Diagnostics & medtech
Liquid biopsy and circulating tumour DNA
What cell-free DNA is and why its length betrays tumour origin, how sequencing error itself becomes the main obstacle, what molecular barcodes and partitioned reactions do about it, and why residual-disease monitoring beats imaging on timing alone.
Dying cells do not keep their genomes to themselves. Whenever tissue turns over — normally, under treatment pressure, or through necrosis — pieces of degraded DNA slip out of the wreckage and circulate dissolved in plasma. Most of this cell-free DNA in any person derives from unremarkable blood cells renewing themselves, but a tumour anywhere in the body contributes its own broken shards, carrying the same mutations the parent cells held. Fishing them from a routine vial replaces surgical biopsy with a needle test whose real scientific problems are arithmetical.
The signal is drowning in the matrix
Early or treated disease leaves vanishing traces: the tumour-derived share of all plasma DNA can sit near hundredths of a percent, meaning perhaps a dozen genuinely mutant molecules swim among tens of thousands of healthy lookalikes drawn from ten millilitres of blood. Two natural properties help. Cell-free DNA arrives cut to lengths matching a spool of DNA wrapped around protein cores — about one hundred sixty-seven bases — and tumour copies run marginally shorter on average, so careful size selection mildly concentrates the quarry before analysis begins. Beyond that, everything depends on reading fidelity.
Sequencing’s own lies are the wall
Modern sequencers miscall individual molecules at rates around one percent — catastrophically frequent when hunting variants expected once per thousand reads. Raw depth cannot fix this; deeper runs simply add more instrument mistakes alongside more truth. The decisive inventions are all ways of grouping reads that descend from one original molecule. Barcodes attach random tag sequences before amplification, so the many descendants of each genuine template carry identical labels and can be collapsed into a consensus that cancels copying slips; requiring agreement between both strands of the same double helix suppresses artefacts yet further. Digital PCR reaches equivalent reliability differently — emulsion droplets isolate individual template molecules, Poisson statistics convert presence-or-absence calls per partition, and counting positive wells yields quantification robust down to mutant alleles hiding among millions of normal ones.
The quiet triumph is longitudinal
A single snapshot asks whether tumour DNA exists now; serial monitoring asks something far more powerful, whether disease is growing back. Treatment records teach the test each patient’s private mutation set, customised assays then hunt exactly those sequences months afterwards, and rising quantities flag regrowth long before any scan could resolve a lesion — molecular evidence simply accumulates faster than anatomy. This explains monitoring’s clinical weight despite modest detection ranges: below some tumour burden no signal exists at all, which is silence rather than reassurance.
One biological decoy deserves permanent notice. Blood-forming cells accumulate their own age-related mutations, cloning up in elderly patients and shedding indistinguishable-looking fragments — the reason rigorous pipelines also sequence white-cell DNA to blacklist anything arising there rather than in tumour. Fragility completes the picture: cell-free DNA degrades within hours of collection, so logistics guard the measurement as much as chemistry does.