base in a thousand, genuine rare mutations are buried in noise. The standard solution is to tag every original fragment with a unique molecular barcode before amplification, so that copies of the same starting molecule can be grouped. A real mutation appears in every copy of that group; a sequencing error appears in one. This turns an error-prone read into a consensus.

Then there is a confounder that took the field years to appreciate properly. As people age, blood stem cells accumulate mutations, and clones carrying them expand — clonal haematopoiesis. Those cells shed DNA into plasma like everything else, carrying mutations in exactly the genes cancer panels look for. For a while, a meaningful share of apparent tumour signals were coming from ordinary blood cells. Distinguishing them requires sequencing the patient's white cells alongside the plasma, which is now standard practice and was not always.

Where the evidence is actually strong

The dream application — a blood test that screens healthy people for any early cancer — is the hardest version and the least settled. Detecting a tumour before it causes symptoms means detecting it when it is small and shedding very little, and specificity has to be extraordinary because the test is applied to people who are overwhelmingly well.

The applications where this is already changing practice are narrower and more convincing. Choosing a treatment when a tissue biopsy is difficult or unsafe: a blood sample can identify the mutation that determines which targeted drug will work. Monitoring: because the half-life is hours, ctDNA falls fast when a treatment works and rises when resistance emerges, often well before imaging shows anything. And residual disease: detecting ctDNA after apparently curative surgery identifies patients whose cancer will return, sometimes months before a scan can, which is the strongest evidence base in the field and the direction clinical practice is genuinely moving.

Why it matters for students and researchers

This is a good subject for understanding what modern diagnostics actually are. The sequencing is routine; the difficulty is entirely in the statistics of rare signals — limits of detection, error models, how many molecules were in the tube to begin with, and what happens to a probability when the condition is rare. A student who understands those ideas can evaluate a test they have never seen. One who only knows the biology cannot.

For India there are two specific openings. Reference data used to interpret these tests comes largely from other populations, and locally derived baselines matter for exactly the same reason they do in the rest of genomics. And the capacity gap around these tests is counselling rather than laboratory: the tests are available and expanding fast, while the ability to explain what a screening result does and does not mean has not kept pace. A test is only as good as the decision made after it, which is a sentence worth attaching to every diagnostic anyone builds.

Frequently asked questions

What is cell-free DNA?

It is short fragments of DNA, around 170 bases long, released into the bloodstream by cells as they die. It circulates for minutes to a couple of hours, so it reflects what has been dying recently rather than accumulating over time.

How does a non-invasive prenatal test work?

It sequences millions of DNA fragments from the mother's plasma and counts how many map to each chromosome. An extra fetal chromosome makes that chromosome slightly over-represented in the mixture, which the test detects as a statistical excess.

Why is NIPT a screening test rather than a diagnosis?

Because a positive result can arise without the fetus being affected — most importantly because the circulating DNA comes from the placenta, which does not always match the fetus. Rare conditions also produce a high proportion of false positives, so positives need confirmatory testing.

Why is detecting cancer DNA in blood so difficult?

Because tumour DNA can be less than one per cent of the cell-free DNA present, so real mutations are hidden among sequencing errors. Molecular barcoding and sequencing the patient's blood cells separately are used to separate true signals from artefacts.

What is clonal haematopoiesis and why does it matter here?

It is the age-related expansion of blood cell clones carrying mutations. Those cells shed DNA carrying mutations in cancer-associated genes, which can be mistaken for tumour signal unless white cells are sequenced alongside the plasma.