Your Plasmid Consensus Does Not Line Up With Your Map
6 min read · Updated October 30, 2026
A whole-plasmid sequencing service returns one assembled circular consensus. Lined up against your map as written, it usually disagrees everywhere — and almost always for reasons that are not about your clone.
There are three of them, and each has to be resolved before a difference list means anything.
A circle has no first base
The assembler started the circle wherever the assembly happened to close, which has nothing to do with where your map starts. Comparing the two as written aligns base 1 of the consensus with base 1 of the design, and from there every position disagrees.
The fix is to find the rotation. In a worked example on a 4,361 bp plasmid, the consensus was rotated by 2,860 bases relative to the design — once that is applied, the two line up and the real differences stand out as the two they are.
This is also why a difference list has to be reported in the design's coordinates rather than the consensus's. Position 428 of the consensus and position 1928 of the design are the same base, and only one of those numbers matches the map you are reading.
Half the time it is the other strand
Nothing about a circular assembly fixes which strand is written out, so roughly half of consensus sequences come back as the reverse complement of the design. In the same worked example the consensus was both rotated and reverse-complemented, and neither is a property of the clone.
A sequence that is correct but written on the other strand does not produce a mutation list. Compared as written, a 1,000 bp stretch of that plasmid against its own reverse complement reads 43.76% identity — the number two unrelated sequences of the same composition would give, and nothing you can act on. Resolving rotation and strand first turned the same comparison into 99.954% identity with two differences.
A consensus can be two or three copies of your plasmid
Rolling-circle amplification and some assembly paths produce a consensus that is a dimer or trimer — the same plasmid written out two or three times end to end. The length is then a clean multiple of the design's, which is the signature to look for.
Compared naively this reads as a plasmid of the wrong size. Compared one copy at a time, it is usually your plasmid, and the copies agreeing with each other is reassurance rather than evidence: they came from the same molecule.
An indel in a homopolymer is the artefact that looks exactly like a mutation
The commonest long-read basecalling error is a single base inserted into or deleted from a run of identical bases, and the commonest real slippage mutation is the same event. From the difference alone they cannot be told apart.
So the run length travels with the difference. In the worked example the deletion at design position 1928 sits inside a run of six A's, which is flagged; the substitution at position 200 is not in a run and is flagged as nothing. The first one needs confirming by a second method before you act on it; the second is a difference you can believe.
The same applies to the sequence contexts providers document as error-prone, such as Dam (GATC) and Dcm (CCWGG) methylation sites. Marking them is not a verdict — it is a note that this particular difference deserves a Sanger read before it changes a decision.
With an annotated design, a position becomes a consequence
Comparing against bare bases gives you positions. Comparing against the annotated GenBank record gives you what those positions do: in the worked example, the substitution at design position 200 falls inside the tetA coding sequence at codon 39, changing Asp to His — a missense change, reported as p.Asp39His. The homopolymer deletion at 1928 falls outside both features of the design used here — which annotates only tetA and bla — and is reported as non-coding. Annotate the same plasmid more fully and that position sits inside the ROP coding sequence, where the same deletion is reported as a frameshift: the consequence column is only as complete as the map you supply.
That difference in output is the difference between a clone you have to think about and a clone you can use, and it costs nothing except pasting the map rather than the sequence.
What this comparison does not establish
- That the clone is pure. A consensus is one sequence; a culture grown from two colonies collapses to one answer and the comparison reports it as clean.
- That a region is covered. A consensus can be assembled over a region few reads crossed, and the per-base depth that would show it is in a separate file.
- That a flagged difference is an error. The flag says the context is one where errors are common, not that this one is.
- That an unflagged difference is real. It says only that the commonest artefact does not explain it.
Frequently asked questions
Why does my sequencing consensus start at a different base from my map?
Because a circular assembly has no defined first base: it starts wherever the assembly closed. The offset carries no information about the clone. Find the rotation, apply it, and report differences in the map's coordinates — otherwise every position in the report is unusable against the map you are reading.
My consensus is the reverse complement of my plasmid. Is something wrong?
No. Which strand gets written out is arbitrary for a circular assembly, and roughly half come back reverse-complemented. Compared without resolving it, a perfect clone shows essentially no identity — the alarming number is the artefact, not the clone.
Is a 1 bp indel in a homopolymer a real mutation?
It cannot be decided from the consensus alone. A single base inserted into or deleted from a run of identical bases is both the commonest long-read basecalling artefact and what a real slippage mutation looks like. Confirm it with a Sanger read across the run, or with a second library, before acting on it.
My consensus is exactly twice the length of my plasmid. What happened?
The assembly produced a concatemer — your plasmid written out twice end to end, which rolling-circle amplification readily produces. Compare one copy at a time. The two copies agreeing tells you nothing extra, because they are the same molecule read twice.
Related references
Related tools
Drop the results your sequencing service sent and the map it should be — every sample of the order checked at once.
Align your reads to your own annotated reference and get what each difference DOES — silent, missense at a named codon, or a premature STOP that ends the protein — not just where it sits.
Inspect chromatograms, compare a read with its reference, or screen a whole plate of clones.
Auto-detect promoters, tags, origins and resistance markers in a plasmid, then run a deep scan against pLannotate's full feature databases.
Align a query to a reference and list substitutions, insertions and deletions with effects.