How to Design a Diagnostic Digest for Clone Screening
11 min read · Updated September 11, 2026
You do not design a diagnostic digest by finding an enzyme that cuts. You design one by asking whether the lane for the construct you want would still be recognisable beside every construct you might actually have.
That change of question matters. A one-cut enzyme can linearise all of your candidates into the same single band and tell you nothing, while an enzyme you would never have chosen from the intended map alone can separate an empty vector, a reversed insert and the correct clone at a glance. This guide shows how to define those alternatives, turn each digest into an observable band pattern, and choose a screen without pretending that a gel proves every base is right.
A diagnostic digest is a comparison, not a site search
A restriction-site finder answers a sequence question: where does each enzyme cut this molecule? A diagnostic digest answers an experimental decision: which enzyme gives the intended molecule an observable lane pattern that none of the alternatives shares? The first needs one sequence. The second cannot be designed honestly without at least two.
This is why "cuts once" is not a useful default. One cut in a circular plasmid produces one linear fragment equal to the plasmid's full length. If the correct construct and an incorrect construct are the same length, both become the same band. Even when their lengths differ, a single uncut or singly-cut band often says little about insert orientation or a bad junction.
A screen also has to cut the intended construct at least once. An enzyme that leaves it uncut may appear to distinguish it from a cut alternative, but an uncut lane is indistinguishable from a failed digest. Building a decision on the absence of enzyme activity turns a pipetting error into a passing clone.
Name the wrong constructs before choosing the enzyme
The alternatives define what the digest can prove. If you compare only the correct construct with empty vector, the winning enzyme proves only that the clone is not empty vector. It says nothing about a reversed insert, a one-part dropout or a tandem duplication you never supplied.
Start from the failure modes of the assembly you actually ran rather than a generic checklist. For a two-part restriction ligation, empty vector and both insert orientations are the obvious set when the ends are compatible. For a multi-part Gibson or Golden Gate assembly, useful candidates include an omitted part, an inverted part, a duplicated part, a pair swapped in order and backbone recircularisation. Only keep outcomes that are chemically possible for your design.
Do not attach invented probabilities to that list. A possible product is not a likely product, and measured Golden Gate overhang cross-talk does not directly predict colony counts after transformation. The purpose of the list is coverage: if this wrong molecule were in the tube, would the proposed lane expose it?
- Write down the intended full sequence, not only the insert.
- Build the full sequence of each plausible wrong product.
- Remove duplicates that are merely rotations or reverse-complement representations of the same circular molecule.
- Label every alternative by mechanism — empty vector, insert reversed, part 3 omitted — so a failed comparison tells you what remains unresolved.
Get topology right before reading any fragment sizes
A circular molecule cut at n distinct positions gives n fragments. A linear molecule cut at the same n positions gives n + 1, because it already has two physical ends. That one setting changes every predicted lane.
The start of a plasmid sequence is not a physical end. It is an arbitrary place where a circular record was written as text, so a recognition site can cross that join. A linear scan silently loses such a site and can turn a true two-cutter into an apparent single cutter. Treat clone-screening plasmids as circular unless the material on the bench was deliberately linearised.
The same fact gives a useful impossibility check. A plasmid rotated to start at another base, or reverse-complemented and rotated, is the same molecule. No restriction digest can tell those representations apart, and a planner should identify them as identical rather than searching forever for an enzyme that distinguishes them.
Fragments are not automatically visible bands
For each candidate enzyme, first compute the complete-digest fragments for every molecule. Their lengths must sum to that molecule's length; if they do not, the digest calculation is already wrong. Then convert those fragments into what a person would score in a lane.
Two fragments close in size can comigrate and appear as one band. Two different fragments with exactly the same size form one brighter band, not two positions. A very short fragment can run off with the dye front, and a fragment carrying only a small fraction of the plasmid's DNA mass can be too faint to call reliably. Comparing raw fragment arrays without this second step recommends screens that exist on paper and disappear on the gel.
