Genetic audit: PCR vs karyotype. What is the difference?

14 September 2026

Cell lines drift. Pluripotent lines drift faster than most, and gene editing, single-cell cloning and extended passage all accelerate it. At some point every laboratory working with a long-lived line has to ask whether the cells in the flask are still the cells that were banked.

Two families of methods answer that. One is cytogenetic: prepare metaphase chromosomes, stain them, and look. The other is molecular: most commonly, run a PCR assay at a set of chosen genomic positions and compare copy number against a reference. Both are described as checking the genetic integrity of a line, and both were developed for that purpose, which has led to a fairly widespread assumption that they are interchangeable and the choice is one of convenience.

They are not interchangeable, and the difference is not really about resolution or price. It is about what kind of question each method is capable of asking.

What Question Are You Asking?

A karyotype asks an open question: what does this genome look like? The analyst does not tell the assay what to find. Whatever is present in the metaphase spread is available to be seen, including things nobody had considered.

A PCR copy number panel asks a closed question: at these specific loci, is the copy number what it should be? It cannot answer any other question, because the assay is only present at the positions someone chose in advance, and those positions were chosen from what was known when the panel was designed.

That last point deserves more weight than it usually gets. A panel is a static set of markers fixed at a moment in time, while the field it serves is not static at all. Gene editing methods are changing quickly, and the abnormalities they produce are still being catalogued. A rearrangement that looks like a one-off today may turn out to be characteristic of a particular editing approach, a particular nuclease or a particular delivery method, and it can only acquire that status if somebody was in a position to notice it in the first place.

A laboratory relying entirely on a fixed marker set, whether PCR or FISH, has no mechanism for spotting an emerging pattern. It will detect the abnormalities of the previous decade with excellent precision and miss whatever the current decade is generating. Recurrence lists only grow because open-ended methods keep feeding them.

What G-banded Karyotyping Actually Shows

Cells are arrested in metaphase, swollen in hypotonic solution, dropped onto slides, treated with trypsin and stained with Giemsa. The chromosomes display a reproducible pattern of light and dark bands, and a trained cytogeneticist reads each spread against the standard ideogram for the species.

The important structural feature of the method is that it is performed cell by cell. Twenty spreads means twenty independent observations of twenty individual cells, each one scored for chromosome number and for band pattern on every chromosome. That gives a genuinely genome-wide view at a resolution of roughly 5 to 10 Mb, and it detects a category of event that no quantitative method can reach:

Numerical aneuploidy, whole chromosome gains and losses
Unbalanced structural rearrangements, including derivative chromosomes and isochromosomes
Balanced rearrangements: reciprocal and Robertsonian translocations, inversions, whether acquired in culture or constitutional
Polyploidy and near-polyploidy
Marker chromosomes, ring chromosomes, dicentrics and other structural oddities
Anything unexpected, including abnormalities that have never been catalogued in that cell type

The method needs an actively dividing cell culture so that enough metaphase cells can be obtained for a skilled cytogeneticist to interpret.

What PCR Copy Number Testing Does

The molecular approach uses a set of primer and probe assays at defined genomic positions, quantified against one or more reference loci and against a control DNA of known normal copy number. The readout is a copy number ratio at each tested position. In quantitative PCR this is derived from cycle threshold values; in digital PCR (e.g. digital drop PCR or ddPCR) the sample is partitioned and target molecules are counted directly, which removes the amplification efficiency term and improves precision.

Commercial panels for human pluripotent stem cells target the regions that recur most often in culture: gains at 1q, 12p, 17q, 20q11.21 and Xp, losses at parts of 10p and 18q. For mouse embryonic stem cells the equivalent short list is chromosomes 1, 6, 8 and 11 together with loss of the Y.

G-banded karyotyping compared with PCR copy number panels
Attribute	G-banded karyotype	PCR copy number panel
Genome coverage	Complete, every chromosome in every cell scored	Only the loci on the panel, typically 6 to 24 positions
Resolution	Approximately 5 to 10 Mb, everywhere	Sub-kilobase, but only where an assay sits. Everything else is missed
Balanced rearrangements	Detected, acquired and constitutional alike	Not detectable in principle
Polyploidy	Detected immediately	Not detectable in principle
Novel or uncatalogued abnormalities	Detected	Never detected
Rearrangement structure and origin	Described directly	Not available, only a dosage value
Mosaicism sensitivity	Set by number of cells counted, see below	Approximately 20 to 30% (qPCR), 5 to 10% (dPCR)
Sample requirement	Live, actively dividing cells which can be prepared and fixed in any lab	Genomic DNA, including from frozen pellets
Output	ISCN formula, a formal description of the genome	Pass or flag at each tested locus

Four Things a PCR Panel Cannot See

Balanced rearrangements

A reciprocal translocation exchanges material between two chromosomes without gaining or losing any. Copy number at every locus in the genome is unchanged. An inversion flips a segment in place, again with no dosage change. A Robertsonian translocation fuses two acrocentric chromosomes and loses only the short arm satellite material, which no panel targets.

