How Long Should a Gibson Assembly Overlap Be?
11 min read · Updated August 11, 2026
Every Gibson protocol tells you to add an overlap to your primers, and then quotes a number — 15 bp, 20 bp, 40 bp — without saying what the number is for. So people pick one, get a 60-mer they have to pay for, and have no idea whether the assembly failed because the overlap was too short or because the primer would not anneal.
The confusion has a single cause: a Gibson primer carries two melting temperatures that are independent of each other. This guide separates them, shows what each one is worth with measured numbers, and covers where the overlap should sit.
What the overlap is actually doing
Gibson assembly is one isothermal reaction with three enzymes (Gibson et al., Nat Methods 2009;6:343-5, doi:10.1038/nmeth.1318). A 5' exonuclease chews back one strand from the ends of every fragment, exposing single-stranded 3' overhangs. Overhangs that are complementary anneal to each other. A polymerase fills the gaps and a ligase seals the nicks.
The overlap you design is the sequence that becomes those complementary single strands. It is not a primer in the reaction — nothing anneals to it in order to be extended by a thermal cycler. It is a sticky end, made enzymatically instead of by a restriction enzyme, and its job is to find its partner among all the other fragment ends in the tube and stay annealed at the assembly temperature.
That is a completely different job from the one the rest of your primer does, which is to anneal to your template in a PCR and prime synthesis. Two jobs, two temperatures, one oligo.
The two temperatures, measured
Here is the same two-fragment circular assembly designed five times, changing nothing but the requested overlap length. The annealing arm is designed to a 60 °C target in every case; only the overlap moves.
- 15 bp overlap — overlap Tm 53.6 °C — primer 34 nt — annealing Tm 61.8 °C
- 20 bp overlap — overlap Tm 62.8 °C — primer 39 nt — annealing Tm 61.8 °C
- 25 bp overlap — overlap Tm 69.1 °C — primer 44 nt — annealing Tm 61.8 °C
- 30 bp overlap — overlap Tm 75.4 °C — primer 49 nt — annealing Tm 61.8 °C
- 40 bp overlap — overlap Tm 78.1 °C — primer 59 nt — annealing Tm 61.8 °C
Reading that table
The annealing temperature does not move. Not by a tenth of a degree across a 25 bp range of overlap lengths, because the annealing arm is a separate piece of sequence chosen against your template and the overlap is a 5' tail hanging off it that the template never sees. If your PCR is failing, lengthening the overlap will not help; if your assembly is failing, retuning the annealing arm will not help either.
What the overlap length buys is overlap stability, and it buys it with diminishing returns. The first 5 bp, from 15 to 20, is worth 9.2 °C. The last 10 bp, from 30 to 40, is worth 2.7 °C. Past about 30 bp you are mostly paying for oligo.
What it costs is length, linearly: 34 nt at a 15 bp overlap, 59 nt at 40 bp. That matters commercially and practically — synthesis price brackets step up, and the proportion of full-length product in a crude oligo falls as length rises, so a 60-mer delivers a smaller fraction of correct molecules than a 35-mer does.
So the design question is not "what is the recommended overlap" but "what is the shortest overlap that is comfortably stable at my assembly temperature". For a reaction incubated around 50 °C, where these assemblies are normally run, a 20 bp overlap at 62.8 °C has room to spare. A 15 bp overlap at 53.6 °C is closer to the edge, and on a GC-poor junction it would be closer still — which is the case where lengthening is the right move.
GC content moves the answer, so design in base pairs and check in degrees
The overlaps above ran 50-56% GC, which is why the length-to-temperature relationship looks so tidy. Junctions in real constructs are not so obliging: an AT-rich junction at 15 bp can sit 10 °C below a GC-rich one of the same length.
This is the argument for reading the overlap's melting temperature rather than trusting a length. Two junctions in the same assembly can be the same nominal design and differ by more than the margin you were relying on, and it is always the weaker junction that fails — assembly success is set by the worst junction, not the average.
It is also the argument for checking every junction rather than the one you were thinking about. A three-fragment assembly has three junctions and you probably only designed two of them consciously.
Where the overlap goes
There are two conventions and they produce different oligos for the same assembly.
One primer carries the whole overlap. At each junction, the primer amplifying one fragment gets the full overlap as a 5' tail, and the primer amplifying its neighbour gets none — it just anneals. In the worked example above this is visible in the output: each fragment's forward primer carries a 25 nt tail and its reverse primer carries none, so the primers come in two flavours, long and short.
The overlap is split across both. Each of the two primers meeting at a junction carries roughly half, so both are moderately long and neither is very long. This keeps all your oligos in one length bracket, at the cost of the overlap only existing once both PCRs have worked.
Neither is wrong. The single-primer convention is simpler to reason about — the overlap sequence is exactly what you typed and it lives in exactly one oligo — and it is what a generated design will normally give you. What matters is knowing which one you are looking at before you paste sequences into an order form, because a 44-mer and an 18-mer at the same junction are not a mistake.
The junction is what you should verify, not the product length
Assemblies rarely fail by producing nothing. They fail by producing something plausible: a backbone that recircularised without the insert, two fragments joined in the wrong order, a junction that lost a few bases.
So build the intended product in silico before you order anything and read the junction sequences themselves. A simulation that returns the assembled sequence plus each junction's overlap, GC and Tm gives you three things to check: that the product is the length you expected, that each junction's overlap is the sequence you intended, and that the weakest junction is still comfortable.
Two failure modes are worth looking for specifically. Overlaps that resemble each other — which happens easily when fragments share a promoter or a terminator — can mis-pair, and the fix is to move a junction rather than to lengthen it. And an overlap that lands inside a repeat gives the exonuclease-exposed strand more than one place to anneal, with the same result.
