Loading
Old version — revision 1
This is a fixed snapshot of Nanoscale diagnostics, saved by Import as part of the initial corpus import. It is not edited and it is not updated; the article may have changed since.
Edit summary: Initial import of content/nanoparticle-diagnostics.md — the filesystem corpus, unchanged. Not an edit.
Detection methods that read single molecules or use nanoscale labels, including liquid biopsy, multi-cancer blood tests, and nanopore sensing, together with their overdiagnosis costs.
Nanoscale diagnostics covers detection methods that operate on individual molecules or use nanoscale particles as labels and sensors. The category spans the mundane and the frontier: the coloured line on a home antigen test is a stripe of gold nanoparticles, while single-molecule sequencing through an engineered pore reads a DNA strand base by base. The clinically consequential branch as of 2026 is the blood test for fragments of tumour DNA, which is closer to changing cancer screening than any other technology in this wiki's nanomedicine section.
The oldest working application uses nanoparticles as visible or magnetic reporters. Colloidal gold conjugated to antibodies aggregates at a capture line on a nitrocellulose strip, producing the red band of a lateral flow test; the format is cheap, needs no instrument, and was manufactured in billions of units during the COVID-19 pandemic. Superparamagnetic iron oxide particles serve as magnetic resonance contrast agents and, in one cleared platform, as the readout for detecting bloodstream pathogens by their effect on water relaxation times, avoiding the delay of culture.
Quantum dots — semiconductor nanocrystals with narrow, size-tunable emission — became standard in research imaging and multiplexed assays but have made little clinical progress, partly because the brightest formulations contain cadmium.
These are labels rather than sensors. The engineering that matters is surface chemistry: keeping particles from aggregating, attaching antibodies at the right density, and preventing non-specific binding. It is the same body of knowledge that governs Targeted drug delivery, applied to a much easier problem, because a particle in a test strip does not have to survive blood, evade phagocytes, or find an organ.
Cell-free DNA circulates in plasma in everyone. Tumours shed their own, and that circulating tumour DNA carries the mutations, copy-number changes, and methylation patterns of the cancer that produced it. Sampling it means sampling a tumour without a needle, repeatedly, and capturing heterogeneity a single core biopsy would miss.
The technical difficulty is signal fraction. In advanced disease tumour DNA may be a few percent of the total; in early-stage disease it is often below one part in a thousand, so detection requires deep sequencing, molecular barcoding to distinguish real variants from sequencing errors, and statistical models that aggregate weak evidence across many sites. Methylation-based methods have proved more sensitive than mutation-based ones for early detection, because a tumour genome carries far more differentially methylated positions than driver mutations, and the pattern also indicates the tissue of origin. The same methylation arrays and sequencing panels underlie the Epigenetic clocks literature, which reads the marks for a different purpose and faces a much weaker validation standard.
Three applications have separated out. Genotyping advanced cancers to select targeted therapy is routine practice. Minimal residual disease testing after surgery detects recurrence months before imaging and has been used in randomised trials to decide who needs adjuvant chemotherapy, in one colon cancer study reducing chemotherapy use without worsening recurrence-free survival.1 Screening asymptomatic people is the most consequential and the least settled.
Multi-cancer early detection tests look for a shared signal across dozens of cancer types in a single blood draw, most using methylation patterns. In a prospective study of several thousand adults, just over a third of positive signals corresponded to a confirmed cancer, with specificity above 99 percent.2 A specificity of 99 percent sounds excellent and is not, when the prevalence of undiagnosed cancer in a screened population is well under one percent: most positives in such a setting are false, and each one triggers imaging, biopsy, and months of uncertainty.
A single-cancer version cleared the regulatory bar first. A blood test for colorectal cancer approved in the United States in 2024 detected about 83 percent of cancers at 90 percent specificity in a large screening cohort, but only about 13 percent of advanced precancerous lesions.3 That asymmetry matters: colonoscopy prevents cancer by removing adenomas, whereas a blood test that misses nearly nine in ten of them mostly finds disease that has already arrived. Its plausible value is in reaching people who decline colonoscopy at all.
The largest randomised trial of multi-cancer screening, enrolling roughly 140,000 people in England, was designed to measure whether the approach reduces late-stage diagnoses. As of 2026 no national screening programme has adopted a multi-cancer blood test.
The screening trapFinding cancer earlier is not the same as helping patients. Lead-time bias makes survival from diagnosis look longer whenever diagnosis moves earlier, even if death occurs at the same moment. Length bias means screening preferentially catches slow-growing tumours, which are the ones least likely to kill. Only a reduction in mortality, measured against an unscreened control arm, distinguishes benefit from statistical artefact.
