The Control Plates That Decide Why There Are No Colonies
6 min read · Updated October 30, 2026
A failed cloning experiment is not short of explanations; it is short of a way to tell them apart. The plate is blank, and the reaction, the cells, the plate and the construct are all candidates.
Three control plates settle most of it, and they are worth running before the experiment rather than after it fails.
What each control eliminates
- No DNA, same cells, same plate. Colonies here mean the plate is not selecting — the antibiotic has lost activity or the cells are already resistant. Nothing else you conclude is safe until this is zero.
- An intact, uncut plasmid into the same aliquot of cells. Colonies here prove the cells took up DNA and grew, which eliminates the competent-cell branch entirely. It is the single most informative plate.
- The identical reaction with the insert, or one part, left out. This separates the assembly from everything downstream: if that plate is as blank as the experimental one, the reaction is not the difference.
- The digested but unligated vector. Colonies here measure intact vector that survived the digest with no help from the ligase, which is where empty-vector background comes from.
Worked: a Golden Gate with no colonies
Take a Golden Gate assembly with BsaI, a kanamycin vector plated on kanamycin, and a blank plate. On the design alone, three causes remain in play: the assembly produced no circular DNA, the cells never took up DNA, and the construct kills the host.
The cheapest next observation is named along with why it works: run the identical reaction with one part left out and plate it alongside. That settles the assembly branch and cannot change the cell-competence branch, so whichever way it comes out, one of the two is eliminated.
Now add the controls: 0 colonies on the experimental plate, 0 with no DNA, 240 from an intact plasmid carrying the same marker, and 3 from the unligated vector. The cell-competence cause is eliminated by name — 240 colonies from an intact plasmid into the same cells means those cells took up DNA and grew — and two causes remain.
The useful part is which observation did the eliminating
An elimination is only as good as the observation behind it, and the two are easy to confuse. A positive control that carried a different resistance marker proves the cells are competent and says nothing about whether your plate selects, so those two conclusions have to stay apart.
Leaving a control field blank is also not the same as entering zero. Not having run the no-DNA plate and having run it and seen nothing are different states of knowledge, and only the second one eliminates anything.
When nothing separates the last two
In the worked case, the two causes left are the construct killing the host and the assembly producing nothing. No single observation separates them: both are settled by the same evidence, so the honest output says so rather than ranking one above the other.
That is where the experiment has to change instead of being observed harder. Plating the same reaction at 30 °C, and in parallel into a tighter-repression host or onto a lower-copy backbone, separates them — a construct that appears only at 30 °C is a toxicity result, not a cloning result.
A diagnosis that always produces a ranked list is less useful than one that admits when the inputs do not discriminate.
Things the design settles without any plate
Some causes are decided by the sequences and the paperwork rather than by observation. A kanamycin vector plated on ampicillin is the clearest: no colony can grow however well the assembly worked, and the mismatch is a fact about the design.
Others in the same class: a part with no Type IIS site left in it, two vector ends that are compatible with each other, a primer pair that pairs through its own 3' ends, and a methylation site sitting across a recognition sequence in DNA prepared from a dam+/dcm+ strain. Each of those is checkable before anything is plated.
Which is why the methylation question asks about the strain the DNA was PREPARED from rather than the one it was transformed into — the enzyme met the methylation during the digest, not after.
No probability is computed anywhere
The ordering is by evidence, not by likelihood: a cause with a design fact and two observations behind it ranks above one with a single assumption, and nothing here estimates how often each cause occurs in the world.
That is a deliberate limit. The frequencies would have to come from somewhere, they would be lab-specific, and a number attached to a guess is harder to argue with than the guess itself.
Frequently asked questions
Which control plate should I run if I only run one?
An intact, uncut plasmid into the same aliquot of cells, plated on the same plates. Colonies there eliminate the whole competent-cell branch in one observation, and the absence of colonies points at the cells or the plates rather than at your assembly.
I got colonies on my no-DNA control. What does that mean?
The plate is not selecting. Either the antibiotic has lost activity, the concentration is wrong, or the cells already carry resistance. Nothing else from that experiment is interpretable until it is fixed, including a result that looks like success.
What does the unligated vector control tell me?
How much intact vector survived the digest. Colonies on that plate are vector that recircularised or was never cut, and they set the empty-vector background you will screen through. A high count points at an incomplete digest or at skipping dephosphorylation on a single-cut vector.
Why will it not just tell me the most likely cause?
Because likelihood would have to be invented. The ordering is by the evidence your design and observations actually provide — a cause backed by a design fact and two observations ranks above one resting on an assumption — and when the inputs separate nothing, the honest output is to say so and name the experiment that would.
Related tools
No colonies, every clone empty vector, or no PCR band: get the causes your design and your control plates actually implicate, and the cheapest experiment that separates the top two.
Assemble fragments and design junction primers for Gibson, Golden Gate or restriction cloning.
Find the restriction digest that tells your intended construct apart from the empty vector, a flipped insert or a mis-assembly — with the bands you would actually be able to read.
Diagnose gel problems (no bands, smearing, wrong size, primer-dimer, degradation) with a fully deterministic symptom guide that links to the right SeqBench tools.
Paste what you got back and the PCR you say made the insert. It re-derives the insert, finds it in either orientation, and gives you the exact mismatch positions.