These events are invisible to any method that works by measuring dosage, and that includes digital PCR and array CGH as well as qPCR. They are not rare enough to ignore. In the large mouse ES cell survey run as part of the International Mouse Phenotyping Consortium, around 1% of clones carried a translocation, and work on individual mouse ES lines has documented Robertsonian rearrangements involving chromosome 11 arising in culture. In human lines, unbalanced translocations onto a partner chromosome are a common route by which a 1q or 17q gain is acquired, and the balanced precursor states pass unremarked by dosage assays.

Not all of them are acquired in culture either. Some balanced rearrangements are constitutional, present in the donor and carried by the line from the moment it was derived. A karyotype identifies these at first characterisation, which matters for interpreting any phenotype the line subsequently shows and for the provenance record that travels with it. A dosage panel gives no signal at all that such a rearrangement exists, so the line proceeds on the assumption that it does not, and a constitutional finding that should have been documented at the outset instead surfaces years later, if at all.

Polyploidy

This one is worth dwelling on, because it is the cleanest illustration of the underlying problem. Take a tetraploid line. Every locus is present at four copies rather than two. Every reference locus is also at four copies. The ratio at every position on the panel is exactly 1.00, and the sample reports as entirely normal.

A karyotype identifies the same line in a single field of view, because the analyst counts 92 chromosomes instead of 46. Tetraploidy and near-tetraploidy are recurrent in long-term culture and in some editing workflows, and a genome-wide dosage method cannot help here either. Chromosome counting is the only routine assay that catches it.

Anything not on the panel

Commercial hPSC panels are honest about their coverage. STEMCELL Technologies states that the target regions in its qPCR kit account for approximately 75% of abnormalities observed by cytogenetics, and that the kit is not intended as a substitute for karyotyping. The ddPCR-based iCS-digital test from Stem Genomics claims better than 90% coverage of recurrent abnormalities, where "recurrent" was defined in the underlying meta-analysis as an abnormality reported in at least five separate publications.

Those are respectable figures, and the honesty is welcome. But notice how the target lists were built. They come from meta-analysis of published karyotype and array data. Every locus on every panel is there because cytogenetics found it first. A panel is by construction a summary of what open-ended methods have already discovered, which means it is always looking backward, and it cannot contribute anything to the catalogue it depends on.

The practical consequence is a specific reading of a negative result. A clean panel does not mean the line is normal. It means that none of the tested regions is abnormal. Those are very different statements, and the gap between them widens the further the line has travelled from a standard workflow, which is exactly the situation in which people reach for a genetic audit.

Structure and origin

Suppose a panel reports a gain at 17q. That is useful, but it does not tell you whether the extra material sits as a tandem duplication in place, as an isochromosome, or translocated onto another chromosome and displacing material there. Those three outcomes have different implications for how the clone will behave over subsequent passages and for whether the event is likely to be stable or to progress.

The karyotype gives you the structure directly, written out in ISCN notation, and that description is what allows a decision about whether the line can be recovered from an earlier vial or should be discarded.

Mosaicism: the Comparison Most People Get Backwards

A frequent claim is that PCR is the more sensitive method for detecting a variant subpopulation. The reasoning is that a karyotype samples only twenty cells while a PCR assay measures the whole population at once. It sounds right, but the arithmetic does not support it.

If a fraction of cells in a culture carries an abnormality and every metaphase examined comes back normal, the upper confidence limit on that fraction follows directly from the binomial. The classic tabulation is Hook's, and the figures are these.

Mosaicism excluded by counting metaphases, all scored normal
Metaphases counted	Abnormal fraction excluded at 95% confidence	Comparable to
20	14%	Better than a typical qPCR panel
30	9.5%	Comparable to digital PCR
50	5.8%	Better than digital PCR
100	3.0%	Beyond the reach of any PCR panel

A standard twenty-cell count already excludes mosaicism above about 14%, which is better than the 20 to 30% limit that qPCR panel suppliers quote for their own kits. Thirty cells brings the karyotype level with digital PCR. Fifty puts it ahead.