Size differences must be judged as ratios rather than fixed numbers. Agarose migration is approximately linear with the logarithm of fragment length over the useful range, so 1.0 versus 1.2 kb is much like 5.0 versus 6.0 kb, while a 200 bp gap that is obvious near 1 kb can be invisible near 10 kb. This log-size relationship is the foundation of agarose separation, established in the early quantitative work of Aaij and Borst (Biochim Biophys Acta 1972;269:192-200, doi:10.1016/0005-2787(72)90426-1).
The gel rules are explicit conventions, not laws of agarose
A planning tool has to draw a line somewhere, so SeqBench states every line it uses. At 1.0% agarose, its tabulated working window is 500 bp to 10 kb. Two bands at least 1.20-fold apart are called clearly resolved, 1.10- to 1.20-fold is marginal, and below 1.10-fold they are treated as comigrating. A fragment below 30 bp is unscoreable on an ordinary screening gel, and a band carrying under 2% of the lane's total DNA mass is flagged as faint.
Those are deliberately conservative planning heuristics, not a claimed performance specification for your gel. Buffer, voltage, run length, comb, stain, imaging system and how quickly someone reads the plate all move real resolution. The upper end is especially unforgiving: two fragments both above the gel's exclusion range collect near the well and are treated as one blob even when their size ratio looks generous.
The agarose value used is reported because it decides these calls. If you request a percentage between the tabulated values, it is snapped to the nearest supported one — for example, 0.95% is evaluated as 1.0%, with the 500-10,000 bp window named in the result rather than hidden.
Worked example: correct insert, empty vector and reversed insert
SeqBench's built-in example is synthetic on purpose: a 1,400 bp backbone plus a 600 bp insert makes a 2,000 bp intended circular construct. The alternatives are the 1,400 bp empty vector and another 2,000 bp construct carrying the reverse complement of the insert. They are generated demonstration sequences, not a biological plasmid.
On a 1.0% gel, ApoI gives the intended construct fragments of 1,084 and 916 bp. It leaves the empty vector as one 1,400 bp fragment. Reversing the insert moves the internal geometry of the sites and gives 1,242 and 758 bp instead.
That is the property a diagnostic digest needs. Total length alone cannot separate the correct and reversed 2,000 bp plasmids, but their two-band fingerprints differ. The empty vector is a third, single-band pattern. ApoI is useful here not because it is a favourite enzyme or because it cuts a prescribed number of times, but because all three lane-level answers are distinct under the stated gel rules.
The example also shows why orientation screens need asymmetry. If the two insert orientations produce the same unordered fragment sizes, the gel cannot know which fragment came from which side of the insert. Move to another enzyme, use a pair, or screen one junction by PCR.
Prefer one enzyme; add a second only when it buys discrimination
A single-enzyme screen is cheaper to run and easier to interpret, with fewer partial-digest bands to mistake for real products. Search those first. If one enzyme separates the intended lane from every listed alternative, adding another cutter is complexity without new information.
A double digest is valuable when no single enzyme creates enough asymmetry — one cutter distinguishes the backbone state while the other splits the insert or one junction. SeqBench's automatic policy searches pairs only when no single enzyme separates everything; you can ask it to search pairs regardless when you have a specific reason.
The pair still has to work in one tube. The planner reports the best shared NEB buffer, the lower of the two tabulated activities, and whether their incubation temperatures agree; when no one-buffer route is suitable it recommends a sequential digest. These are compatibility facts from a static enzyme table, not a prediction of yield. DNA methylation, enzyme age, glycerol concentration and star activity remain bench conditions.
What a clean diagnostic lane does not prove
The calculation assumes complete cutting at every recognition site. A partial digest adds larger fragments made from adjacent pieces that failed to separate. Undigested supercoiled plasmid migrates by conformation rather than by base-pair length. Star activity adds cuts the sequence model did not request, and Dam or Dcm methylation can block a site that is perfectly present in the text. All four can create or remove bands without changing the clone.
The screen also knows only the alternatives you supplied. A clone can share the expected fragment pattern while carrying a point mutation, a small indel away from a cut site, or a different structural error that preserves the same fragment lengths. Position-based scoring deliberately ignores brightness, so a doublet made of two comigrating fragments is not treated as distinct from a singlet at that position.