After the assembly works, the junctions are also the thing to sequence. Whole-plasmid sequencing has made this cheap enough that there is no longer a good reason to check only the insert.
A design recipe
- Decide the fragment order and where each junction falls. Prefer junctions in unique, moderately GC-balanced sequence over ones inside repeats or shared regulatory elements.
- Design the annealing arms first, to a single Tm target so every primer works in the same PCR — or in PCRs you are willing to run separately, which is the usual reality.
- Start at a 20 bp overlap and read its melting temperature rather than assuming. Lengthen only the junctions that come back cool.
- Check every junction in the assembly, not the one you were thinking about. The weakest sets the outcome.
- Simulate the full product and read the assembled sequence, not just its length.
- Check the oligo lengths you are about to order. If a long overlap has pushed a primer past a synthesis price bracket, ask whether 5 bp less would still be comfortable.
- Linearise the backbone cleanly, and if you are cutting it with an enzyme, confirm that enzyme cuts exactly once first.
The tool
SeqBench's Cloning Simulator assembles fragments by Gibson/overlap, Golden Gate or restriction-ligation and returns the product plus the junction primers. For an overlap assembly you give it the fragments, whether the product is circular, the overlap length and a target Tm for the annealing arms; it returns the assembled sequence, every junction with its overlap sequence, GC and melting temperature, and a forward and reverse primer per fragment with the tail length and annealing Tm reported separately from each other.
That separation is the point. Every number in this guide came from running it and changing one parameter at a time, which is also the fastest way to answer the question for your own construct rather than adopting a number from a protocol written for someone else's.
Frequently asked questions
How long should a Gibson assembly overlap be?
Long enough to be comfortably stable at your assembly temperature, which is a temperature question rather than a length one. On one measured design at ~50% GC, a 15 bp overlap gave an overlap Tm of 53.6 °C, 20 bp gave 62.8 °C, 25 bp gave 69.1 °C and 40 bp gave 78.1 °C — so the first 5 bp past 15 is worth 9.2 °C and the last 10 bp up to 40 is worth 2.7 °C. Start around 20 bp, read the melting temperature, and lengthen only the junctions that come back cool.
Why are my Gibson primers so long?
Because the oligo is two sequences: an annealing arm that primes the PCR and a 5' overlap tail that becomes the sticky end. In one design the same annealing arm produced a 34 nt primer with a 15 bp overlap and a 59 nt primer with a 40 bp overlap. Length costs money and quality — the fraction of full-length product in a crude oligo falls as length rises — so trim the overlap to the shortest one that is still comfortably stable rather than defaulting to the longest number in the protocol.
Does a longer overlap raise my PCR annealing temperature?
No, and this is the most useful thing to know about Gibson primer design. The overlap is a 5' tail the template never sees, so it does not participate in the first rounds of annealing. Across overlaps from 15 to 40 bp on the same design, the annealing temperature stayed at 61.8 °C throughout while the overlap Tm moved from 53.6 to 78.1 °C. If the PCR is failing, fix the annealing arm; if the assembly is failing, fix the overlap.
Should both primers at a junction carry the overlap?
Either convention works. Giving the whole overlap to one primer produces one long oligo and one ordinary one at each junction, which is simpler to reason about and is what most generated designs return. Splitting it across both keeps every oligo in a similar length bracket. The important thing is knowing which you have — a 44-mer and an 18-mer meeting at one junction is the single-primer convention, not a mistake.
Why did my Gibson assembly give the wrong product?
Usually a junction problem rather than an overlap-length problem. Check whether two junctions in the assembly have similar overlap sequences, which happens easily when fragments share a promoter or terminator and lets ends mis-pair; whether an overlap falls inside a repeat, which gives the annealing strand more than one home; and whether the backbone can recircularise on its own. Simulate the intended product first and read the junction sequences rather than only checking the expected length.
How does Gibson assembly actually work?
One isothermal reaction with three enzymes, as described by Gibson and colleagues in 2009. A 5' exonuclease recesses one strand from each fragment end, exposing single-stranded 3' overhangs; complementary overhangs anneal; a polymerase fills the remaining gaps and a ligase seals the nicks. The overlap you design is what becomes those complementary single strands — an enzymatically generated sticky end, not a primer.
How many fragments can one Gibson reaction join?
In routine practice a handful, and the reason to be conservative is combinatorial: every additional fragment adds a junction, success is set by the weakest one, and the chance of two junctions resembling each other closely enough to mis-pair rises with the count. If an assembly needs many pieces, staging it — assembling sub-parts and then joining them — is usually faster than optimising a single large reaction.
What should I sequence after a Gibson assembly?
The junctions at minimum, since that is where the method fails, and preferably the whole plasmid. Whole-plasmid nanopore sequencing has made full verification cheap enough that checking only the insert no longer makes sense — an assembly that closed correctly at both junctions can still carry a PCR-introduced mutation anywhere in a fragment you amplified.
Related references
Common restriction enzymes: recognition sites, cut positions, NEB buffer activity, star activity and an interactive double-digest buffer finder.
Nucleotide ambiguity codes and their complements.
Average and exact molecular weights for DNA and RNA sequences.
Related tools
Assemble fragments and design junction primers for Gibson, Golden Gate or restriction cloning.
Design ranked PCR primer pairs from a template, with Tm, GC and dimer checks.
Estimate primer Tm, GC% and molecular weight from a sequence.
Find recognition and cut sites for common restriction enzymes.
Scan a coding sequence for premature stops, cryptic RBS/polyA signals, unwanted restriction sites, GC extremes and repeats.
Align many NGS/Nanopore/Sanger reads to a reference with minimap2, call multi-read consensus variants, and build a corrected consensus sequence.