Karyotype sensitivity to mosaicism is also a purchasing decision. If a line is destined for a clinical programme or a critical handover, order an extended count and buy the sensitivity you need. PCR sensitivity is fixed by measurement precision, and no amount of extra spend moves it, because the limit is set by how finely the method can resolve a ratio of 1.00 from a ratio of 1.05.

There is a further asymmetry. A karyotype detects a mosaic clone as a set of discrete abnormal cells, so you also learn what those cells are carrying and can track the clone at the next passage. Where a line is already abnormal, the mosaic picture is more informative still. Seeing which additional changes are appearing, in which cells, and in what proportion of the population shows how the overall chromosome complement is evolving, and whether a particular subclone is being selected. That is the difference between knowing a line has an abnormality and understanding where the line is heading.

A PCR panel reduces the same situation to a small shift in an averaged ratio, with no view of the underlying population structure and no information about what else those cells are carrying.

Array CGH

Array comparative genomic hybridisation sits between the two. It is genome-wide, unlike a PCR panel, and its resolution is far higher than G-banding, typically in the tens of kilobases. It picks up the sub-microscopic imbalances that banding misses, including the 20q11.21 amplicon, and it does so without needing to know in advance where to look.

Its blind spots are the ones shared by all dosage methods. Balanced rearrangements do not appear. Polyploidy does not appear. Low-level mosaicism is limited to roughly 20 to 30% depending on platform and segment size.

Array CGH and karyotyping are therefore complementary rather than competing, and a full characterisation of a line for a regulated application usually runs both. Neither is displaced by a targeted PCR panel.

Building an Audit Strategy That Works

The layered version is the one that survives contact with reality, and it does not require choosing a single method.

Full karyotype at the decision points. On derivation, at master and working bank creation, before any critical experiment, before transfer of a line to a collaborator, and before committing to an expensive downstream process such as microinjection or differentiation at scale. These are the moments where an undetected balanced rearrangement or a tetraploid subclone costs real money, and they are also the moments where you want a formal, quotable ISCN description on file rather than a set of pass flags.

Array CGH where resolution is the binding constraint. Particularly for lines that have been through editing, and for anything intended for clinical translation where sub-microscopic imbalance is a documented concern.

PCR panel for the routine gaps in between. Weekly or per-passage screening, with the understanding that a clean result means the tested regions are clean and nothing more. Escalate any flag to cytogenetics for characterisation, because the panel tells you something has changed but not what.

The framing that clarifies most decisions: a PCR panel is a screen, a karyotype is a characterisation. Screens are valuable when their false-negative profile is understood and when there is a defined path from a positive result to a proper answer. Problems start when a screen is treated as though it were a characterisation, which is what happens when a line is released on a panel result alone.

Questions Worth Asking Before You Choose

Has this line been edited, single-cell cloned, or carried past passage 30? If so, the probability of something off-panel rises and open-ended methods earn their cost.
Does anyone downstream need a formal karyotype record, for a regulator, a journal, a funder or a collaborating institution? A panel report does not substitute.
Is there a specific known risk, such as 20q11.21 in human pluripotent lines or trisomy 8 in mouse ES lines? Targeted testing is efficient against a named target and only against a named target.
Would a constitutional abnormality in the donor material change how results from this line are interpreted? Only cytogenetics will find one.
What happens if the result is a false negative? If the answer is a failed germline transmission round or a wasted differentiation campaign, the economics favour the more thorough method by a wide margin.
Are the cells dividing well enough to give metaphases? If not, that is a genuine constraint and a dosage method may be the only option available.

Summary

PCR copy number panels and karyotyping are frequently presented as alternatives, and the typical framing is not accurate. A panel measures dosage at positions chosen in advance. A karyotype describes the genome as it is, cell by cell, without being told what to look for.

Balanced translocations, inversions, polyploidy, marker chromosomes and every abnormality that has not yet made it onto a published recurrence list are invisible to dosage methods and visible in a metaphase spread. Constitutional rearrangements sit in the same category and are found only by looking. On mosaicism, the karyotype matches or beats a qPCR panel at a standard twenty-cell count and pulls further ahead as the count rises, and it shows the population structure behind the number rather than an average across it.

G-banding has a resolution floor at around 5 to 10 Mb, and the 20q11.21 amplicon is a real example of an important abnormality that can fall below it. That gap is closed by FISH or array CGH rather than by abandoning cytogenetics.

Where the target is named and the question is narrow, a PCR panel is the sensible tool. Where the question is whether a line is sound, or where the field is still working out what its new methods do to genomes, the assay has to be one that can return an answer nobody anticipated.