Previews, tracks, and what your project actually keeps
Most tools here draw their answer on the map, not just in the panel. Guides appear as their own ring, PCR products as amplicon lanes, protein domains as coloured blocks over their coding sequence, rare codons as marks along the gene.
None of that is part of your sequence. It is a prediction about your sequence, and the distinction matters the moment you export the file, hand it to a collaborator, or come back to it in six months: a plasmid map cluttered with last week's 3,857 candidate guides is not a record of anything.
So the rule across the whole workbench is one sentence:
What a tool draws is a view — remembered with your project, but never written into your record unless you explicitly keep it.
The three kinds of thing on your canvas
The middle one is the one worth understanding.
| Your record's annotations | A tool track | A highlight | |
|---|---|---|---|
| For example | Features from the file, or ones you drew | CRISPR guides, PCR amplicons, protein domains | The bases an optimization changed, a Tm profile, a digest's peptides |
| Drawn as | Features, in the feature track | Its own ring and lane | Colour over the sequence |
| Comes back when you reopen the project | Yes | Yes | No |
| In an exported GenBank | Yes | No | No |
| In the file's history / provenance | Yes | No | No |
| How it goes away | You delete it | Clear track, or switch it off | Close the panel |
A track is a standing answer — 3,857 candidate guides you want to look at again tomorrow — so it is remembered with the project. A highlight belongs to an open panel and one specific run, so it is not: restoring "the bases some optimization changed" with no result behind it would be a claim you could not check.
Neither is part of the record. That is the point of the middle column: you keep the view without the file quietly acquiring 3,857 features it never had.
How to keep something a tool found
Every tool that predicts something has an explicit "keep it" action, and that action writes a real annotation — the same kind of feature the file would have had if it arrived annotated.
| Tool | Draws | Keep it with |
|---|---|---|
| CRISPR Guide Design | Guides and cut sites, in their own ring/lane | Annotate the guides you selected |
| PCR / Fusion / Multiplex | Predicted amplicons and primer-binding sites | Open the product as a new project, or save the primers |
| Protein Domains | InterProScan matches over the coding sequence | Annotate the matches you want |
| Codon Optimizer | Edited bases and unresolved constraints | Open optimized as a new document, or download GenBank |
| Gibson / assembly | Homology junctions between fragments | Build the assembly — the product is a new record |
| Codon usage | Rare-codon positions and clusters | Export the report, or optimize the region |
| Protease digest, RBS, melting temperature, ICE | Peptides, initiation windows, Tm profiles, edit outcomes | Export the table |
The asymmetry is deliberate. Drawing is free and reversible; keeping is a decision, and a decision should take a click that you meant.
Turning a track on and off
Right-click the map and open Tracks, or use Editor Settings → Display → Visible tracks. Every lane the canvas can draw is listed there, including the ones tools fill — a track that is filled shows its count, so CRISPR guides (3,857) tells you what you are about to hide.
Hiding a track does not discard it. To discard it, use the tool's own Clear track control.
Autosave, and why a preview is not caught by it
Your project saves itself as you work — you never press Save to keep an edit. That makes the split above load-bearing rather than cosmetic: if overlays were ordinary edits, autosave would quietly bake every preview you ever left open into the record, and the file's history would show them as edits you made.
They are not, and this is checked rather than intended. The saved project is written without them — including when a preview is on screen at the moment the save happens, and including when you close the tab with the panel still open. Tracks are saved too, but under their own heading, which is why they come back on the map without ever appearing in the sequence.
There is one historical wrinkle worth knowing if you have old projects: builds before August 2026 wrote some overlays into the record itself. If you open an old project and find guides or amplicons sitting in the feature list, that is where they came from — and simply opening and saving the project now clears them out.
A picture you put on the stage becomes a file
Sending a result to the stage — the Show on the stage button in a section's header, or the tab a finished run opens for you — also writes it into your project, where it sits in the Documents rail beside the files you imported. You can open it, export it, or send it to someone, and it is still there when you come back tomorrow.
What gets written is the result, not a screenshot of it. Where the surface has a table behind the drawing, the table is the file: a virtual gel writes its fragments (lane, coordinates, length, the enzyme and overhang at each end, mass), a guide track writes its guides, a codon matrix writes its counts, and a phylogeny writes Newick rather than a drawing of a tree. A picture that has no table behind it — a dot plot, which is millions of marks — is written as the picture, and it is written at the resolution it was drawn at rather than blown up.
The name says which experiment it was. A tree built by neighbour-joining with aLRT support is
tree-nj-alrt; change the method or the support and the next file is named differently, because it
is a different experiment. Run the same experiment again and you get tree-nj-alrt-2, so the
counter always means "same question, asked again". Re-showing a result you have not changed writes
nothing at all: the file already in your project is that result.
Nothing is written for a surface whose result cannot be honestly serialised, and nothing is written when the run produced no result. An empty file with a plausible name would be worse than no file.
The one place a preview changes the bases
Codon Optimizer can show you the optimized sequence itself rather than an overlay on top of the original — the Showing on map control, set to Optimized.
That is a stronger kind of preview, because the whole map recomputes: translations, GC content and especially restriction sites, which is how you find out that an optimization quietly destroyed the site your cloning plan depends on. See Optimize a CDS for a worked example on a real plasmid.
While those bases are on screen:
- the document is read-only, so nothing you type can land on a sequence that is not the saved one;
- your project still holds the original, whatever the canvas is showing;
- switching back to Original costs nothing and never re-runs the solver.
To turn the optimized sequence into something real, use Open optimized — it becomes a new document, and your starting sequence is untouched.