Overdiagnosis is the detection of disease that would never have caused symptoms. It is not a false positive: the cancer is really there under the microscope. It would simply never have progressed.
The clearest natural experiment came from South Korea, where widespread ultrasound screening of the thyroid increased diagnosed incidence many times over while mortality from thyroid cancer stayed flat — the signature of finding disease that did not need finding.4 Most of those patients had their thyroid removed and took hormone replacement for life.
Blood-based multi-cancer tests could reproduce this at scale, or could avoid it. The optimistic argument is that a tumour shedding enough DNA to be detected is metabolically active and therefore more likely to be consequential, which would make the modality biased toward aggressive disease rather than indolent disease. The argument is plausible and unproven; the evidence that would settle it is mortality data from randomised trials that are due to report over the coming years.
Nanopore sensing threads a molecule through a protein or solid-state pore a few nanometres wide and reads the resulting change in ionic current. The principle was demonstrated for polynucleotides in 1996 and became a commercial sequencing platform in the 2010s, giving very long reads from a device that fits in a hand and that has been used for genomic surveillance during outbreaks in field conditions.5 Work on reading peptides and post-translational modifications by the same method has advanced considerably, and if it matures it would provide something proteomics currently lacks: single-molecule sequencing of proteins.
A different route uses CRISPR–Cas9 machinery as a detector. Collateral cleavage by certain Cas enzymes, triggered when a guide RNA finds its target, cuts a labelled reporter and produces a fluorescent or lateral-flow signal, allowing nucleic acid detection at low cost without thermal cycling.
Activity-based synthetic biomarkers invert the usual approach. Rather than looking for a molecule the body produces, they inject a nanoparticle carrying peptide substrates for disease-associated proteases; cleavage releases reporter fragments that concentrate in urine, amplifying a local enzymatic activity into an easily measured signal.6 The approach remains in early clinical development. It is the closest working thing to a diagnostic nanodevice: an injected particle that performs a computation of sorts in tissue and reports the answer, though without any of the autonomy the designs in Medical nanorobots assume. Structural containers built by DNA nanotechnology and the steerable devices of Medical microrobots have both been proposed as sensing platforms, and neither has entered human testing.
Analytical sensitivity has outrun clinical interpretation. Detecting a hundred tumour DNA molecules in a tube of blood is now tractable; knowing what to do about them frequently is not, and clonal haematopoiesis — mutated blood cell clones that accumulate with age and shed the same variants tumours do — is a persistent source of confusing signals that also connects these assays to the biology of Stem cell exhaustion and, more broadly, to the somatic mutation entry in the Hallmarks of aging.
Cost and access shape who benefits. A test priced for self-pay markets, delivered outside organised screening programmes and increasingly alongside the panels covered under Consumer blood testing, tends to reach people already well served by medicine, which is the pattern described in Access and inequality. Insurance and employment consequences of a positive result raise the questions covered in Genetic discrimination.
The most interesting unresolved application is not cancer at all. If the same sequencing and proteomic depth were applied longitudinally to healthy people, it might supply the surrogate endpoints that the field of Aging biomarkers has been unable to validate, put a measurement behind claims about Biological age, and give trials in Geroscience hypothesis something to measure other than death. Whether a molecular signal read from plasma can track something as diffuse as biological ageing, rather than a discrete tumour, is a much harder question than early cancer detection, and no assay has yet answered it.
paperTie, J. et al. "Circulating Tumor DNA Analysis Guiding Adjuvant Therapy in Stage II Colon Cancer." New England Journal of Medicine, 2022. ↩
paperSchrag, D. et al. "Blood-based tests for multicancer early detection (PATHFINDER): a prospective cohort study." The Lancet, 2023.↩A single-arm cohort with no unscreened comparison group, so it reports how the test performed and cannot show whether screening changed any outcome.
paperChung, D.C. et al. "A Cell-free DNA Blood-Based Test for Colorectal Cancer Screening." New England Journal of Medicine, 2024. ↩
paperAhn, H.S., Kim, H.J., Welch, H.G. "Korea's Thyroid-Cancer 'Epidemic' — Screening and Overdiagnosis." New England Journal of Medicine, 2014.↩An analysis of national incidence and mortality trends rather than a trial; the overdiagnosis inference rests entirely on the divergence between the two curves.
paperKasianowicz, J.J., Brandin, E., Branton, D., Deamer, D.W. "Characterization of individual polynucleotide molecules using a membrane channel." PNAS, 1996. ↩
paperKwong, G.A. et al. "Mass-encoded synthetic biomarkers for multiplexed urinary monitoring of disease." Nature Biotechnology, 2013.↩The demonstration was in mouse disease models, not in patients.