Which Antibiotic Does This Plasmid Need?
6 min read · Updated October 30, 2026
Every lab has a tube from somebody who left. Before it can be grown, three things have to be read off the sequence: what selects for it, what will carry it, and roughly how many copies per cell to expect.
All three are answerable from the sequence, and the fourth thing worth knowing is how an answer can be empty and the plasmid still carry the gene.
Two markers is an answer, not a problem
Scanning pBR322 — 4,361 bp, circular — returns two selection markers: a TetR/TetA region and AmpR (bla), which is ampicillin or carbenicillin. Both are real, and the plasmid is the textbook example of a vector with two.
That makes the practical answer better rather than worse. Plating on either works, and having two means you can use one for selection and the other for screening: an insert cloned into the tetracycline gene destroys it, so colonies that grow on ampicillin and die on tetracycline are the ones carrying the insert. That was the original point of the design.
Where two markers are a problem is a plasmid that acquired a second one by accident, which is the case the duplicate-marker check exists for.
The origin says the host and the rough copy class
The same scan returns one origin on pBR322, matched as the pMB1/ColE1 family: E. coli, high copy class. Family and copy number are not the same thing, and it is worth knowing where the looseness is — pUC19 carries an origin from the same family and sits in the hundreds against pBR322's tens, because a point mutation in the RNA II regulator and the loss of the rop gene removed the copy-number control — not because the origin is a different kind.
A coarse class is still the number you need for a prep. The practical split is whether a plasmid is in the hundreds of copies (a routine miniprep gives plenty), the tens (expect less), or single copy, where a standard prep gives almost nothing and the protocol has to change.
An empty answer is a statement about the vocabulary, not the plasmid
Run the same scan on pACYC177 and it reports AmpR and no origin at all. The plasmid has both a second marker and an origin; what happened is that neither matched the small set of signatures the scan knows, and the report says exactly that — eight marker families and eight origin families were checked.
This is the distinction worth internalising before acting on any absence: nothing matched and nothing is there are different statements, and only the first one is what a screen can establish. A larger annotation library is the next step, not a conclusion.
Paste the record and it reads your annotations instead
Hand the same plasmid over as an annotated GenBank record and the source of the answer changes: the dossier reads the record's own features rather than running the signature scan. That is better when the record is well annotated and worse when it is not, and the result says which happened.
An unlabelled origin feature comes back as an origin with no host and no copy class — we found one and cannot classify it, which is deliberately not the same output as finding none. Label the same feature p15A ori and it resolves to E. coli, medium copy.
So a sparsely annotated record gives a sparse dossier. If your record carries spans without names — which is common in public records, including pUC19's — scan the bare sequence instead and compare the two.
The closest backbone is a containment statement
Matching pBR322's sequence against a curated set of backbones returns pBR322 at 100% identity over 100% coverage, and pET-3a at 100% over 100% as well.
Both are literally true: compared against pBR322, pET-3a is the same molecule with 277 bases inserted at position 376 and two more at 2,296, and nothing else changed. Coverage counts how much of YOUR sequence the reference explains, so a reference carrying extra bases does not lose any. That is how to read this field: it says which known backbone your sequence sits inside, not which plasmid you are holding. Use the length difference to tell them apart — and remember that a vector no public record covers cannot appear in the list at all.
What else the scan flags, and why Type IIS sites are not a hazard by default
pBR322 carries one BsaI site, one BsmBI site, three BbsI sites and one SapI site. None of these is reported as a problem, and that is correct: random sequence contains Type IIS sites and so does nearly every real plasmid. They only matter if you intend to assemble with that enzyme, which is a fact about your plan rather than about the plasmid.
What is flagged without being asked: two markers of the same family, two origins for the same host, and direct repeats long enough for a recA-proficient strain to recombine out. Those are properties of the molecule that will cost you something later whatever you intend to do with it.
Frequently asked questions
How do I find out which antibiotic an unknown plasmid needs?
Scan the sequence for resistance markers. A match names the gene family and the antibiotics it covers — bla means ampicillin or carbenicillin, aph means kanamycin or neomycin, cat means chloramphenicol. If nothing matches, that is a statement about the signature set, not about the plasmid, so try a larger annotation library before concluding it has no marker.
My plasmid has two resistance genes. Which one do I use?
Either, and the second one is useful. Classic vectors such as pBR322 carry two deliberately: select on one and screen on the other, since an insert that disrupts the second marker makes carriers identifiable by their sensitivity. Check first that both genes are intact, since a marker interrupted by a cloning site no longer selects.
Can I tell the copy number from the sequence?
Only to a class. The origin family tells you whether to expect hundreds, tens or one copy per cell, which is what decides how a prep will go. It does not give a number: pUC and pBR322 origins are the same family and differ several-fold, because of a regulatory point mutation rather than a different origin.
Why does it say no selection marker when I know there is one?
Because the screen matched nothing in the families it knows, which is a different statement from the plasmid carrying none. Signature scans of this kind cover a small number of gene families and miss variants — two kanamycin resistance genes from different transposons are different sequences. Run a larger annotation library before deciding.
Related references
Related tools
Which antibiotic, which host, which backbone, and what will bite you later.
Auto-detect promoters, tags, origins and resistance markers in a plasmid, then run a deep scan against pLannotate's full feature databases.
309 publicly deposited vectors with their full GenBank feature tables, and 617 parts harvested from them.
Locate the exact direct repeats that let a construct recombine away the DNA between them, and build the shortened molecule you would actually recover.
Render a circular or linear map with restriction sites.