Figures, reports & curation

Everything CRAIC knows about alignment uncertainty can now be turned into a publication-quality figure, a QC report, or a hands-on edit. All figures are drawn through one vector engine, so each exports identically to PNG, SVG, or PDF (and, for the alignment view, an interactive HTML page) — no extra software needed.

Exporting figures

File ▾ → Export figure ▸ offers:

Figure	What it shows
Alignment figure	The wrapped block alignment with residues faded where the column is unreliable — the trustworthy core stands out and ambiguous stretches recede. Includes a reliability strip and consensus.
Confidence report	A per-column reliability profile with the ambiguous hotspots shaded, plus a text QC summary (mean reliability, % high-confidence columns, the hotspot coordinates).
Sequence logo (selection)	An uncertainty-weighted logo: confident columns stand tall, ambiguous ones shrink into multi-letter stacks.
Homology arcs (explorer pair)	The two posterior-explorer sequences as tracks, with arcs linking residues that could be homologous — opacity ∝ posterior. A confident region is a tight braid; an ambiguous one fans out.
Co-occurrence map (selection)	How often each pair of residues ends up aligned across the ensemble of alternative alignments — the crowd's view of the region, complementing the single-model posterior.
Homology card (last-clicked residue)	The "cloud" for one residue: which columns it could occupy and with what confidence. Click a residue first to choose it.
Interactive HTML	A self-contained, shareable page: coloured residues, a reliability strip, sticky names, and hover tooltips.

Figures use the level you're viewing (nt / codon / amino acid) and, where relevant, the current column selection.

Navigating ambiguity

Edit ▾ → Go to next ambiguous region (Ctrl/⌘+G) jumps the selection to each reliability hotspot in turn, worst first — so you go straight to the columns that need a human instead of scrolling.

Comparing two alignments

File ▾ → Compare with alignment… loads another alignment of the same sequences and shows, as the column track, how often the two place residues together — red where they disagree. A natural extension of the multi-aligner agreement track to an external alignment (e.g. yours vs a collaborator's, or before/after editing).

Manual editing

The fastest way to curate is the keyboard, and it acts on the selected sequences only — so you can realign one (or a few) sequences without disturbing the rest. Click a cell to place the edit cursor (a bold amber box) and select that sequence; Cmd/Ctrl-click more cells or names to add them, or Shift-click to grab a contiguous range. Selected sequences slide together and stay fixed relative to one another. If you reuse the same set often, save it as a named sequence group (Edit ▸ Sequence groups) and re-select it in one click. Then:

Key	Action
← → ↑ ↓	Move the cursor
- (minus) or Space	Slide the selected sequences right (open a gap; a trailing gap is absorbed, so the alignment doesn't widen)
Backspace / Delete / Shift+Space	Slide the selected sequences left (consume the gap to the left, if every selected sequence has one)

Holding - or Backspace auto-repeats, so you can slide continuously. You can also drag a selected (amber) sequence with the mouse to slide it — one column per column dragged — or Alt/Option-drag any sequence to grab it directly.

The unselected sequences stay exactly where they are. Editing works at every view level: at the nucleotide level a slide moves one base, and at the codon or amino-acid level it moves a whole codon (three bases), keeping the reading frame intact. The view you're in stays put after each edit.

The same operations are also on Edit ▾ → Edit alignment (handy for discovery), all recorded on the provenance log and fully undoable:

Action
Undo / Redo	Ctrl/⌘+Z, Ctrl/⌘+Shift+Z — covers every edit, splice, sort, and degap.
Nudge residue left / right	Ctrl/⌘+, and Ctrl/⌘+. — slide the last-clicked residue into an adjacent gap; the status bar shows the column's support score change so you see whether the edit helped.
Insert gap column	Open a gap in every sequence at the selected column.
Delete empty column	Remove an all-gap column.
Pin selected columns (Ctrl/⌘ P)	Mark the selected columns as trusted. Pinned columns are drawn with a gold bar above the grid.
Unpin selected columns (Ctrl/⌘ ⇧ P) / Clear all pins	Remove pins.
Realign around pinned columns	Keep the pinned columns fixed and let the engine re-solve only the uncertain stretches between them. The pins move with their columns, so the blocks you trust stay pinned afterwards.
Mask selected residues (Ctrl/⌘ K)	Mark the residues under the selection (the selected sequences across the selected columns, or just the residue under the cursor) as unreliable. The column stays; the residues are drawn grey and hatched.
Unmask selected residues (Ctrl/⌘ ⇧ K) / Clear residue masks	Remove residue masks.
Mask residues below the threshold	Mask every residue whose reliability is below the Mask slider's value, keeping every column.

Masking residues instead of columns

Masking a column removes every residue in it, including the ones that are aligned correctly, and on a tree that costs more signal than the error it removes. A residue mask removes only what you judge wrong. Export masked… writes each masked residue as missing data — N for nucleotides, X for amino acids — which every tree program ignores without losing the rest of the column; if a column mask is also showing, both are applied.

A residue mask belongs to the residues, not the columns, so it follows them through edits and realignment, is saved in the session, and every change to it is on the undo stack and the History. On the command line, craic mask aln.fasta --residues -o masked.fasta masks residues below the threshold instead of dropping columns.

Manual edits to gap structure drop the coding annotation (a hand-made gap can break the reading frame); residue nudges keep it, since they don't move columns.

Column annotations

Label regions of the alignment — α-helices, active sites, domains, anything — as named, coloured column features. Select a column range (drag), then Edit ▾ → Annotations → Add annotation from selection… and give it a name. Annotations show in a labelled colour band along the bottom of the alignment.

• Fixed by position. An annotation marks a column range; if you later edit the alignment, it stays at those columns (it does not follow the residues).
• Select to highlight. Edit ▾ → Annotations → Select “…” highlights that feature's columns and scrolls to it.
• Export the regions only. Export annotated columns only… writes a new alignment containing just the columns covered by your annotations (the union).
• Saved with the alignment. Annotations live in the same sidecar file as sequence groups (<file>.craic.json), so the alignment file stays standard and the annotations come back when you reopen it.

Annotations and the conservation Mask slider are independent channels and can be used together: the slider dims low-scoring columns and exports by score, while annotations colour-band specific regions and export by region — pick whichever export you want from its own menu, so the two never conflict.