How to design a Golden Gate assembly

Updated October 2026.

Golden Gate assembly succeeds or fails at the design stage. The enzyme, the four-base junctions and the primer tails are decided before anything goes in a tube, and a bad junction shows up two weeks later as a plate of wrong colonies. This guide covers each decision, the common mistakes, and a checklist to run before you order oligos. For cycling, enzyme ratios and bench troubleshooting, read our guide to optimizing Golden Gate assembly.

What is Golden Gate assembly?

Golden Gate assembly is a one-pot cloning method that uses a Type IIS restriction enzyme plus T4 DNA ligase to join several DNA fragments in a fixed order. Type IIS enzymes cut outside their recognition sequence, so each fragment is released with a short overhang whose sequence you designed, and the recognition site ends up outside the final construct.

That property drives everything else. The overhangs are four bases you pick, and two fragments join only when their overhangs are complementary. Since the correct product loses its recognition sites, digestion and ligation run in one tube at the same time, while wrongly religated fragments keep their sites and get cut again. Cycling pushes the mixture toward the construct you designed, with no gel purification between steps.

When should you use Golden Gate instead of Gibson assembly?

Choose Golden Gate for standardized parts, many fragments, or a combinatorial set; choose Gibson for a few long, one-off fragments whose junction sequences you want to leave untouched. Golden Gate fits:

Modular parts you reuse across projects (promoters, coding sequences, terminators, backbones).
Assemblies of four or more fragments in a defined order.
Scarless protein fusions, because the junction can fall inside a codon.

Gibson fits better when fragments are long PCR products or synthetic blocks used once, when the sequence contains Type IIS sites you would rather not edit, or when the junction can be any convenient overlap. Our Gibson assembly design guide covers that path, and the two combine well: build modules by Gibson, then compose them by Golden Gate.

Which Type IIS enzyme should you pick?

BsaI is the default for most Golden Gate work, BsmBI (or its isoschizomer Esp3I) is the usual partner enzyme for a second assembly level, and BbsI (or BpiI) is the common third option. All three leave four-base overhangs, so a design moves between them if the recognition sites change.

BsaI: recognizes GGTCTC and cuts one base away. The default for most part standards.
BsmBI / Esp3I: recognizes CGTCTC. Used at the second level of a hierarchical scheme so level-one junctions are not cut again.
BbsI / BpiI: recognizes GAAGAC. Useful when a sequence already carries BsaI or BsmBI sites you would rather not edit.

Pick the enzyme that cuts your parts the fewest times, and copy the exact tail layout from the supplier's specification rather than from memory.

How do you design the four-base fusion overhangs?

Fusion site overhangs are the four-base sticky ends that define part order. Every overhang in a reaction must be unique and must not pair with any other overhang in the mix. Three constraints do most of the work:

No palindromes. An overhang such as AATT or GATC is its own reverse complement, so a fragment ligates to a second copy of itself and to its neighbor in the wrong orientation.
No near-duplicates. Overhangs differing at a single base mispair at a measurable rate; keep at least two differences between any two overhangs in a reaction.
Balanced base content. Avoid all-A/T and all-G/C overhangs; one or two G or C bases per overhang is the usual compromise.

Where the junction falls matters too. Inside a coding sequence, place it on a codon boundary so the fusion stays scarless. Between parts, reuse the junctions your standard already defines, so parts stay interchangeable.

What is overhang fidelity, and how do you use published sets?

Overhang fidelity is the measured tendency of a set of overhangs to ligate only to their intended partners under real reaction conditions. Published fidelity data, from sequencing the ligation products of every overhang pair, shows that some combinations mispair far more often than base-pairing rules predict.

You do not have to measure it yourself. Groups including New England Biolabs have published tested, high-fidelity overhang sets for different assembly sizes; take the set matching the number of junctions you need and use it as given. Fidelity drops as junctions multiply, which is why larger assemblies are usually split into levels.

How do you domesticate internal Type IIS sites?

Domestication means removing the enzyme's recognition sites from inside a part so only the designed junctions are cut. Search every fragment, backbone included, for the recognition sequence on both strands, then change one base in each hit.

In a coding sequence, use a synonymous codon change, preferring a codon common in the host organism. In a promoter, terminator or origin, change a base outside any known motif. If a part is being synthesized anyway, fold the change into the order; if it exists as a plasmid, one overlap PCR or mutagenesis step fixes it, and the domesticated version becomes the stock part. A single missed site opens a fragment mid-assembly, and the failure looks like a low colony count with no obvious cause.