A diagnostic digest is therefore a fast structural filter, not base-level verification. Use it to avoid sequencing obvious failures. Confirm the surviving clone across the junctions, or across the whole plasmid when the construct matters.
A repeatable design procedure
- Build the intended full-length molecule and every plausible wrong molecule as explicit sequences.
- Mark the topology correctly and remove alternatives that are merely another rotation or strand representation of the same circle.
- Restrict the search to enzymes you can actually run, but reject unknown names rather than interpreting "not tried" as "not useful".
- Compute complete-digest fragments for every candidate and check that each lane still sums to its molecule's length.
- Convert fragments to observed bands at the agarose percentage you will use, including comigration, the working size range and faint fragments.
- Choose the simplest digest whose intended pattern differs from every listed alternative without relying on a faint-only difference.
- Check buffer, temperature and methylation sensitivity before running it.
- Sequence the survivors; do not promote a matching gel pattern into a claim that every base is correct.
Frequently asked questions
What makes a good diagnostic restriction digest?
A good digest gives the intended construct an observable band pattern that every plausible wrong construct lacks, using an enzyme that actually cuts the intended molecule. The useful property is discrimination, not a prescribed cut count. It should work within the resolving range of the gel you will run and should not depend only on a very short or faint fragment.
Why is an enzyme that cuts my plasmid once usually a weak clone screen?
One cut turns a circular plasmid into one linear fragment equal to its total length. Two constructs of the same length therefore give the same band, regardless of insert orientation or junction sequence. A single cutter is excellent for linearisation and usually poor for identification; a diagnostic digest wants a fingerprint with enough asymmetry to separate the candidates.
Which wrong constructs should I include?
Only products your assembly could plausibly make, but include the ones that would change your decision: empty vector, both insert orientations for non-directional ligation, part omissions, duplications, inversions or swaps for multi-part assembly, and backbone recircularisation. The digest can exclude only the alternatives you provide, so leaving one out narrows the claim even when the gate passes.
Can a restriction digest tell which way round my insert is?
Yes when the site geometry is asymmetric enough that the two orientations produce different fragment-size sets. In the synthetic example, both orientations are 2,000 bp, but ApoI gives 1,084 + 916 bp for the intended orientation and 1,242 + 758 bp when the insert is reversed. If both orientations give the same unordered sizes, no reading of that gel can recover orientation; use another enzyme, a double digest or junction PCR.
How far apart do two DNA bands need to be?
Think in ratios, not a fixed number of base pairs. SeqBench uses conservative planning lines: at least 1.20-fold apart is clearly resolved, at least 1.10-fold but below 1.20-fold is marginal, and below 1.10-fold is treated as comigrating, with a downgrade outside the selected gel's working range. Those are explicit conventions, not a universal specification — voltage, run length, agarose, buffer and imaging all affect a real gel.
Should I use a single digest or a double digest?
Use one enzyme when it separates the intended construct from every alternative; it is simpler and produces fewer partial-digest bands. Add a second only when it creates discrimination no single enzyme provides, then check that both enzymes have a usable common buffer and compatible incubation temperatures or plan a sequential digest.
Why must I mark a plasmid as circular?
Because a circular molecule cut n times gives n fragments, while a linear molecule gives n + 1, and a recognition site can cross the arbitrary place where a plasmid file starts. Treating a plasmid as linear can lose that site and change a true two-cutter into an apparent single cutter, producing the wrong screening pattern without an obvious error.
Does a matching diagnostic digest prove my clone is correct?
No. It supports a structural claim against the alternatives you supplied. A point mutation, small indel away from a cut site, or an unlisted rearrangement can preserve the same band sizes, while partial digestion and methylation can change the physical lane without changing the sequence. Use the digest to triage clones, then confirm junctions or the whole plasmid by sequencing.
Related tools
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.
Pick two restriction enzymes and get the single buffer that runs both, each enzyme's activity in every buffer, and a straight answer on when to digest sequentially instead.
Predict restriction fragments and see the simulated agarose gel with a ladder.
Locate the exact direct repeats that let a construct recombine away the DNA between them, and build the shortened molecule you would actually recover.
Paste the recipe you ran and the sequence you got. It re-runs the assembly from the parts and diffs the whole molecule against your claim.