Diagnostic Digest Planner — Pick the Enzyme That Screens Your Clones
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.
🌐 Nothing you paste is logged or stored — every tool is also callable via REST & MCP, and in bulk from the batch tools
Picking a screening enzyme by eye off a plasmid map answers the wrong question. What matters is not where an enzyme cuts your intended construct, but whether the lane it produces looks DIFFERENT from the lane the wrong constructs produce — and two bands only look different if they are far enough apart on the gel you are running. This planner does that comparison properly: paste the construct you designed and the ones you might have got instead (empty vector, insert in backwards, a part dropped from a Golden Gate reaction), and it digests all of them with every curated enzyme, works out which fragments would resolve at your agarose percentage, which are too faint to score, and which pile up outside the gel's resolving window — then ranks the enzymes by whether the intended construct ends up with a band pattern no wrong construct shares. If no single enzyme manages it, it searches double digests and checks whether one NEB buffer runs both. For a multi-part Golden Gate or Gibson assembly it will also generate the candidate wrong plasmids for you, as real sequences, ranked by how few mis-ligations each one needs.
Working on a whole construct rather than constructs you paste in? Open SeqStudio — the full editor, with live feature annotation, plasmid maps, restriction and primer panels, undo/redo, multi-document tabs and GenBank / SnapGene import and export.
How to use the Diagnostic Digest Planner tool
- 1Paste the construct you designed into Intended construct, and tick Circular if it is a plasmid.
- 2Add the constructs you might have got instead — or, for a multi-part assembly, open “Generate from assembly parts”, paste the parts in order and let it enumerate them.
- 3Set the agarose percentage of the gel you will actually run, and optionally restrict the search to the enzymes on your bench.
- 4Read the recommended digest, the lane-by-lane band pattern, and the note on any comparison that rests on a faint or unresolved band.
Frequently asked questions
How is this different from the Virtual Gel?
The Virtual Gel answers “what does this digest give me?” for a digest you have already chosen. This answers the earlier question — “which digest should I choose?” — by comparing several constructs at once and reporting whether their lanes are actually tellable apart. It is a choice between enzymes, not a simulation of one.
When are two bands ‘the same band’?
Migration on agarose is linear in log(size) over the gel’s resolving range, so bands separate on a size RATIO rather than a size difference: 1.0 vs 1.2 kb reads like 5 vs 6 kb, while 9.8 vs 10.0 kb reads like nothing. Bands 1.2x apart or more are treated as clearly separate, under 1.1x as one band, and the gap between as marginal. That 1.1x threshold is not invented — it is set by the ladders: the tightest adjacent rung pair any ladder here puts inside a working window is 1000 vs 900 bp on the 100 bp ladder, a ratio of 1.111, and a ladder is sold on the claim that its rungs are separately visible.
Why does it care about the agarose percentage?
Because the resolving window moves with it, and outside that window sizes stop being comparable. Above the exclusion limit every fragment piles just under the well — 12 kb against 20 kb on a 1% gel is one blob whatever the ratio — so a digest that looks discriminating on paper can be unreadable on the gel you actually pour. Below the lower limit resolution degrades rather than vanishing, which is handled separately.
Does it tell me how likely each wrong construct is?
No, deliberately. For Golden Gate it reports the MEASURED overhang cross-talk a mis-assembly would have to exploit, from published T4 ligation-count data, because that is what has actually been measured. Turning those counts into “4% of your colonies” would need transformation efficiency, partial digestion and vector background, none of which that data covers. A cross-talk count of zero means ‘not observed in that assay’, which is a detection limit rather than an impossibility.
Why does it only need to separate the RIGHT clone from the wrong ones?
Because that is the question a screen asks: is this colony the one I wanted? Telling two different failures apart from each other only says which way it went wrong, which is useful but never the job — so it is reported as a bonus and never allowed to outrank a digest that reads more cleanly. Ranking on expected information gain instead would have required a prior over how often each failure happens, and no such prior is available.
What does this NOT check?
That the alternatives you listed are the ones your reaction will actually produce — it screens the hypotheses it is given and does not invent them. It also assumes a complete digest: partial digestion, star activity and leftover supercoiled plasmid all add bands a clean digest would not show. Verdicts rest on band position only, so a doubly-bright band is never counted as distinguishable from a single one at the same size.
Can I run it from code?
Yes. diagnostic_digest and assembly_outcomes are both on the REST API and the MCP server, and they chain: assembly_outcomes returns a forDiagnosticDigest list you can hand straight to diagnostic_digest. Every calculation here is deterministic — no model, no fitted constants beyond the published ligation data — so the API returns the same answer for the same input every time.
More
Related tools
Score your 4-base Golden Gate / MoClo overhangs against real published T4-ligase ligation counts: the weakest junction, every risky pair, and how your set compares with a published one.
Locate the exact direct repeats that let a construct recombine away the DNA between them, and build the shortened molecule you would actually recover.
Work backwards from the size you measured: which pairs of priming sites on your template could make a product that big, and can their 3' ends actually extend?