What do MoClo and other part standards give you?

MoClo (Modular Cloning) is a Golden Gate standard that fixes the overhang sequences for each part type in advance, so any promoter swaps for any other without redesigning junctions. These schemes are hierarchical: level zero holds parts in entry vectors, level one builds transcription units, level two combines those into multigene constructs, with alternating enzymes between levels.

The gain is reuse: once parts are in the standard, a new combination is a matter of choosing which tubes go in the reaction. The cost is upfront, because every part must be domesticated and cloned into an entry vector first. Standards pay off when a lab builds similar constructs repeatedly, and are overkill for one three-part clone.

How do you plan combinatorial and hierarchical builds?

A combinatorial Golden Gate design keeps the junction overhangs fixed and varies which part sits at each position, so every variant uses the same reaction layout. With several promoters at position one and several coding sequences at position two, the overhangs stay identical and the library is the product of the choices. Two practices keep large builds manageable.

Design the junction set once, for the largest number of positions, then reuse it for every variant and follow-up library.
Split deep assemblies into levels. Build four-part to six-part modules first, then join the modules with the second enzyme. Fidelity per reaction stays high, failed modules are caught early, and each module is sequence-verified once.

What should the primer tails look like?

A Golden Gate primer reads 5' to 3' as four to six spacer bases, the enzyme recognition site, any spacer the enzyme specification requires, the designed four-base overhang, then the region that anneals to your template. Without the spacer at the front, the enzyme has nowhere to bind at the end of the amplicon.

Design the annealing region like any cloning primer, typically 18 to 25 bases with a melting temperature near 60 °C, leaving the tail out of that calculation. Our primer design guide covers the rest.

What are the most common Golden Gate design mistakes?

The failures that recur are design errors, not bench errors:

An internal recognition site left in a part or the backbone.
Two overhangs in one reaction that differ by a single base.
A palindromic overhang that lets a fragment self-ligate.
Recognition sites facing outward, so they stay in the product instead of being cut away.
No spacer bases at the 5' end of the primer.
Reusing an old junction set without rechecking it against the new parts.
Too many fragments in one reaction when levels would have worked.

What is a quick Golden Gate design checklist?

Run this before ordering anything:

Confirm the enzyme, its recognition sequence and any required spacer.
Search every fragment, both strands, for that sequence and domesticate the hits.
Assign an overhang to each junction, from a tested set when possible.
Check that no overhang is palindromic and that no pair differs by a single base.
Confirm junctions inside coding regions keep the reading frame.
Write the primer tails in a fixed order: spacer, recognition site, overhang, annealing region.
Simulate the assembled product end to end and read every junction in it.
Save the junction set with the project so the next variant reuses it.

How do you design a Golden Gate assembly in TeselaGen Community (free)?

TeselaGen Community is a free edition of the TeselaGen platform with Gibson and Golden Gate assembly design, automatic assembly protocols, primer design, and Twist and IDT order forms.

Put the parts in the library. Import or create the sequences in the DNA sequences library. The cut site tools in the sequence editor show where a Type IIS enzyme cuts, which is how you find internal sites to domesticate.
Create a design. From Molecular Biology, open Design and start a blank design. Pick the combinatorial layout if any position has more than one option, or the list layout for one fixed combination.
Lay out the positions. Each column is one position in the construct: label it with a glyph, then insert the sequence or part. Extra rows in a column give that position several options, and the platform enumerates the combinations.
Set the assembly method. In the Operation tab of the right inspector, choose Golden Gate or MoClo. The method drives what is planned: overhangs, flanking Type IIS sites and the primers that add them.
Clear the validation tab. It flags junction and digest problems before you submit, including overhangs that do not form a unique ordered set and missing or internal Type IIS sites. The design's Golden Gate parameters, among them the four-base overhang length set by the enzyme, its recognition sequence and the extra 5' bases added to each piece, are visible and editable.
Submit for assembly. The platform generates an assembly protocol with the input sequences and parts, the primers and oligos to order, the PCR reactions, the digest fragments, the assembly pieces and the assembled constructs.
Save and order. Save the constructs and oligos back to the libraries, export the oligo list as CSV, or send sequences to the Twist or IDT ordering tools, which score them against the vendor's build constraints and produce an order form.

Hierarchical tree layouts and lab execution are on paid plans. For the wider design tooling, see the molecular biology toolkit. To run this flow on your own parts, start with the free Community edition.